Research and Development in the Computer and Information Sciences

DOCUMENT RESUME

ED 040 727 LI 002 047 AUTHOR Stevens, Mary Elizabeth TITLE Research and Development in the Computer and Information Sciences. Volume 3, Overall System Design Considerations, A Selective Literature Revie4. INSTITUTION National Bureau of Standards (DOC), Washington, D.C. Center for Computer Sciences and Technology. REPORT NO NBS-monogr-113-Vol-3 PUB DATE Jun 70 NOTE 149p. AVAILABLE FROM Superintendent of Documents, U.S. Government Printing Office, Washington, D.C. 20402 (NO. C13.44:113, Vol 3, $1.25)

EDRS PRICE EDRS Price MF-$0.75 HC Not Available from EDRS. DESCRIPTORS *Computer Programs, Computers, *Information Networks, *Information Processing, Information Science, *Input Output, Literature Reviews, *Program Design, Programing Languages IDENTIFIERS *On Line Systems

ABSTRACT A selective literature review of overall system design considerations in the planning of information processing systems and networks. Specific topics include butare not limited to: (1) requirements and resources analysis, (2)problems of system networking, (3) input/output and remote terminal design, (4) character sets, (5) programming problems and languages,(6) processor design considerations, (7) advanced hardware developments, (8) debugging and on-line diagnosisor instrumentation and (9) problems of simulation. Supplemental notes anda bibliography of over 570 cited references are included. Parts 1 and 2 of thisseries on research and development efforts and requirements in thecomputer and information sciences are availableas ERIC documents: LI 001 944 and LI 001 945 respectively. (Author/NH) A UNITED STATES DEPARTMENT OF COMMERCE PUBLICATION NBS MONOGRAPH113,VOLUME3

°Pius0

ssfi

U.S. DEPARTMENT OF COMMERCE Volume Overall System Design Consideration National A Seledive Literatle Review Bureau of EN- Standards

U,S, DEPARTMENT OF HEALTH, EDUCATION (=) & WELFARE C\t OFFICE OF EDUCATION THIS DOCUMENT HAS BEEN REPRODUCED EXACTLY AS RECEIVED FROM THE PERSONOR ORGANIZATION ORIGINATING IT POINTS OF c=t VIEW OR OPINIONS STATED DO NOT NECES- SARILY REPRESENT OFFICIAL OFFICE OF EDU- CATION POSITION OR POLICY NATIONAL BUREAU OF STANDARDS

The National Bureau of Standards ' was established by an act of Congress March 3, 1901, Today, in addition to serving as the Nation's central measurement laboratory, the Bureau is a principal focal point in the Federal Government for assuring maximum application of the physical and engineering sciences to the advancement of technology in industry and commerce. To this end the Bureau conducts research and provides central national services in four broad program areas, These are:( 1 )basic measurements and standards, (2) materials measurements and standards, (3) technological measurements and standards, and (4) transfer of technology. The Bureau comprises the Institute for Basic Standards, the Institute for Materials Research, the Institute for Applied Technology, the Center for Radiation Research, the Center for Computer Sciences and Technology, and the Office for In formation Programs. THE INSTITUTE FOR BASIC STANDARDS provides the central basis within the United States of a complete and consistent system of physical measurement; coordinates that system with measurement systems of other nations; and furnishes essential services leading to accurate and uniform physical measurements throughout the Nation's scientific community, industry, and commerce. The Institute consists of an Office of Measurement Services and the following technical divisions: Applied MathematicsElectricityMetrologyMechanics--HeatAtomic and Molec- ular PhysicsRadio Physics 2Radio Engineering "Time and Frequency 2Astro- physics 2Cryogenics.' THE INSTITUTE FOR MATERIALS RESEARCH conducts materials research leading to improved methods of measurement standards, and data on the properties of well.characterized materials needed by industry, commerce, educationalinstitutions, and Government; develops, produces, and distributes standard reference materials; relates the physical and chemical properties of materials to their behavior and their interaction with their environments; and provides advisory and research services to other Government agencies. The Institute consists of an Office of Standard Reference Materials and the following divisions: Analytical ChemistryPolymersMetallurgyInorganic MaterialsPhysical Chemistry. THE INSTITUTE FOR APPLIED TECHNOLOGY provides technical services to promote the use of available technology and to facilitate technological innovation in industry and Gov- ernment; cooperates with public and private organizations in the development of technological standards, and test methodologies; and provides advisory and research services for Federal, state, and local government agencies. The Institute consists of the following technical divisions and offices: Engineering StandardsWeights and Measures Invention and Innovation Vehicle Systems ResearchProduct EvaluationBuilding ResearchInstrument ShopsMeas- urement EngineeringElectronic TechnologyTechnical Analysis. THE CENTER FOR RADIATION RESEARCH engages in research, measurement, and application of radiation to the solution of Bureau mission problems and the problems of other agencies and institutions. The Center consists of the following divisions: Reactor RadiationLinac RadiationNuclear RadiationApplied Radiation. THE CENTER FOR COMPUTER SCIENCES AND TECHNOLOGY conducts research and provides technical services designed..to aid Government agencies in the selection, acquisition, and effective use of automatic data processing equipment; and serves as the principal focus for the development of Federal standards for automatic data processing equipment, techniques, and computer languages. The Center consists of the following offices and divisions: Information Processing Standards--Computer Information Computer Services Sys- tems DevelopmentInformation Processing Technology. THE OFFICE FOR INFORMATION PROGRAMS promotes optimum dissemination and accessibility of scientific information generated within NBS and other agencies of the Federal Government; promotes the development of the National Standard Reference Data System and a system of information analysis centers dealing with the broader aspects of the National Measure- ment System, and provides appropriate services to ensure that the NBS staff has optimum accessibility to the scientific information of the world. The Office consists of the following organizational units: Office of Standard Reference DataClearinghouse for Federal Scientific and Technical Information "Office of Technical Information and PublicationsLibraryOffice of Public InformationOffice of International Relations.

Headquarters and Laboratories at Gaithersburg. Maryland, unless otherwise noted: mailing address Washington, D.C. 20234. Located at Boulder, Colorado 80302. Located at 5285 Port Royal Road, Springfield, Virginia 22151, NBS TECHNICAL PUBLICATIONS

PER!ODICALS NONPERIODICALS

JOURNAL OF RESEARCH reports National Applied Mathematics Series. Mathematical tables, Bureau of Standards research and development in manuals, and studies. physics,mathematics, chemistry, and engineering. Comprehensive scientificpapers give complete details Building Science Series. Researchresults,test of the work, including laboratory data, experimental methods, and performance criteria of building ma- procedures, and theoretical and mathematical analy- terials, components, systems, and r,tructures. ses. Illustrated withphotographs, drawings, and Handbooks. Recommended codes of engineering charts. and industrial practice (including safety codes) de- Published in three sections, available separately: veloped in cooperation with interested industries, professional organizations, and regulatory bodies. Physics and Chemistry Special Publications.Proceedings of NBS confer- Papers of interest primarily to scientists working in ences, bibliographies, annual reports, wall charts, these fields.This section covers a broad range of pamphlets, etc. physical and chemical research, with major emphasis Monographs.Major contributions to the technical on standards of physical measurement, fundamental literature on various subjects related to the Bureau's constants, and properties of matter. Issued six times scientific and technical activities. a year. Annual subscription: Domestic, $9.50; foreign, $11.75*. National Standard Reference DataSeries. NSRDS provides quantitive data on the physical Mathematical Sciences and chemical properties of materials, compiled from the world's literature and critically evaluated. Studies and compilations designed mainly for the mathematician and theoretical physicist.Topics in Product Standards. Provide requirements for sizes, mathematical statistics, theory of experiment design, types, quality and methods for testing various indus- numerical analysis, theoretical physics and chemis- trial products. These standards are developedcoopera- try, logical design and programming of computers tively with interested Government and industrygroups andcomputersystems.Shortnumericaltables. and provide the basis for common understanding of Issuedquarterly.Annual subscription:Domestic, product characteristics for both buyers and sellers. $5.00; foreign, $6.25*. Their use is voluntary. Technical Notes. This series consists of communi- Engineering and Instrumentation cations and reports (covering both otheragency and Reporting results of interest chiefly to the engineer NBS-sponsored work) of limited or transitory interest. and the applied scientist. This section includesmany Federal Information Processing Standards Pub- of the new developments in instrumentation resulting lications. This series is the official publication within from the Bureau's work in physical measurement, the Federal Government for informationon standards data processing, and development of test methods. adopted and promulgated under the Public Law It will also cover some of the work in acoustics, 89-306, and Bureau of the Budget Circular A-86 applied mechanics, building research, and cryogenic entitled, Standardization of Data Elements and Codes engineering.Issued quarterly. Annual subscription: in Data Systems. Domestic, $5.00; foreign, $6.25*.

TECHNICAL NEWS BULLETIN CLEARINGHOUSE The best single source of information concerning the The Clearinghouse for Federal Scientific and Bureau's research, developmental, cooperative and Technical Information, operated by NBS, supplies publicationactivities,this monthly publicationis unclassified information related to Government-gen- designed for the industry-oriented individual whose eratedscience and technologyindefense,space, daily work involves intimate contact with science and atomic energy, and other nationalprograms.For technologyfor engineers, chemists, physicists,re- further information on Clearinghouse services, write: search managers, product-developmentmanagers, and Clearinghouse company executives. Annual subscription: Domestic, U.S. Department of Commerce $3.00; foreign, $4.00*. Springfield, Virginia22151 Difference in price is due to extra cost of foreign mailing. Order NBS publications from: Superintendent of Documents Government Printing Office Washington, D.C. 20402 UNITED STATES DEPARTMENT OF COMMERCE Maurice H. Stans,Secretary NATIONAL BUREAU OF STANDARDS Lewis M. Branscomb, Director

Research and Development in the Computer and Information Sciences

3.Overall System Design Considerations A Selective Literature Review

Mary Elizabeth Stevens Center for Computer Sciences and Technology National Bureau of Standards Washington, D.C. 20234

"PERMISSION TO REPRODUCE THIS COPY- RIGHTED MATERIAL BY MICROFICHE ONLY HAS BEEN GRANTED BY

G P e TO ERIC AND ORGAN( TIONS OPERATING UNDER AGREEMENTS WITH THE U.S. OFFICE OF EDUCATION, FURTHER REPRODUCTION OUTSIDE THE ERIC SYSTEM REQUIRES PER- MISSION OF THE COPYRIGHT OWNER,"

National Bureau of Standards Monograph 113, Vol. 3 Nat. Bur. Stand. (U.S.) Monogr. 113 Vol. 3, 147 pages (June 1970) CODEN: NBSMA Issued June 1970

For sale by the Superintendent of Documents, U.S. Government Printing Office, Washington, D.C. 20402 (Order by SD Catalog No. C 13.44:113, Vol. 3), Price $1.25 Foreword The Center for Computer Sciences and Technology of the NationalBureau of Standards has responsibility under the authority of Public Law 89-306 (the Brooks Bill) forautomatic data processing standards development, for consultation and technical assistanceto Federal agencies, and for supporting research in matters relating to theuse of computers in the Federal Government. This selective literature review isone of a series intended to improve interchange of information among those engaged in research and development in the fields ofthe computer and information sciences. Considered inthis volume are the specificareas of overall system design considerations, including the problems of requirements analysis,system networking, terminal design, character sets, programming languages, and advanced hardwaredevelopments. Names and descriptions of specific proprietary devices andequipment have been included for the convenience of the reader, but completeness in thisrespect is recognized to be impossible. Certain important developments have remained proprietaryor have not been reported in the open literature; thus major contributors to key developments in the fieldmay have been omitted. The omission of any methodor device does not necessarily imply that it is considered unsuitable or unsatisfactory, nor does inclusion of descriptive materialon commercially available instruments, products, programs, or processes constitute endorsement.

LEWIS M. BRANSCOMB, Director

III Contents Page 1. Introduction 1 2. Requirements and resources analysis... 1 2.1. Requirements analysis 3 2.1.1. Clientele requirements 3 2.1.2. Information control requirements 3 2.1.3. Other system design requirements 5 a 2.2. Resources analyses 5 2.2.1.System,modularity, configuration and reconfiguration 6 2.2.2. Safeguarding and recovery considerations 6 3. Problems of system networking 7 3.1. Network management and control requirements 7 3.2.' Distribution requirements 8 3.3. Information flow requirements 8 4. Input-output, terminal design, and character sets 9 4.1. General input-output considerations 9 4.2. Keyboards and remote terminal design 10 4.3. Character set requirements 12 5. Programming problems and languages and processor design considerations 13 5.1. Programming problems and languages 13 5.1.1. Problems of veriy large programs and of program documentations 14 5.1.2. General-purpose programming requirements 14 5.1.3. Problem-oriented and multiple-access language requirements 16 5.1.4. Hierarchies of languages and programming theory 17 5.2. Processor and storage system design considerations 18 5.2.1. Central processor design 19 5.2.2. Parallel processing and multiprocessors 20 5.2.3. Hardware-software interdependence 20 6. Advanced hardware developments 21 6.1. Lasers, photochromics, holography, and other optoelectronic techniques 21 6.1.1. Laser technology 21 6.1.2. Photochromic media and techniques 22 6.1.3. Holographic techniques 23 6.1.4. Other optoeleetronic considerations 24 6.2. Batch fabrication and integrated circuits 25 6.3. Advanced data storage developments 26 6.3.1. Main memories 26 6.3.2. High-speed,special-purpose,andassociativeorcontent-addressable memories 27 6.3.3. High-density data recording and storage techniques 28 7. Debugging, on-line diagnosis, instrumentation, and problems of simulation 29 7.1. Debugging problems 29 7.2. On-line diagnosis and instrumentation 30 7.3. Simulation 31 8. Conclusions 33 Appendix A 35 Appendix B 129

List of Figures Page Figure 1. A generalized information processing system 2 Figure 2. Photochromt, ,iata reduction 23 ,V Research and Development in the Computer and Information Sciences 3. Overall System Design Considerations: A Selective Literature Review Mary Elizabeth Stevens

This report, the third in a series on research and development efforts and requirements in the computer and information sciences, is concerned with a selective literature review involving overall system design considerations in the planning of information processing systems and networks. Specific topics include but are not limited to: requirements and resources analysis, problems of system net-

2 working, input/output and remote terminal design, character sets, programming problems and languages, processor design considerations, advanced hardware developments, debugging and on-line diagnosis or instrumentation, and problems of simulation. Supplemental notes and a bibliography of over 570 cited references are included. Key words: Data recording; debugging; holography; information control; input-output; integrated circuits; lasers; memory systems; multiprocessing; networks; on-line systems; programming; simulation; storage. 1. Introduction This is the third in a planned series of reports Affecting all of the system design requirements for involving selective literature reviews of research specific functions of generalized information proc- and development requirements and areas of con- essing systems are those of hierarchies and inter- tinuing R & D concern in the computer and informa- action of systems, and of effective access-response tion sciences and technologies. In the first report,* languages; the client, system-configuration, and the background considerations and general purposes system-usage considerations (especially in terms of intended to be served by the series are discussed. multiple-access, time-shared systems), and of sys- In addition, the general plan of attack and certain tem evaluation, including such on-going "evalua- caveats are outlined.** tions" as debugging aids and on-line instrumented checking or monitoring facilities. In the first two reports in this series, we have been Under overall system design requirements, we concerned with generalized information processing are concerned with input-output capabilities and systems as shown in Figure 1, more particularly terminal display and control equipment, with proces- these reports were concerned respectively with in- sor and storage systems design, with advanced formation acquisition, and sensing, and input opera- technologicaldevelopments,withprogramming tions and with information processing, storage, and language requirements, and with problems of on- output requirements. In this report we will be con- line debugging, client protection, instrumentation, cerned with some of the overall system design con- and simulation. siderations affecting more than one of the processes First, however, let us consider some of the overall shown, such as programming languages, remote system design considerations involved in require- terminals used both for input and output, and ad- ments and resources analysis and in problems of vanced hardware developments generally. system networking.

2. Requirements and Resources Analyses The introduction of automatic data processing effective systems planning and implementation; it techniques has not changed the kind of fact-finding, has changed the degree, particularly with respect to analysis, forecasting, and evaluation required for extent, comprehensitivity,detail in depth, and questions of multiple possible interrelationships. *Information Acquisition, Sensing, and Input: A Selective Literature Review. ** Appendix A of this report contains notes and quotations pertinent to the running For example, a "single information flow" concept 2.1 text. For the convenience of the reader, notes "1.1" and "1.1" recapitulate some of the becomes realizable to an extent not possible before. considerations discussed in the first report. Appendix B provides a bibliography of cited references. On the other hand, distributed 2.2 and decentralized

1 1. Information Acquisition r

2. Information 7. Processing Sensing an I r- Client Service I Input Requests I I I I i-- 3. Preprocessing

1 I I I II 10. 9. Processing I Matching i . ..4 Specifications i I I I 6a. I b. I On-Line Stor- Direct I I- Storage age Oatputs I I

I 11. I earch a I I Selec- I tio I

I 12. Retrieval I I

I I 14.: Outputs I I I I I I I 15. Use and I Evaluation I I I I eMI 410=11 es=11. 011111M MOM 111, =NO Im

Legend: process flow feedback flow FIGURE 1.A generalized information processing system.

2 7.

systems also become more practical and efficient of information input, flow, processing, storage, re- because of new possibilities for automatic control trieval, and output are essential to effective system of necessary interactions. design. Davis in a late 1967 lecture discussed many A major area ofcontinuing.R & D concern with of the multifacetted problems involved in informa- respect to both requirements and resources analysis tion control in both system planning and system is that of the development of more adequate meth- use. The varied aspects range from questions of odologies." Nevertheless, the new business on the information redundancy in information items to be agenda of the national information scenethat is, processed and stored to those of error detection and the challenge of system networking offers new correction with respect to an individual item record possibilities for a meshing of system design criteria as received, processed, stored, and/or retrieved. that have to do with where and how the system is to Among these information control requirements be operated and with where and how it is to be used. are: input and storage filtering and compression; quality control in the sense of the accuracy and 2.1. Requirements Analysis reliability of the information to be processed in the system; questions of file integrity and the de. "aerate Requirements analysis, as an operational sine qua introduction of redundancy; problems of formatting, non of system design, begins of course with suitable normalization, and standardization, and error de- assessment of present and potential user needs. tection and error correction techniques. Elsewhere in this series of reports, some embar- More particularly, Davis (1967) is concerned with rassinglycritical commentaries with respect to problems of information control in a system with actual or prospective usage are selectively covered." the following characteristics: Assuming, however, that there are definitive needs of some specifiable clientele for processing system "1. It has several groups of users of differing services that can be identified, we must first attempt administrative levels. some quantifiable measures of what, who, when, "2. The information within, the system has im- where, and why, the information-processing-system- posed upon it varying privacy, security and/or service requests are to be honored." In particular, confidentiality constraints. improved techniques of analysis with respect to The information entering the system is of clientele requirements, information control require- varying quality with respect to its substantive ments, and output and cost/benefit considerations content; that is, it may be raw or unevaluated, are generally desired. h may have been subjected to a number of evaluation criteria or it may be invariant 2.1.1. Clientele Requirements (grossly so) as standard reference data. "4.The user audience is both local and remote. It is noted first that "lack of communication be- 45.Individual users or user groups have individ- tween the client, that is, the man who will use the ual access to the informathm contabed within system, and the system designer is the first aspect the system. of the brainware problem." (Clapp, 1967, p.3). Considering the potential clients as individual users "6. The information within the system is multi- of an information processing system or service, the source information." following are among the determinations that need (Davis, 1967, p. 1-2). to be made: 2.8 We may note first the problems of controls that 1. Who are the potential users? will govern the total amount of information that is to 2. Where are they located? 2.7 be received, processed, and stored in the system. 3. If there are many potential users, user groups, These may consist of input filtering operations 2.11 and user communities, how do needs for in- as in sampling techniques applied to remote data formation and for processing services differ acquisition processes 2.12 or in checking for duplica- zinong them? 2.8 tions and redundancies in the file 2.13 4. What are the likely patterns and frequencies of Other information control requirements with re- usage for different types of potential clients? spect to the total amount of information in the sys- 5. To what extent are potential clients both moti- tem relate to problems of physical storage access, vated and trained to use the type of facilities withdrawals and replacements of items to and from and services proposed? 2.9 the store, maintenance problems including questions of whether or not integrity of the files must be pro- However obvious these and other requirements vided (i.e., a master copy of each item accessible at analysis considerations may be, a present cause of all times) 214 provisions for the periodic purging of critical concern is the general lack of experimental obsolete items,2.13 revisions of the file organization evidence on user reaction, user behavior, and user in accordance with changing patterns of usage,2.18 effectiveness 2.10 response requirements,2.17 and requirements for display of all or part of an item and/or indications of 2.1.2. Information Control Requirements its characteristics prior to physical retrieval.2.18 Detailed consideration and decision-making with Another important area of information control is respect to controls over the quality and the quantity that of identification and authentication of material 3 entering the system, with special problems likely to copying a numerical data display, and accidental or be involved, for example, in the dating of reports. inadvertentmisspellings of names are obvious (Croxton, 1955). As Davis (1967) also points out, the examples. timeliness of information contained in the system With respect to errors introduced by transmission, depends not only on the time of its input bta also examples of R & D requirements and progress were upon the date or time it was recorded or reported cited in the first report in this series ( "Information and the date the information itself was originally ac- Acquisition, Sensing, and Input", Section 3.4). Two quired, including the special case of the "elastic further examples to be noted here include the dis- ruler" (Birch, 1966).2.19 Another typical problem is cussion by Hickey (1966) of techniques designed to that of transliterations and transcriptions between handle burst-type errors 2.26 and a report by Menk- items or messages recorded in many different haus (1967) on recent developments at the Bell Tele- languages.2.2° phoneLaboratories.227 For checking recording A crucial area of B. & D concern is that of the accu- and/or transmission errors, a variety of error de- racy, integrity, and reliability of information in the tection devices (such as input interlocks,2.29 parity system, although these questions are all too often information,229 cheek digits,23° hash totals,2.31 for- neglected in system design and use.2.21 Again, Davis mat controls and message lengths2.32) have been emphasizes the importance of information content widely used.2.33 controls. These may be achieved, on input, either by Problems introduced by alphanumeric digit trans- error-detecting checks on quantitative data or by positionsorsimplemisspellings can often be "correctness control through 'common sense' or attacked and solved by computer routines, provided logical checks." (Davis, 1967, p. 10.) 2.22 Thus, the that there is some sort of master authority list, or use of reliability indicators and automatic inference file, or the equivalent of this in terms of prior con- capabilities may provide significant advantages in ditional matching.2.34 For example, Alberga (1967) improved information handling systems in the discusses the comparative efficiency of various future.223 methods of detecting errors in character strings. One of the obvious difficulties in controlling accu- The use of contextual information for error detec- racy and reliability of the information content of tion and possible correction in the case of auto- items in the system is that of correction and up- matic character recognition processes has been dating cycles.2.24 More commonly, however, errors noted in a previous report in this series, that on affecting the accuracy and reliability of information information acquisition, sensing, and input. This is, are those of human errors in observation, recording, of course, a special case of misspelling.2.35 Some of or transcription and those of transmission or equip- the pertinent literature references include Edwards ment failure during communication and input. The and Chambers (1964), Thomas and Kassler (1967) incidence of such errors is in fact inevitable and and Vossler and Branston (1964). The latter investi- poses a continuing challenge to the system designers gators, in particular, suggest the use of lookup which becomes increasingly severe as the systems dictionaries specialized as to subject field and analy- themselves become more complex.2.25 sis of part-of-speech transitions 2.36 It is to bo noted, of course, that a major area of Context analysis is important, first, because for R & D concern in the communication sciences is the human such capabilities enable him to predict that of information theoretic approaches to error (and therefore skim over or filter out) message detection, correction, and control. In terms of gener- redundancies and to decide, in the presence of alized information processing systems, however, we uncertainties between alternative message readings, shall assume that advanced techniques of message the most probably correct message contents when encoding and decoding are available to the extent noise, errors, or omissions occur in the actual required, just as we assume adequate production transmission of the message.2.37 quality controls in the manufacture and acceptance Context analysis also provides means for auto- testing of, say, magnetic cores. Thus our concern matic error detection and error correction in the here is with regard to the control, detection, and input of textat the character level, the word (where feasible), correction, of errors in informa- level, and the level of the document itself such tion content of items in an information processing as the detection of changes in terminology or the system or network, regardless of whatever protec- emergence of new content in a given subject field. tive encoding measures have been employed. For example, "various levels of contextcan be It should be recognized first of all that any formu- suggested, ranging from that of the characters lation of an information-carrying message or record surrounding the one in question to themore nebu- is an act of reportage, whether it is performed by lous concept of the subject class of the document man or by machine. Such reportage may itself be being read." (Thomas and Kassler, 1963,p.5). in error (the gunshots apparently observed during In automatic character recognition, in particular, riot conditions may have been backfiring from aconsideration has been given to letter digrams, truck, the dial indicator of a recording instrument trigrams, and syllable analysis approaches 2.38as may be out of calibration, and the like). The record- well as to dictionary lookups. ing of the observation may be in error: misreading of, Special problems, less amenable to contextual say, the dial indicator, transposition of digits in considerations, arise in the case of largefiles 4

4,15thr. containing many names (whether of persons or rnents, their locations and the probable workloads of drugs, for example) which are liable to mis- (both as to types and also as to throughputs re- spellings or variant spellings or which are homo- quired), are the necessary analyses of the resources nyinous .2'39Information control requirements in presentlyorpotentiallyavailable.Resources such cases may involve the use of phonetic indexing analysistypicallyinvolvesconsiderationsof techniques 2.40aswellaserrordetection and manpower availabilities, technological possibilities, correction mechanisms.2.41 and alternative procedural potentialities. Automatic inference and consistency checks The question may well be raised with respect to may be applied to error detection and error cor- an obvious spectrum of R & D requirements. rection as well as to identification and authentica- Certainly there will be continuing areas of R & D tion procedures. Waldo and DeBacker (1958) give concern with respect to advanced hardware tech- an early example as applied to chemical structure nologies in processor and storage system design, data.2,42 A man-machine interactive example has and in materials and techniques that are related been described by North (1968).2'43 For the future, to these requirements. Next there are problems of however, it can be predicted that: "Ways must be "software" that is, of programming techniques found for the machine to freely accept and use to take full advantage of parallel processing capa- incomplete, qualitative information, to mix that bilities, associative memory accessing and organiza- informationwithinternally-derivedinformation, tion, multiprogrammed and multiple-access and to accept modifications as easily as the original system control. information is accepted." (Jensen, 1967, p. 1-1). Certain requirements are obviously overriding Finally, we note that, in its broadest sense, the because they permeate the total system design term "control" obviously implies the ability to and because they interact with many or all of the predict whether a given machine procedure will sub-systems involved. These include the problems or will not have a solution and whether or not a of comparative pay-offs between various possible given computer program, once started running, assemblies of hardware and software, the questions will ever come to a halt. The field of information ofprogramminglanguagesandofsuitable control may thus include the theories of automata, hierarchies of such languages, and the problems computability, and recursive functions, and ques- of man-machine interaction especially in the case tions of the equivalence of Turing machines to of time-shared or multiple access systems. other formal models of computable processes. Similarly, the requirements for handling a variety of input and output sensing modalities and for 2.1.3. Other System Design Requirements processing more than one 1-0 channel in an effectively simultaneous operation clearly indicate needs Other system design considerations with respect for continuing research and development efforts torequirementsanalysisincludequestions of in the design and use of parallel processing tech- centralizationordecentralizationoffunctions niques,multi-processornetworks,time-shared andfacilities,including compromises such as multiple access scheduling, and multi-programming. clusters; 2.44questionsofbatch-processingas against time-sharing or mixtures of these modes,2.45 Hierarchies of languages are implied, ranging and questions of formatting, normalization,2.46 and from those in which the remote console user speaks standardization.2.47 to the machine system in a relatively natural language (constrained to a greater or lesser degree) to A final area of requirements analysis involves those required for the highly sophisticated execu- the questions of system design change and modifica- tive control, scheduling, file protection, accounting, tion2.48 and of system measurement.2.49 In particular, monitoring, and instrumentation programs. For the information on types of system usage by various future, increasing consideration needs to be given clients provides the basis for periodic re-design not only to hierarchies of languages for using sys- of system procedures and forappropriatere- tems, but to hierarchies of systems as well .2.51 organization of files. Such feedback information There are, of course, concurrent hardware re- may also provide the client with system statistics search, development, and effective usage require- that enable him to tailor his interest-profile or ments in all or most of these areas. Improvements in search strategy considerations to both the available microform storage efficiency, lower per bit informa- collection characteristics and to his own selection tion-representation costs, communication channel requirements. As Williams suggests32" this kind utilization economies, improved quality of facsimile of facility is particularly valuable in. systems where reproduction and transmission of items selected or the client himself may establish and modify the retrieved, are obvious examples of directly fore- categories of items in the files that are most likely seeable future demands. Some of the above con- to be of interest to him. siderations will be discussed in later sections of this report. Here we are concerned in particular with 2.2. Resources Analysis resources analysis in terms of system modularity, configuration and reconfiguration, and with provi- Collateralwithcomprehensiveanalysesof sions for safeguarding the information to be handled potential system clienteles, their needs and require- in the system. 5 2.2.1, System Modularity, Configuration, and and the emergence of computer-communication Reconfiguration networks obviously imply some degree of both "x'oday, in increasingly complex informationproc- modularity and replication of components,pro- essing systems, there are typically requirements for viding thereby some measure of safeguarding and considerable modularity and replication of system recovery protection.2,62 An extensive bibliographic components in order to assure reliable, dependable, survey of proposed techniques for improving system and continuous operation. "52 The possibilities for reliability by providing variousprocesses for intro- the use of parallel processing techniques are reducing redundancy is provided by Short (1968).2,63 ceiving increased. R & D attention. Such techniques Protective redundancy of system components is, may be used to carry out data transfers simultane- as we have seen, a major safeguarding provision ously with respect 2"53 to the processing operations, in design for high system reliability and availa- to provide analyses necessary to convert sequential bility."' In terms of continuing R. & Dconcerns, processing programs into parallel-path programs however, we note the desirability of minimizing or to make allocations of system resources more the costs of replication 2.88 and the possibilities for efficiently because constraints on the sequence in development of formal models that will facilitate which processing operations are executedcan be the choice of appropriate trade-offs between risks relaxed.255 and coss.2'6° In terms of system configuration and reconfigura- Finally, there are the questions of resources tion, there is a continuing question of the extent of analysis with respect to the safeguarding of the desirable replication of input-output units and other information in the system or networkthat is, components or sub-assemblies. This may be particu- the provisions for recovery, backup, rollback, and larly important for multiple-access and multiple-use restart or repeat of messages, records, and files.7 systems.2.58 A particularly important systemcon- The importance of adequate recovery techniques figuration feature desired as a resource for large- in the event of either system failure or destruction scale information processing systems is that of or loss of stored data, can hardly be overesti- open-endedness.2.57 mated."8 Systemreconfigurations,oftennecessaryas The lessons of the Pentagon computer installa- changing task orders are received, are particularly tion fire, in the early days of automatic dataprocess- important in the area of shifting the system facilities ing operations, still indicate today that, inmany for system self-checking and repair.2'58 Thus Amdahl situations, separate-site replication of the master notes that "the process of eliminating and introduc- files (not only of data but also often of programs) ing components when changing tasks is reconfigura- is mandatory 2 °° Otherwise, the system designer tion. The time required to reconfigure uponoccur- must determine whether or not the essential rence of a malfunction may be a critical system contents of the machine-usable master files can be parameter," (Amdahl, 1965, p. 39) and Dennis and recreated from preserved source data.2.7° If the Glaser emphasize that "the ability of a system to file contents can be recreated, then the designer adapt to new hardware, improved procedures and must decide in what form and on what storage new functions without interfering with normal sys- media the backup sourcerecords areto be tem operation is mandatory." (Dennis and Glaser, preserved.27' 1965, p. 5.) In terms of system planning andresource analysis for information processing network design,we note 2.2.2. Safeguarding and Recovery Considerations the following questions: Can the network continue to provide at least A first and obvious provision for "fail-safe" (or, minimal essential services in thecase of one or more realistically, "fail-softly") 2.59 operation of an more accidental or deliberate breaks in the links? information processing system network is that of What are the minimal essential servicesto adequate information controls(for example, as be maintained at fail-safe levels? To what extent discussed above) on the part of all membersystems will special priorities and priority re-scheduling and components in the network 2.6° This require- be required? ment reflects, of course, the familiar ADP apho- Must dynamic re-routing of information flow rism of 'garbage in, garbage out'. Again, the total be applied, or will store-and-forward with delayed system must be adequately protected from inad- re-routing techniques suffice? vertent misuse, abuse, or damage on the part of There are known techniques for evaluating its least experienced useror its least reliable com- optimum or near-optimum paths through complex ponent. Users must be protected from unauthorized paths in the sense of efficiency (economic, work- access and exploitation by other users, and they also load balancing, and throughputor timeliness must be protected from the system itself, not only considerations). Can these techniques bere- in the sense of equitable management, scheduling, applied to the fail-safe or fail-softly requirements and costing but also in thesense that system ormust new methods andalgorithmsbe failures and malfunctions shouldnot cause intoler- developed? able delays or irretrievable losses.2.61 What are the fallback mechanisms at all levels Tie-ins to widespread communication networks and nodes of the system for: (a) specific failures 6 at a particular node, (b) breaks of one ormore for changing the systemresponse to internal change specific link(s), (c) massive failures, suchas the in direction or to external stimuli, of thelength of New York area power blackout? time necessary fora newcomer to be inserted into his assigned role in the system, of theredundancy, in general, with respect toareas of R & D backup, or alternatives availableat times of partial concern affecting safeguarding and recovery provi- or total system destruction, and so forth. Clearly, sions, we may conclude with Davis that "Rarely, there will beno adequately constructed system until if ever, are measurements made of theability such measures of effectivenessare understood and of the system to respond when partially destroyed incorporated into system design." (Davis,1964, or malfunctioning, of the length of time required p. 28).

3. Problems of System Networking

Steadilymountingevidenceofthenearly "a, Load sharing among majorcomputer cen- inevitable development of information-processing- ters.. system networks, computer-communication utili- "b. Data pick-up from remotetest sites (or from ties,and multiply-shared, machine-based, data airborne tests). Insome cases real-time proc- banks illuminates a major and increasingly critical essing and retransmission of resultsto the area of R & D concern. In this area, the problems test site would be desirable. of "organized complexity" 3.'are likely to be at "c. Providing access for Plant Ato a computer least an order of magnitudemore intractable than center at Location B. Plant A might havea they are today in multiprogrammedsystems, much medium-scale, small-scale,or no computer of less in those systems requiring extensiveman- its own. machine interaction. "d. Data pick-up from dispersed plants andoffices It is probable, in each of these three fieldsof for processing and incorporation in overall development, that there has been and will continue reports. The dispersed points might be in the to be for some time to come: (1) inadequate require same locality as the processing center, or ments and resources fact-finding and analysis,3.2 possibly as much as several thousand miles (2) inadequate tools for system design,3.3 and(3) away." (Perlman, 1961,p. 209.) the utter lack of appropriatemeans for evaluation in advance of extensive (and expensive) alternatives Three special areas of system network planning of system design and implementation.3.4 Certainly may be noted in particular. These are the areas of the problems of system networking will involve network management and control, of distribution those of priority scheduling and dynamic alloca- requirements, and of information flow requirements. tion and reallocation in aggravated form.3.5 More- over, the extensive prior experience in, for example, message-switching systems, is likely to be of rela- 3.1. Network Management and Control tivelylittlebenefitintheinteractivesystem Requirements network.3.6 Effective provisions for networkmanagement and In particular, the practical problemsof planning control derive directly from the basic objectivesand for true network systems in theareas of documenta- mission of the network to be established.First, tion and library services have scarcely begunto be there are the questions withrespect to the potential attacked." Nevertheless, the developmentof com- users of the system such as the following: puter-communicationsnetworkshasbegunto emerge as the result of some or all of the following 1. What are the objectives of thesystem itself? Is factors: it to be a public system, free and accessibleto all? 3." Is it to servea spectrum of clientele (1) Requirements for data acquisition and collec- interests, privileges, priorities, and different tion from a numbe),,. of remote locations." levels of need-to-know? Is it subject, (2) Demands for services and facilitiesnot readily in the available in the potential user's immediate provision of its services, to constraints ofna- locality. tional security, constitutional rights(assurance of protection of the individual citizen's rightto (3) Recognized needs to share data,programs and thesecurity, among other things, of his subroutines, work loads, andsystemre- "papers" from unreasonable searches and sources." In addition, varioususers may share seizures), laws and regulations involving penal- the specialized facilities offered byone or more ties for violation suchas "Secrecy of Com- of the other members of the network."° munications," and copyright inhibitions? Similar requirementswere considered by various 2. What are the charging and major members of the pricing policies, if aerospace industry as early any, to be assessed against differenttypes as 1961, as follows: of service, differenttypes of clients, and different priorities ofservice to the different "6. What members of the clientele? 3.12 are the budgetary constraints and financially allowable 1 What differentprotections may be built into rate of development?" the system for differentcontributors with (Davis, 1968,p. 4- 5). varying degrees of requirementsfor restric- The factors of geographicalcoverage, location tions uponaccess to or use of their data? 3'13 and facilitiesof participants, andmembership 4. Whatare the priority, precedence, and inter- point to some of thedistribution requirements, ruptprovisions requiredin terms of the to be considered next. clientele? a." Next are the questions,in terms of the potential client-market, of the location,accessibility, cost, 3.2. DistributionRequirements volume of traffic, andscheduling allocations for A major some determinate number ofremote terminals, area of concern with \respect to distri- user stations, and communication links. bution requirements ininformation processingnet- work planning is that ofthe question of thetype Then there are the questionsof the performance and extent of centralization and technologicalcharacteristics required with or decentralization of the varioussystem functions. There isfirst the respect to these terminals, stations,and links.3,15 possibility of Are the centralsystem and the communication a single master, supervisory, and networkboth capable control processingcenter linkedto many geo- of handling,effectively graphically dispersed satellite simultaneously, the number ofindividual stations centers (which carry or links required? Does the communication out varying degrees ofpreprocessing and post- system processing of the informationhandled by the itself impose limitationson bandwidths available, central system) and datatransmissionrates,number of channels terminals. Secondly, several operable interconnected but independentprocessors may effectivelyinparallel?Are alternate interchange control and transmission modes available inthe event of channel supervisory functionsas usurpation or nonavailability for workload and otherconsiderationsdemand.3.18 other reasons? Still another possibilityis regional centralization Is effectively on-lineresponsiveness of the communi- such has been recommended cation system linkages requiredand if so to what for a national docu- extent? mentation network, for example.") Different compromisesin network andsystem More generally, the followingdesign and planning design to meet distribution questions should be studiedin depth if there is requirements are also to be effective management and obviously possible.g"However, a variety of special control: problems may arise withrespect to distribution "1. What is thescope of the network? requirements whensome of the network functions a. Its geographical coverage are decentralized."' b. Services to be providedby and to whom Then there is thequestion of whetheror not c. Location and facilities of participants the network isto be physically distributed that d. Existing capabilitiesavailable is, "the term 'distributednetwork' is best usedto e. Required rate of development delineate those communicationsnetworks based "2. What are the relevantsoftware and data on connecting each station to alladjacent stations, characteristics? rather than to justa few switching points,as in a a. Privacy requirements centralized network." (Baran,1964, p.5). This b. Accessibilityand/oravailabilityof distribution requirementconsideration isclosely program services related to informationflow analysis and planning, c. System managementprograms especially withrespect to assurance of continuing "3.What are the networkmanagement and productive operation whencertain parts of the control requirements? network are inoperative.3.22It should be noted, a. Standardization 116 moreover, that "solving the data basemanagement b. Membership problem has been beyondthe state of the art." c. Information and program manipulation (Dennis, 1968,p. 373). d. Feedback 3.17 e. Documentation f. Cost of services 3.3. Information Flow "4. What are the pertinentlegal regulations and Requirements practices? In general, it a. FCC regulations may be concluded that "to determine the correct configuration,certain basic factors b. Carrier ratestructure must be investigated. c. Common carrier use These factorsgenerally relate to the informationflow requirements and d. Responsibilities forinformation content include the following: e. Privacy versus broadcast methods 1. The kind of information f. Federalagency jurisdiction... to be transmitted "5. through the communicationsnetwork and the What are the technologicalconstraints? types of messages. 8 2. The number of data sources and pointsof tions with respect to information flow requirements distribution to be encompassed by the network typicallyinvolvecalculationsof average daily and their locations. volume of message and data traffic, peakloads 3. The volume of information (in termsof mes- anticipated, average message length, the numberof sages and lengthsof messages) which must messages to be transmittedin given time intervals, flow among the various locations. total transmission time requirements, and ques- 4. How soon the information must arrive tobe tions of variable duty cycles for different system useful. What intervals the information is to and network components.3.23 be transmitted and when. How mur,,hdelay Examples of relatively recent developments in is permissible and the penalty for delays. this area include RADA (Random AccessDiscrete 5. The reliability requirements with respect to Address) techniques 3'24 and a "hot-potato" routing the accuracy of transmitted data, or system scheme for distributed networks.3.25 A continuing failure and the penalty for failure. R & D challenge in terms of scientific andtechnical 6. How the total system is going to growand the information services has been posed by Tell (1966) rate of growth." by analogy with the techniques of input-output economics.3.26 Of major concern is the problem of (Probst, 1968, p. 19). high costs of communication facilities necessary to More specifically, overall system design considera- meet network information flowrequirements."'

4. Input-Output, Terminal Design, andCharacter Sets

The area of input-output, especially for two- without the need for elaborate or inefficient pro - dimensional and even three-dimensional informa- grammidg and related software requirements." tion processing, is currently receivingimportant The overalloutput capability design should emphasis in overall information processing sys- provide ability to reformat conveniently and effi- tem design. One reason forthis, as we have seen, ciently 4.4 as well as to select certain character is the increased attention being given toremotely sequences. Because ofthe variety of equipments accessed, time-sharing, or man-machine interaction needed for various tasks, provision should be made systems. In particular, as notedby Tukey and Wilk: for reversal of the bit order of input and output "The issues and problems of graphical presentation data so that either high or low order bits canbe in data analysis need and deserve attentionfrom processed first. In the case of displays, special many different angles,ranging from profound psy- provisions may be required to prevent overlapping chological questions to narrow technological ones. of symbols." These challenges will be deepened by the evolution Related to format control is the question of vari- of facilities for graphical real-timeinteraction." able byte size for input and output. For the future, (Tukey and Wilk, 1966, p. 705). system design will require ASCII (AmericanStand- ard Code for Information Interchange) code sorting 4.1. General Input/Output Considerations and ordering capabilities, but in many circumatances it will also be necessary to handle collapsed sub- Since a multiplicity of input and output lines are sets of ASCII and other codes,longer byte lengths assumed for a variety of types of information to such as 10- and 14-bit codes for typesetting, and be processed (including feedback informationfrom even longer codes for monotype,numeric process users and from the systemitself), development re- control,datalogging, and equipmentcontrol. quirements with respect to both equipmentand Analog-digital anddigital-analogconvertibility sortware processing operationsinclude batching of is needed for experimental applications in source various input units, buffering of atleast some data automation, measurements automation, map types of input (as required, forexample, to provide analysis, map and contour plotting, pattern process- necessary reformatting),and multiplexing of input ing, and the like. One example of convergentefforts operations. Such considerations also apply even in the field is provided by Ramseyand Strauss more forcefully to interfacesbetween the various (1966) who discuss interrupt handling in the area nodes of a network involving more than one typeof of hybrid analog-digital computers as representa- participating system." tive of more general on-line schedulingproblems. Format control istypically needed both into For some of these investigations, at leastvirtual and out of the system, preferably under dynamic real-time clocks will be needed.'This implies program control. The formatcontrol subsystem, processor main frame and transfertrunks versatile by means of address storage registers or other enough to handle these requirements whether im- techniques, should enable the input data itself plemented by software or built into the hardware. to determine where it should go in storage,and other Another important requirement is for versatile means of "self-addressing"should be provided and varied graphic input and outputcapability, 9 includinglightpen,microfilm,FOSDIC-type data, and not to display only the data needed for the scanning, mark-sensing, OCR (Optical Character decisions the system is called upon to make." Recognition), MICR (MagneticInkCharacter (Fubini, 1965, p. 2). Recognition), color-code input (such Lovibond In general, the client of the on-line, graphical color network), and three-dimensional ,robe data input-output, and problem-solving system needs in (see the first report in this series), and large- convenient means for the input of his initial data, vocabularycharacterandsymbolgeneration; effective control of machine processing operations, diagram retrieval, construction and reconstruction, effectivelyinstantaneous system response,dis- and perspective or three-dimensional projection plays of results that are both responsive to his capabilities out (as discussed in the second report needs and also geared to his convenience, and in this series). Photographic and TV-type input handy means for the permanent recording of the and output with good resolution, hard-copy re- decisions and design choices he has made. production capability, varying gray-scale facility, With respect to these client desiderata, the identi- and at least the possibility of handling color input fiable R & D requiNatents relate to keyboard or output display techniques will be required in function key overlay design; 4.14 improvements in future system design.4.7 Audio input-output capa- both problem-oriented and client-oriented languages bilities should include dataphone, acoustic signal for man-machine communication and interaction;4.15 inputs, and voice, with speech compression on fast,high-resolution,flicker-free display genera- output, requiring controlled timing of "bursts" or tion; 4.16ability to selectively emphasize various "slices." areas of display,417 furtherdevelopment of the com- In many system design situations, we should bination of static displays (such as maps) with com- be able to switch peripheral equipment configura- puter-controlled dynamic displays, and rapid tions around for special purposes and we may responsivity of the system to feedback from the need to have multiple access to various types of client. peripheraldevicessimultaneouslyduringthe Since remote, reactive terminals are an in- same processing run, e.g.,to be able to shift creasingly important factor in systems involving between character recognition and graphic scanning dynamic man-machine interaction, the question tasks for input of material where text and graphics of design of remote inquiry stations and consoles are intermixed. necessarily raises problems of human engineering Related to these problems of input, output, and for whose solution there is inadequate experimental on-line responsiveness (especially for clients incost-benefit, and motivational data 4.19available volvedinproblem-solvingapplications),isthe to date. Also involved are questions of acceptance concept of graphical communication generally. and interactive response by the client to feed- This presupposes, first, a suitable language for the back outputs from the system, including requests exchange of both pictorial data and control infor- for further information or additional inputs and mation between the designer and the machine, and display of re-processed results. secondly, provisions for the dynamic manipulation of data and controls." 4.2. Keyboards and Remote Terminal Design Recent programming techniques under investi- Where graphic input and output facilities are to gation for the display of two-dimensional structure be available to on-line users, there are unresolved information are exemplified in work by Forgie,4.1° questions of interrelated and interlocking system by Hagan et al. (1968) 4.11, and at the University of and human factors. How clumsy are light pens Michigan (Sibley et al., 1968).4.12 Then there is the or pointers to use? Are they heavy or difficult to DIALOG programming system developed at the ITT aim? 4.20Should light-pen imputs be displayed Research Institute in Chicago for graphical, textual, a little to the left or to the right of the actual light- and numeric data input and display, online and off- pen location so that the active part of the input line programming facilities, and hard-copy options. is not blocked from view by the moving light-pen (Cameron et al., 1967). A special feature is a char- itself? 4.21 Can flicker-rate be kept to a tolerable acter-by character man-machine interaction mode, level without undue and costly regeneration de- sothatthe programmer may use only those mands on a multiply-accessed central processor input symbols that are syntactically correct. For used by the many clients, or must the remote more efficientmachine useinproduction-type terminal have storage and display re-generation operations,a DIALOG compiler forthe IBM capabilitiesat added cost and design complex- 7094 has been prepared following the "Trans- ity? 4.22For graphicinput and display should mographer" of McClure.4.13 (McClure,1965). the input surface be flat, upright, or slanted? 4.23 Then we note that "in the area of displays, de- It has been pointed out, in the case of the recent termining the information to be displayed and development of a solid state keyboard, that "the generating the procedures for retrieval and format- requirements of today's keyboards are becoming ting of the information are the difficult problems." more complex.Increasedreliabilityand more (Kroger,1965,p.269).Further,as of today, flexibility to meet specialized demands are essential. "too many systems are designed to display all the Remote terminals are quite often operated by 10 relatively untrained personnel and the keyboard terms of desired display and console improve- must be capable of error-free operation for these ments.4.30 people. At the same time it should be capable The desirable design specifications for remote of high thru-put for the trained operator as will be inquiry stations, consoles, and terminals and dis- used on a key tape machine. play devices as discussed in the literature variously "Some of the limitations of existing keyboardsare: include: economy, dependability, and small enough size for convenient personal use.4.34 Some misgiv- Mechanical interlocks which reduceoperator ings continue to be expressed on thisscore. Thus, speed. it is reported that Project Intrex will consider the Excessive service (increasingly important for design of much more satisfactory small consoles4.32 remote terminals). and Wagner and Granholmwarn that "at the mo- Contact bounce and wear of mechanicalment, itisdifficultto predict whether remote switches. personal consoles can be economically justified to Non-flexible format." (Vorthmann and Maupin, the same extent that technological advances will 1969, p. 149). make them feasible." (1965, p. 288). Cost certainly For automatic typographic composition applica- appears to be a major factor in the limited nature tions,itis emphasized that "the application of of the use that has been made of remote terminals computerstotypesetting only emphasizesthe to date.4.33 scope and the need for a radical re-thinking on A second requirement is for the provision of keyset design," and that, although "onemay ima- adequate buffering facilities including, for at least gine that the keyboard is a relatively simple piece some recent systems, capabilities for local display of equipment. ..in fact,it presents a unique maintenance 4.34 From the hardware standpoint, it combination of mechanical, electrical and human is noted that "the major improvements in displays problems." (Boyd, 1965, p. 152). Current R & D will be in cost and in the determination and imple- concerns with respect to keyboard redesign in- mentation of the proper functions from theuser volve consideration of principles of motion study standpoint. The cathode-ray tube will probably as applied to key positioning, key shape, key be dominant as the visual transducer for console pressures required, and the like.4.24 displays through 1970, but there are severalnew Nevertheless, itis to be emphasized that "in- techniques for flat-panel, digitally addressed dis- put-output devices are still largely the result of an plays presently under development thatmay even- ingenious engineering development anda some- tually replace the CRT in many applications. what casual and often belated attention to operator, The advances in memory and logic component system attachment, and programming problems" technologolies will permit significant improvements and that "...no input-output device, including in the logic and memory portions of console dis- all terminals combined, has yet received thecare- plays." (Hobbs, 1966, p. 37). ful and competent human factors study afforded Other features that are desirable may includea the cockpit of a military aircraft."4.25(Brooks, capabilityforrelativelypersistentdisplay,for 1965, p. 89). example, up to several hours or several days,4.35 and the capability, as in the Grafacon 1010 (acom- Beyond this are questions of design requirements mercially available version of the RAND Tablet) for dynamic on-line display. Thus we arecon- for the tracing of material such as maps or charts to cerned with requirements for improved remote be superimposed on the imput surface,or the Syl- input console and terminal design.4.28 Relatively vania Data Tablet ET 1, which also allows a modest recent input-output terminal developments, espe- third axis capability. cially for remote consoles or dynamic man-machine As in the case of system outputs generally, interaction, have been marked by improved po- hard-copy options are often desired through the tentialities for two- and even three-dimensional terminal device. For example, the console "should data processing and by further investigation ofhave a local storage device on which theuser prospects for color, as discussed, for example, by can build up a file of the pieces of information he Rosa (1965),4.27 Mahan (1968) 4.28 and Arora et al. is retrieving, so that he can go back and forth in (1967), among others. Van Dam (1966) has provided referring to it. It should have means of giving him an informative state-of-the-art review of such scanning and input/output techniques. Vlahos (1965) low-cost hard copy of selected material he has been considers human factor elements in three-dimen- shown and temporarily stored." (King, 1965, p. 92). sional display. Ophir et al. (1969) discuss computer- The use of markers and identifiers for on-line generated stereographic displays, on-line.4.28 text editing purposes should be simplified or eliminated to the maximum extent possible. If, forex- In the area of input-output engineering andsys- ample, elaborate line and word sequence identi- tem design, what is needed for more effective man- fications must be used both by the machine system machine communication and interaction will include and by the client, then the virtues of machine the provision for remote consoles that are truly processing for this type of application will be largely convenient for client use. Hardware, software, and lost. Such systems should not only be easy to use, behavioral factors are variously interrelated in but easy to learn how to use. I1

376-411 0 - 70- 2 An important questionto be asked by the sys- processing system design anduse, and especially tem designer is whether the outputresponses will in information selection and retrievalapplications. be of the types and in the formatsthat the client More specifically: "The majorproblem today in will expect to receive. It is noted inparticular that the design of displaysystems is that we cannot "the closer the correspondence of thecomputer output is specify in more than qualitativeterms such critical to the methods of presenting textual criteria as 'context' and `meaning'."(Muckier and and tabular material familiarto the user, the greater Obermayer, 1965,p, 36). Swanson adds: "Other his information absorptionrate will be." (Morenoff restrictions derive from the need ofprograms to and McLean, 1967,p.20). Thus, in engineering solve hidden line problems,to recognize context, applications, for example, the clientmay want and to make abstractions." (Swanson,1967, p. 39). results to be displayed ina familiar format such Finally, we note the specific problemin documen- as a Nyquist plot.4'36 Similarly, in operationson tary and library applications that large character files or data banks, theuser should be able to struc- repertoiresare important to input and output ture and sequence files and subfiles for display representations of various levels of referenceand and selection to suit hisown purposes.4.37 emphasis in technical texts (especially,for example, For effective online operation, thesystem should in patent applications with multilevelreferrals to provide responses within the readingrates of typical accompanying drawings and diagrams) andto de- users, and with good resolution, little or no flicker, lineation of different types of possible and with both access points upper and lower case 4.38 A some in indexes and catalog cards.In addition,a wide what more specific and stringent list ofremote termi- variety of special symbols and/orexotic alphabets nal desiderata is provided by Licklider,with par- are typically employed in texts dealing primarily ticular reference to the requirements ofmultiple with mathematical, logicalor chemical subjects. A access to the body of recorded knowledge. His de- text written principally in one particular language sired features include, butare not limited to, and alphabetmay frequently use the alphabet set color, oratleast gradations ofgray scale; 4'39 of one or more other languages,as in reference to terse or abbreviated modes of expressionto the proper names, citations, and quotations. machine with full or "debreviated"response;4.40 selective erasibility of the display by eitherprogram 4.3. Character Set Requirements or user command, 4.41 and capabilities suchas the following:"Shortly thereafter, thesystem tells Formultiple-accesssystems "mostcreative me: 'Response too extensive to fit onscreen. Do users want large character sets with lower-caseas you wishshortversion, multipage display,or well as capital letters, with Greekas well as Latin typewriter-only display?'." (Licklider, 1965,p. 50). letters, with an abundance of logicaland mathe- Another design criterion affected by thefactor of matical signs and symbols, and with allthe common client convenience is that of theextent of display punctuation marks." (Licklider, 1965,p. 182). In on the console face of meta-information.4.42 addition, subscripts, superscripts anddiacritical Continuing needs for technologicaldevelopments marks may be required.4.52 have also been indicated for improvedterminal Attacks on problems of character-setrequire- and output display design. Examplesinclude the ments for output begin with on-line printervaria- development of new, fast phosphorsand other tions to provide larger output character-setvocabu- materials,4.43 use of analog predictivecircuitry to lariesatthe expense both ofoutput printing improvetracking performance,'" and variable speed and of prior input precedence-codingand/or sequencing of processor operations incomputa- ofprocessorprogramming.Largercharacter- tion and display.4.45 A number ofadvanced tech- set vocabularies are provided both by photocom- niques are also being appliedtolarge-screen position techniques and byelectronic character- displays,4.46 althoughsome continuing R & D dif- generation methods, but againat the expense of ficulties are to noted.4.47 Multiplexingof graphic either pre-codingor programming requirements 4.53 display devicesmay also be required.4.48 It should be noted, ofcourse, that the internal Returning, however, to the humanbehavioral language of most general-purposeinformation proc- factors in input-output and terminaldesign, we note essing systems is limitedto no more than 64 dis- that man-machine relationshipsrequire further cretecharacters, symbols, andcontrol,codes. investigation both from the standpoint ofhuman Thus there must be extensiveprovision for multi- engineering principles and also from that ofattitudes ple table lookups and/orfor decoding andre- of clients andusers 4.49 that there are continuing encoding of precedence codesor transformation requirements for research and developmentefforts sequences, on both input and output, ifany internal on both sides of the interface 4.50 and that, in all manipulations are to be performedon the textual probability, "industry will requiremore special material. In general, the largerthe character set, prodding in the display-controlarea than in the the more elaborate the precodingand/or program- other relevantareas of computer technology." ming efforts that will be required. (Licklider, 1965, p. 66). Then there is the problem ofsetting up key- We note furtheran area of R & D concern that board character sets that will recur in many other are adequate for applica- aspects of information tionrequirementsandyetwithinreasonable 12 human engineering limitations. Oneproposed solu- designed to facilitate the input ofa variety of lan- tion, the use of keyboard overlays andcontrol codes guages and diacriticals." (1965, p: 4). These authors to enable rapid shifting fromone character subset point out further that "if the problems toanother,is associated exemplified by developmentsat with the design of input keyboardsand photocom- Bunker-Ramo.4.54 Another possible solutionto the position printing devices character set problem can be resolved for the as related to human engi- multiplicity of alphabets, there stillwill remain neering factors that is receivingcontinuing R & D the formidable task of searchingin a machine file concern is to provide multiple inputs viaa single which contains them." (Avramet al 1965, p. 89). keystroke, suchas "chord" typewriters or Steno- Similarly, Haring (1968) points type devices.4.55 out that the 128 Regardless of what symbols provided in the ASCII codeis inadequate may be available through for the augmented catalog underdevelopment at various conversion, transliteration,or translation Project INTREX. '59 processes on either input or output, there remains the s.,,,estion of the effects ofinternal character The very number anddiversity of varied but set v,2on sorting, ordering, filing and interfiling realistic cataloging, filing, andsearch considera- operations for a specificprocessor-storage system. tionsin terms of character-setand sort-order For example, "other control elementswhich are requirements that exist todaymay indeed surprise frequently required in the designof information the typical computerscientist facing libraryauto- systems are special sorting elements. Ina directory mation implementationproblems.Nevertheless, of personalnames, such as those which might be particular problems of sorting,filing, andreassem- found in an author bibliography,if names begin- bly orders in terms ofpractical usage needs and ning with 'Mc' and those beginningwith 'Mac' acceptability to the clients ofmechanized systems are to sort together, then special sorting codes and services should besubjects ofconcern to must be entered into the computer for this purpose." designers of machine languages,machine char- (Austin, 1966, pp. 243-244)4.56 acter-sets, and of theprocessors as such.4.6° The size of an adequate character set is apar- The even more difficultcase of extensive mathe- ticularly critical problem inat least two signifi- matical, chemical, and other special cant areas: those of automatic typographic-quality symbols desired on output imposes additionalhardware re- typesetting and of library automation.4.57Complex quirements,whetherforhigh speed character set requirementsare also to be noted printers, in such multiple-access applications photocomposition devices,or character genera- as computer- tors.4.64 This, then, is thearea that has been called aided-instruction (CAI)4.55 Avramet al., considering "Caligraphy by Computer." 4.62 A final automation requirements example of at the Library of Con- unusual character set requirementsis provided by gress, stress that "keyboard entry devicesmust be "Type-A-Circuit" developments.4.63

5. ProgrammingProblems and Languages andProcessor Design Considerations The questions of design and development of The overall system designrequirements of the appropriate programming languages andof proc- future indicate R & Dconcerns with programming essor design are obviously pertinent to all of the languages, and 'especially with hierarchies operations shown in Figure 1. As of of such 1967-68, how- languages at the present time.5.2Controversies ever, special emphasis in terms of research require- certainly exist as between advocatesof more and ments lies in three principalareas: user-oriented input, response and display languages; more "universal" languages and proponents of prob- symbol lem-oriented or user-oriented machinecommunica- manipulation languages capable of handlingar- tion techniques. rays of multiply interrelated data, and increasing interpenetration of hardware and softwareconsid- erations in both system design andsystem use. 5.1. Programming Problemsand Language For example, in the operations ofdeveloping Continuing R & D requirementsfor program- processing specifications from clientrequests for minglanguageimprovementsrepresenttwo service (Boxes 8 and 9 of Fig. 1),we need: new and contradictoryrequirements:on the one hand, more powerful problem-oriented languages;versa- thereis tile supervisory, executive, scheduling, recognized need for increasinglyuni- and account- versal,common-purposelanguagescompatible ing programs; hierarchies ofprogramming lan- with a. wide variety ofsystems, hardware configura- guages;multiprogrammingsystems;improved tions, and types of applications; microprogramming; new approaches and on the other to increasing hand for hierarchies of languagesystems.In interdependence of programming andhardware; addition, a number of specialrequirements for and more versatile and powerfulsimulation lan- more flexible, versatile and powerful languages guages. are just beginning to emerge, especially in such 13 areas as graphical communication, on-line problem tages of on-line documentation systems." Thus the solving, multipleaccess and multiprocessor con- area of program documentation requires further trol systems, simulation, and on-lineinstrumenta- study and concern. tion. We are concerned here, then, with A related difficulty is that ofinadequate means general- for translation between machinelanguages, al- purpose language and compatibility requirements; though some progress has been made." with special-purpose requirements suchas prob- An in- lem-oriented programs, list-processing triguing possibility deserving furtherinvestigation and other has been raised by Burge (1966,p. 60) as follows: techniquesfornon-numericdataprocessing; with special problemareas such as very large "Presented here isa problem and a framework for programs and the requirements of multiple-access its solution. The problem isas follows: Can we and multiprogrammed, get a computer program to scana library of pro- systems; with hierarchies grams, detect common parts of patterns, extract of programming languages, and with theincreasing them, and re-program the library interdependence of software and hardwarecon- so that these siderations. common parts are shared?" Another current question of R &0 concern 5.1.1. Problems of Very Large Programs and ofProgram with respect to programmingproblems is of the Documentation generality with whicha given language system can or cannot cope with a wide variety of In terms of continuing R & Dconcern, we system note first the problems of handling configurations and reconfigurationsover time."° very large The questions of developmentof more effective programs, defined as those that demandmany common-purpose times the available mainstorage capacity and orgeneral-purposelanguages involve very real problems ofmutually exclusive that are sufficiently complex instructure to require features and of choices more than ten independent programmers to work as between a number of meansofachievingcertaindesirable on them." An obvious requirement is to develop features." built-in efficienttechniquesforsegmentation: "When many programmers are involved, there is the prob- Areas of continuing R & Dconcern in program- lem of factoringthe systemintoappropriate ming language developmentsreflect,first,the subtasks. At the present time this isan art rather need for increasing generality,universality, and than a science, andvery few people are good at compatibility (theseobjectivesare followedin its practice, because of the inabilityto find useful general-purpose language construction andstand- algorithms for estimating the size anddegree of ardization, on theone hand, and by increasing difficultyof programming tasks."(Steel,1965, recognition of the needs for hierarchies oflanguage, p. 234). The questions of automatic segmentation, on the other); secondly, the special requirements of although recognizedas critical and difficult prob- multiple-access, multiprogrammed, multiprocessor, lems, have therefore been raised." and parallel processingsystems; thirdly, require- In particular, the checkout ofvery large pro- ments for problem-oriented and other special-pur- grams presents special problems.* Forexample, pose languages, and finally, needs for continuing "another practical problem, whichis now be- advances in hardware-software balancesand in ginning to loom very large indeed and offerslittle fundamental programming theory. prospect of a satisfactory solution, is that of checking the correctness ofa large program." (Gill, 5.1.2. General-Purpose Programming Requirements 1965, p. 203). Further, "theerror reporting rate from a program system of severalmillion instruc- The presently indicated transitionfrom exclu- tions is sufficient tooccupy a staff larger than most sively batch or job-shop operationto on-line, mul- computinginstallationspossess."(Steel,1965, tipleaccesssystemmanagement 5.12sharply p. 233). aggravates the problems of programming language Other specific requirementsin the programming requirements in a number of different,rays. First, problem areasinclude improved provisionfor there are very real difficulties intranslating from adequate program documentationand related con- programming languages andconcepts geared to trols." For example, Pravikoff(1965) presents sole occupancy anduse of system facilities to cogent arguments for the improved documenta- those required in the multiple-access,and multi- tion for programs generally. Mills programmed, much less themultiprocessor and (1967) points network environment." to the special documentation problems inmultiple access systems where usersare less and less apt As Brooks (1965) pointsout: "Today's excite- to be trained programmers." Dennis (1968)points ment centers chiefly around (1) multiprogramming to the present high costs of large-scaleprogramming fortime-sharing,(2)multiple-computer systems efforts as due to inadequatedocumentation that using a few computers forultra-reliability, and prevents taking advantage of programming already (3) multiple-computersystems using a highly par- achieved," while Kay (1969) considersthe advan- allel structure for specialized efficiencyon highly structured problems." 5.14 In all ofthese cases, *See also Section 7.1 of this report,on debugging problems generally. moreover, the R & D requirementsare typically 14 aggravated by a persistent tendency to underesti- analysis processing through those of alphanumeric mate the difficulties of effective problem solution.515 file management to those of list and multi-list An obvious first common-purpose requirement is processing.533 Liststructures have been noted fortrulyefficientsupervisory, accounting, and in several of the file management systems mentioned monitoring control programs5.16that will effectively and it is to be noted further that: "Linked indexes allocate and dynamically reallocate system re- and self-defining entries are an extension of list sources, thatwill be secure from either inad- processing techniques." (Bonn, 1966,p.1869). vertent or malicious interference, and that will be For non-numeric data processing applications, in flexible enough to accommodate to changing cli- general, symbol string manipulation, list processing entele needs, often with new and unprecented and related programming techniques have been of applications.517 particular concern.5.34 Also in the area of general-purpose programming While many advantages of list-processing tech- requirements are the various emerging programs niques have been noted, a number of disadvantages for generalized fileor data base management, are also reported. Among the major advantages are maintenance, and use.5.18 The first requirement, the recursive nature of list processing languages 5'35 here, is for reconciliation of variable input, file and adaptability to dynamic memory allocation and storage, and output formats, together with flexible reallocation.5.36 Also, languages of this type are means for dupe checking on input, combinatorial "well suited to symbol manipulation, which means selection, and output reformatting.5.1° Closely re- that it is possible to talk about the names of vari- lated are the questions of the so-called "formatted ables, and perform computations which produce file systems."5.20 them" (Teitelman, 1966, p. 29), and "by means of A first approach to such general-purpose pro- list structures... athree-dimensional spatial net- gramming systems was undoubtedly that of the work can be modeled in computer memory." Univac B-0 or Flow -Matic system developed in (Strom, 1965, p. 112). the mid-1950's.5.21 More recent examples include Typical disadvantages to be noted include lack General Electric's GECOS III (General Compre- ofstandardization,537degreeofprogramming hensive Operating SupervisorIII) 5.22and Inte- sophistication required,5.38 and wastage of stor- grated Data Store concepts;5.23the TDMS (Time- age space.5.39 Major difficulties are often encoun- Shared Data Management System) at System De- tered in updating and item deletion operations 5.40 velopment Corporation5.24and GIS (Generalized and in dealing with complex data structures.5.41 Information System) of IBM.5.25 Heiner and Leish- We may ask also to what extent the available man (1966) describe a generalized program for list-processing and symbol manipulation program- record selection and tabulation allowing variable ming languages are adequate for current applica- parameters for sort requirements, selection cri- tion needs?5.42To what extent are they useful for teria, and output formats.5.26 the investigation of further needs? In particular, In 1965, comparative operation of file manage- itis noted that "unfortunately, while these lan- ment was demonstrated bydifferentsystems guages seem in many ways directly tailored to the including COLINGO (Mitre Corporation), Mark information retrieval work, they are also in other III (Informatics, Inc.), BEST (National Cash Regis- respects very awkward to use in practice." (Salton, ter), Integrated Data Store (G.E.), and an on-line 1966, p. 207). management system of Bolt, Beranek, and New- List-processing languages and structuresare man, Ine.5.27 It was concluded that: "All of these particularly clumsy, moreover, for multiply inter- systems were able to accomplish the processing related and cross-associateddata.543Short of required, but their approaches varied considerably, truly large-scale associative memories, effective particularly in the file structures chosen for the compromises are needed as between list struc- application, executive control procedures, and level tures, including multilist programming languages, of language used in specifying the processing to be and file organizations that will provide economy performed." (Climenson, 1966, p. 125). of both storage and access.5.44 In addition, we may note the examples of a file Beyond listprocessing procedures there are organization executive system,5.28 a program man- trees, directed graphs, rings, and other more complex agement system involving a two-level file,5.29 and associative schemes.5.45 A relatively early example a file organization scheme developed for handling isthat of "Rover" with "up links" and "side various types of chemical information.5.3° Another links" as well as "down links".5.46 Savitt et al., 1967, documentation application involving a list-ordered describealanguage, ASP (Association-Storing fileis provided by Fossum and Kaskey (1966) Processor), that has been designed to simplify the with respect to word and indexing term associa- programming of nonnumeric problems, together tions for DDC (Defense Documentation Center) with machine organizations capable of high-speed documents.5.31 parallel processing.5.47 Intheareaof common-purpose languageo, Ring structure language systems are represented also, there is need to reconcile the differing re- by Sketchpad developments5.48and the work of quirementswithrespecttodifferentclasses Roberts at M.I.T. for graphical data processing5'49 of data structures,5.32 from those of numerical and by systems for use in question-answering 15 and similar systems, suchas Project DE 4CON.5.5° Still other associative and, secondly, that "the taskof developinga structure developments are complete new real-time languagewould be inter- discussed for example, by Ashand Sibley (1968),55' esting and challenging since Climenson (1966),5.52 Craig it would require et al., (1966),553 Dodd reconsiderationoftheconventionalprocedure (1966),554 and Pankhurst (1968).5,55 and data definingstatements from the viewpoint There are certainly those whoneed far more of thereal-timerequirements."(Opler,1966, generalizedmultipurposelanguagesand,with p. 197). equivalent force, those who advocatespecialized, Further, on-line problem-solvingapplications re- man-oriented,andspecial-purposelanguages. quire dynamic and flexibleprogramming capabil- A specific current design problemhas to do with ities.5.59 Experimentation withthe working program an appropriate compromise between theseappar- should be permitted with ently contradictory due regard for system comments. A future R & D protection and control."° Generalizedproblem- requirement is to seekmore effective solutions. solving capabilities shouldbe availablein the Thus Licklider insistsupon the need to bring language system withoutnecessary regard to spe- on-line"conversational" languagesmore nearly cific applications."' Suchprograms should be abreast with themore conventional programming extensivelyself-organizing,self-modifying,and languages as an R & D challenge.556 capable of adaptationor tentative "learning." 5.62 Then it is to be noted thatlanguages for on-line 5.1.3. Problem-Oriented andMultiple-Access Language use should be relatively immune to inadvertentuser Requirements errors.5.63 Teitelmancomments, for example, that Increasing needs for advanced "in languages of thistype, FORTRAN, COMIT, special-purpose MAD, etc., it is difficult program language developments have beenrecog- to write programs that con- nized for theareas of man-machine interactive struct or modify procedures because thecommuni- problem-solving and computer-aidedinstruction cation between procedures isso deeply embedded applications,graphical manipulation operations, in the machine instruction coding,that it is very and simulation applications,among others. It is difficult to locateentrances, exits, essential vari- noted inparticular that "their ables, etc." (1966,p. 28). power [that of In language problem-orientedlanguages]intheextension systems such as TRAC,554 thepower- of computing lies in the fact thatthe expert knowl- fulness is largely buried inspecific machineproc- ledge of skilled analysts inparticular fieldscan essors for this language theuser is freed from be incorporated into theprograms, together with learning a number of arbitraryexceptions, he is a language which is native to the field,or nearly so, not able to "clobber" the underlyingprocessor by and can thereby break down thebarriers of eco- inadvertent goofs, he is able iterativelyto name and nomics, of lack of familiarity withthe computer, rename strings and procedures he wishesto use and of time." (Harder, 1968,p. 235). largely in terms of hisown convenience, and he is able to makeuse of the language and system with Important requirements inprogramming lan- minimal training guagesreflect, or prior experience. predictably,thoseof multiple- Related problems affect thedesign of special- access and time-shared systems involvinga suitable hierarchy of languages purpose languages ("the design of special-purpose so that the relatively un- languages is advancing rapidly,but it hasa long sophisticated usermay converse freely with the way to go," Licklider, 1965, machine without interference p. 66), the interaction to other users or of executive and consolelanguages ("wherever to the control and monitorprograms and yet also draw remote consoles are used,we find the users en- upon common-useprogramsanddata. thusiastic. However, they always hasten The effectiveuse of such systems in turn demands to add that extremely tight there is much to learn about thedesign of execu- and sophisticated programming tivesand processorsforconsolelanguages," to keep "overhead" to a minimum 5'57 andto provide (Wagner and Granholm, 1965, dynamic program and data p. 287), the provision location and relocation. of adequate debugging aids ("thecreation of large Relatively recent developmentsin this areaare discussed by Bauer, Davis, Dennis programming systems usingremote facilities re- and Van Horn, quires a number of debugging aidswhich range in Licklider, Clippinger, Op ler,and Scherramong complexity from compilers others.558 Bauer (1965,p. 23) suggests in particular to simple register con- that: tents request routines." Perlis, 1965,p. 229), and "Entirely new languagesare needed to the development of effective allow flexible and powerfuluse of the computer modularity inprograms from remote stations." and compilers ("it will benecessary for the software to be modular to the greatestextent possible Multiple-access systems and networksobviously because it will needto work up to the next level of require R & D efforts in the developmentof lan- software." Clippinger, 1965,p. 211). guagesystems"optimized forremote-on-line In connection with the latterrequirement Lock use" (Huskey, 1965,p.142). Op ler points out, discusses the desirability of first, that "for the most difficultareas an incremental com- telecom- piler which is characterized byits ability "tocom- munication,processcontrol,monitors,etc. pile each statement real -time languages have provided independently, so thatany little assistance," local change ina statement calls only for there- 16

4,1,.,.1141ak& wociAtadwer,c- 04),w A,WALArt,1,1.,..1kAg-A,Afie...... Aftlialied.,....,151,..,,..S. compilation of the statement, not the complete tion. Certainly, for the future, "the entirespec- program." (1965, p. 462). trum of language from binary machine code to the Time-sharing or time-slicing options and graphical great natural languages will be involved in man's communication possibilities cast new light not only interaction with procognitive systems." (Licklider, upon the problem-oriented but also upon the user- 1965, p. 104). oriented languages and improved possibilities for In the area of desired hierarchies of languages, man-machine communication and interaction. Ef- we note such corroborating opinions as those of fective compromises between the system design, Salton who asks for compiling systems capable of the programmer, and the client have notyet been handling a variety of high-level languages,spe- met, even although the demands of modern systems cifically including list processing and string manip- are increasing in complexity. ulations 5.68 and of Licklider who points out that in The problems of effective ,programming and uti- each of many subfields of science and technology lization reach back beyond specific routines,spe- there are specific individual problems of terminol- cific languages, and specific equipment. Theyare ogy, sets of frequently used operations, data struc- involved in all the many questions ofsystems plan- tures, and formulas, indicating a very real need for ning, systems managment and systems design. For many different user-oriented languages.5.69 How- example, in time-shared operations, "considerably ever, in his 1966 review, the first ACM Turing Lec- greater attention should be focused on questions of ture, Perlis concludes: "Programmers should never economy in deciding on the tradeoff between hard- be satisfied with languages which permit themto ware and software, in using storage hierarchies, and program everything, but to program nothing of in determining the kind of service which the time- interest easily. Our progress, then, is measured by sharing user should be offered." (Adams, 1%5,p. the balance we achieve between efficiency and 488). Finally, the critical problems reach farther generality." (1967, p. 9). back than the potentialities of hardware, software Moreover, even in a singlesystem, it may be and systems planning combined, to fundamental necessary not only to reconcile but to combine the question of why, how, and when we should look to contradictory advantages and disadvantages of dif- machines for substantial aid to human decision- fering levels of programming language in various making and problem-solvingprocesses. ways. Thus Salton states "it is possible to recognize "An analysis of the various requirements that five main process types which must be dealt with a programming language must satisfy (to be known in an automatic information system: string manipu- both by the job-originator and the job-executor lation and text processing methods, vector and and to be capable of expressing both what the matrix operations, abstract tree and list structure first wants to be done and what the second is capable manipulations, arithmetic operations, and finally of doing) invoh es basicresearc hes on the linguistic sorting and editing operations." (Salton, 1966,p. nature of programming languages." (Caracciolo di 205). A second complicating factor relatesto the Forino, 1965, p. 224). still unusually fluid situation with respect to hardware developments, logical design, and the increas- 5.1.4. Heirarehies of Languages and Programming ing interdependence of hardware-software factors Theory in the consideration of future system and network design possibilities. Directmachine-languageencoding andpro- Beyondtheinvestigation,development, and gramming was thefirst and obvious approach experimental application of advanced programming to both numeric and non-numeric data processing languages for specific types of application suchas problems. In both areas, forms of communication graphical data processing, simulation,or on-line with the machine thatare more congenial to the question-answering and problem-solving systems, human user imve been developed,as formal otudy is needed of fundamental problems ofpro- "programming languages" (FORTRAN, ALGOL, gramming theoITy57° For example Halpern si. notes COBOL, and thelike).5.65In addition, special .the increasingly wide gulf between research program languages to facilitate either problem- and practice in the design of programming lan- solving and question-answering systems,or textual guages." (1967, p. 141), while Alt emphasizes that data processing, or both; (suchas list-processing "we do not yet havea good theory of computer techniques, IPLV, LISP, COMIT, SNOBOL, languages, and we are nowherenear the limit of more recently TRAC, and others) 5.66 have been the concepts which can be expressed in such lan- developed. More and more, however, it is beginning guages." (1964, p. B.2--1). to be recognized that hierarchies of language are There is also to be noted in thecurrent state of the essential to present and foreseeable progress.567 art in the computer and information sciencesan Burkhardt (1965, p. 3) listsa spectrum of pro- increasing concern for the relationships between gramming languages on the basis of the "declarative formal modelling techniques, generally, theques- freedom" availableto the user, from absolute tions of formal languages, and the development machine languages withnone, to"declarative of powerful, general-purpose programming lan- languages" which provide a description of the guages.' Karush emphasizes that "the develop- problem and freedom of both procedure and solu- ment of an integrated language of mathematical 17 and computer operations is part of a more general to the design of computing systems" (1965, p. problem associated with the automation of control 137), and Burkhardt warns: "As long as actual functions, This is the problem of formalism which computers arenot well understood therewill embraces both mathematical representations of not be much hope for very successful development systems and representations of the processes of of usefuluniversalprocessors." (1965,p.12). actual execution of systems, by computer or otherwise." (1963, p. 81). 5.2. Processor and Storage System Design A similar concern is expressed by Gelernter: Considerations "Just as manipulation of numbers in arithmetic is the fundamental mode of operation in contempo- In the area of information processing systems rary computers, manipulation of symbols in formal themselves, current trends have been marked systems is likely to be the fundamental operating not only by new extensions to the repertoires mode of the more sophisticated problem solving of system configurations available for the large, computers of the future. It seems clear that while high-cost, high-speed processors but also by a the problems of greatest concern to lay society continuing tendency to the development of com- will be, for the most part, not completely formaliz- putersystem"families."Increasingattention able, they will have to be expressed in some sort is being given to both "upward" and "downward" of formal system before they can be dealt with by compatibility that is,to means by which pro- machine." (1960, p. 275). grams operable on a large system may also oper- High-order relational schemes both in languages ate on a smaller, slower, or less expensive member and in memory structures will thus be required.5.72 of the same family, and vice versa. For example, "the incorporation inprocedural Increasing attention has also been given to pro- programming languages of notations for describing viding adjuncts to existing and proposed systems data structures such as arrays, files, and trees, which will give them better adapatability to time- and the provision to use these structures recur- sharing and on-line multiple-access requirements. sively together with indications of the scope ofSimilarly, the requirements for handling a variety definition, will help greatly with the storage alloca- of input sensing modalities and for processing tion problem and assist the programmer orga- more than one input channel in an effectually nizationally..." (Barton, 1963, p.176). In this simultaneous operation clearly indicate needs for connection, Press and Rogers (1967) have described continuing research and development efforts in the IDEA (Inductive Data Exploration and Analysis) the design and use of parallel processing tech- program package for the detection of inherent niques,multiprocessornetworks,time-shared structure in multivariate data.5,73 multiple access scheduling and multiprogramming. We note also that "even in the more regular Hierarchies of languages are implied as we have domain of formal and programming languages, seen, ranging from those in which the remote con- many unsolvedpractical and theoreticalprob- sole user speaks to the machine system in a rela- lems remain. For example, the matter of recovery tively natural language (constrained to a greater or from error in the course of compilation remains lesser degree) to those required for the highly so- in a quasi-mystical experimental state, although phisticated executive control, scheduling, file pro- some early results, applicable only to the simplest tection, accounting, monitoring, and monitoring of languages suggest that further formal study instrumentation programs. of this problem could be worth while." (Oettinger, Tie-in to various communication links, generally, 1965, p. 16). should include remote consoles, closed circuit TV, Requirementsforcontinuingresearchand facsimile, voice quality circuits, and the like, with development activities in the area of computing capability for real-time processing. Message security and programming theory are increasingly seen as protection facilities are often required, including directly related to needs for more effective multi- encoding and decoding. Access to error detection access time-sharing, multi-programmed, and multi- and error correction mechanisms are also necessary. processor systems. It can be foreseen that "Orga- Overall system design requirements indicate also nizational generality is an attribute of underrated the necessary exploitation of new hardware tech- importance. The correct functioning of on-line nologies, new storage media, associative-memory systems imposes requirements that have been procedures for file and data bank organization and met ad hoc by current designs. Future system management, the use of dynamic reallocations of designs must acknowledge the basic nature of space and access to both programs and files or data theproblems and provide general approaches banks in multiple-access systems, protective and totheir resolution." (Dennis and Glaser, 1965, fail-safe measures, and the development of hier- P. 5). archies of languages of access and usagehier- What are some of the difficulties that can be archies of stored data files, and hierarchies of foreseen with respect to the further development systems.5.74 ofhierarchiesofsystems?Scarrottsuggests Many of the above factors are discussed in other that "the problems of designing and using multi- sections of this report or in other reports in this level storage systems are in a real sense central series. Here, we will consider briefly some of the 18 problems of central processor design, parallel proc- array. The system designer needs to avoid as far essing, and hardware-software interdependence, aspossiblethe problems of unnecessary and clumsy programming in order to apply a single 5.2.1. Central Processor Design procedure to such arrays. For example, Wunderlich points out in connection with sieving procedures With regard to the central processing system, it is for computer generation of sets and sequences that noted that it should be operable in both time-sharing "there are obvious programming difficulties con- and batch processing modes, and that it provide nected with sieving on a field of bits," (1967, p. 13). simultaneous access to many users. Efficient mul- Another hardware-software desideratum here is tiple access should be provided to the hierarchies of obviously the need for efficient bit-manipulation storage with the user having, in effect, virtually un- capabilities. For example, the User would like to find, limited memory.5.75 There should be a capability for for a given gray-level representation of a graphic accessing either internal or auxiliary associative input that, at a given location and blackness level, memory devices. some or all or none of the neighboring locationshave A continuing problem in processor-storage system the same blackness level recordings (this is im- design is that of address- circuitry.5,7° Closely related portant in eliminating "fly specks" from further to this problem are questions of content-identifica- processing, in filling-in "holes" that result from im- tion-matching whether for sequential or parallel perfect printing impressions, and also in deter- access. Indirect addressing and multiple relative mination of the relative locations of centers of addressing via a number of index registers is an blackness when attempting to reconstruct three- important consideration.5.77 dimensional imaging for serial sequences of two- Moreover, direct program access to all registers dimensional image representations). by ordinary instructions and with interlock protec- Problems of computer design as well as program- tion features is often required. If the index registers ming for array processing are discussed by Senzig can be simultaneously and interchangeably used as and Smith (1965) in terms of a worldwide weather instruction counters, there are additional parallel prediction system and by Roos (1965) in terms of processing benefits. For example, such capabilities the ICES (Integrated Civil Engineering System) at can provide for jumping to the nth instruction from M.I.T.5.79 Association matrices present a special the present one, determining whether an instruction form of data arrays requiring efficient manipulation has been executed or not, and other flexible capa- and processing. Such considerations are particularly bilities for debugging, diagnostic, and evaluation important in the experimental research or on-line purposes. It should be possible to manipulate index instrumentationsituations.5.8°Inaddition,bit- registers several at a time. manipulation and array-processing requirements System users may need a relatively long word are severely constrained in commercially available length to provide numerical accuracy in long floating systems. point values, for manipulation of large matrices, to A requirement of major future importance (espe- provide duplex operations on two sections of the ciallyfor chemical informationsearching,file, word simultaneously as in complex number process- organization, mapping functions and graphic data ing, and the like. Automatic unpacking of word processing) is for efficient bit manipulation capa- subsets in variable sized bytes is also recommended bilities, including convenient Boolean processing for future processor design. Salton indicates needs and transplant features. Again, bit manipulation for "flexible instructions operating on individual capabilities are important because many operations bits and characters, and flexible branching orders. require consideration of all the orthogonal neighbors Pushdown store instructions, such as 'pop' or 'push' of a single bit position. should also be useful for the listoperations." In future system designs, increasing needs for (1966, p. 209). multivalued logic approaches can also be foreseen. Also desirable is the ability to access a word In general, a binary (two-valued) logic pervades in- simultaneously from more than one computer sys- formation processing system design and the basic tem or component with automatic protection inter- methods of information representation as of today. locks in case of conflicts. Power-failure protection For the future, however, attention needs to be di- systems should be available and protection should rected toward multivalued logic systems and to also be provided against unauthorized access to direct realizations of the n-ary relations between the various memory and data bank sections. data elements of stored information. There are new More flexible pagination 5.78 is needed for some technological possibilities that point in this direc- important applications. For example, in graphic data tion (e.g., new devices that are capable of at least processing, dynamic memory reallocation proce- ternary response,5." or multiple response by color- dures requiring fixed pagination would be awkward to coding techniques from a single "bit" recording on use and highly inefficient. Storage allocation should advanced storage media). In addition, parallel proc- be under programmer control. For example, if essing, associative processing, and iterative circuit pictorial data overflows the boundaries of a page, at techniques point the way to new complexities of least 4 and as many as 9 pages may be required since program command and control and to new, multi- such data must be processed as a two-dimensional valued, processing opportunties as well. 19 Then, as Wooster comments: "Radically different on the sparcity of attempts made to date on the types of computers may well be needed. At present, potentials of parallel processing in programs for the best way of building these seems to be through many problems. He states further that while the the creation of logical structures tending more and tasks of segmenting the problem itself for parallel more in the direction of distributed logic nets, processing attack are formidable, "they must be wherein vast numbers of processes occur simultane- undertaken to establish whether the future develop- ously in various parts of the structure. Right now, ment lies in the area of parallel processing or the best building blocks for such systems seem to be not." 5.90 Then there is need for judicious inter- multifunctional microprogramable logic elements." mixturesof parallel and sequential processing (1961, p. 21). techniques inspecific design, programming, or Increasing complexity of central processor design application situations.5.9' is indicated by developments in advanced hardware Parallel processing potentials are also closely technologies; 5.82 while increasing flexibility is dic- related to, and may be interwoven with, multi- tated by dynamic reconfiguration requirements."3 processing systems which involve: "The simul- Modularity is an important consideration."' The taneous operation of two or more independent critically challenging, interacting problems of both computersexecutingmore-or-lessindependent design, programming and utilization of multiproc- programs, with access to each other's internal essors and of parallel processing are emphasized by memories..." (Riley, 1965, p. 74). In particular, Brooks (1965), Amdahl (1965), Burkhardt (1965), and the Solomon Computer and Holland Machine con- Op ler (1965), among others. Thus: "The use of multi- cepts may be noted.5.92 For another example, computers implies intercommunication, with the increasing parallelism of operation of a multiple associated implications of interconnection, recon- access processing system has been investigated at figuration and interlocking." (Amdahl, 1965, p. 38). the Argonne National Laboratory in terms of an "Intrinsic Multiprocessing" technique consisting of 5.2.2. Parallel Processing and Multiprocessors n time-phased "virtual" machines which time-share The possibilities for the use of parallel processing very high speed execution hardware (Aschenbrenner techniques should receive increased R & D atten- et al., 1967), while ILLIAC III has been designed tion. Such techniques may be used to carry out data for parallel processing of pictorial data.5.93 transfers asynchronously with respect to the processing operations,"5 to provide analyses necessary 5.2.3. Hardware-Software Interdependence toconvert sequential processing programs into parallel -path programs,5.88 or to make allocations of The earlier dichotomy as between "hardware" system resources more efficiently because con- and "software" considerations is beginning to yield, straints on the sequence in which processing opera- not only to the increasing interdependence of the tions are executed can be relaxed.587 Applications two factors in many information processing applica- to the area of pattern recognition and classification tion requirements, but also to technological de- research and to other array processing operations velopmentsin"firmware"(ineffect,wired-in are obviouS.5'88 microprogramming 5`4) and to growing recognition of However, there are problems of effective use of the critical importance of more precise and compre- parallel processing capabilities. Some examples hensive "brainware" in systems planning, design, of the discernible difficulties with respect to current specification, and implementation.5.95 parallel-processingresearchanddevelopment Is it currently possible to separate computer and efforts have been noted in the literature as follows: storage system design considerations from those of programming language design and of programmed (1) "Multiprogrammed processors will require executive control? Several experts testify that, if it more explicit parallel control statements in is still possible today, it will soon be so no longer 5.98 languages than now occur." (Perlis, 1965, It can be quite clearly seen that the areas of com- p. 189.) puter theory and program design are becoming (2) "Much additional effort will have to be put increasingly interdependent with those of adequate into optimizing compilers for the parallel programming languages,"software", and user- processors that may dominate the computer tolerance levels. At the same time, new possibilities scene of the future." (Fernbach, 1965, p. 84). for multicommunicator, multiprocessor, and multi- (3) "Computers with parallel processing capa- user networks are increasingly coming to the fore. bilities are seldom used to full advantage." The growing interdependence is stressed, for (Opler, 1965, p. 306): example, by Schultz (1967),5.97 and by Lock (1965) There are also problems of R & D concern in who notes the strongly increasing influence of programming language developments involved with multiprogrammed, on-line systems upon the or- increased use of parallel processing capabilities. ganization of the storage facilities. Scarrott (1965, The possibilities of "Do Together" provisions in p.137) for another example, insiststhat "the compilers, first raised by Mme. Poyen in 1959,5.89 problems of designing and using multilevel storage add a new dimension of complexity for analysis, con- systems are in a real sense central to the design of struction, and interpretation. Fernbach comments computing systems." 20 Thus, "as we enter an era of bigger andmore Will we be able to modify thesystem to accom- complex systems some new requirementsare modate new or additional hardware? coming to be of major importance. Will we be able to adda completely new function to an already operating system?" (Perry, Will we be able to minimize the program han- 1965, p. 243). dling by proper allocation to primary,e.g., In the next section, therefore,we will consider core, or secondary, e.g., drum or disk, storage? some of the advanced hardware developments before discussing such overallsystem design con- Will we be able to incorporate changes to the siderations as debugging, on-line instrumentation functional operation of the system? and diagnosis, and simulation.

6. Advanced Hardware Developments Certain obvious overall system design require- into three basic types: solid-state, semiconductor, ments have to do with the further extensive develop- and gaseous. Typical examplesare the ruby solid - mentandapplicationofadvancedhardware state laser, gallium arsenide semiconductor laser, technologies, especially opto-electronics generally and the neon-helium gas laser." (1966,p. 38). (and lasers and holography in particular), with Certainly, lasers, whether of thegas, fluorescent integrated circuit techniques, and with improved crystal, or semiconductor type,are finding many high-density storage media. new possibilities of application in computer, com- Recurrent themes in current progress toward munication, and information processing systems." very-high-speed,computer-controlledaccessto As sources of illumination theycan provide greater primary, secondary and auxiliary storage banks, display efficiency 6.6 and greater resolution with from the standpoint of hardware technology, include respect to display systems involving lightbeam the questions of matching rates of data-and/or- deflection techniques (Soref and McMahon, 1965, instruction access to those of internal processing p. 60), and they provide an effective aid to the cycles, and of the prospects for integrated circuit boundary and contrast enhancement techniques for and batch fabrication advantages in design and image processing developed at the National Physical construction. Laboratory at Teddington, England. More specifi- These new technologies may also be combined in cally, this technology promises new developments in various ways. For example, Lockheed Electronics space communications," in memory construction has been using deflected laser beams to scan and design," in the development of analytical photochromic planes very rapidly and very ac- techniques such as Raman spectroscopy and photo- curately," laser and holographic techniques are microscopypo intheidentificationoffinger- conjoined in equipment designed to photograph fog prints," in quantization of high resolution photo- phenomena in three dimensions," and it has been graphs 6'12 and in the use of holograms for collection, reported that "laser devices show promise ofvery storage, and regeneration of two- and three-dimen- fast switching which together with optical inter- sional data." connections could provide digital circuits that are Small, very high-speed memories may be driven faster than electronic circuits." (Reimann, 1965, by laser beams,6"4 and laser components contribute p. 247). to the design of "all-optical" computers 6.15 and computer circuits and components.Laser in- scribing techniques are being investigated for such 6.1. Lasers, Photochromics, Holograph,, applications as large screen real-time display s,1317 and Other Optoelectronic Techniques and for highly compressed data recording, for example, atPrecision Instrument Company. 618 New hardware developments that are technically As of March 1969, it could be reported that orders promising in terms of the long range research and had been placed for UNICON systems by Pan development necessary to support future improve- American Petroleum Corporation, andwere under ments in information processing and handling sys- consideration by several other organizations, in- tems include the development of special laser tech- cluding U.S. Government agencies." In particular, niques for switching, storage, and other purposes, the National Archives and Records Service has and the possible use of holograms or kinoforms for been studying the possibilities of converting present 3-dimensional pattern recognition and storage. magnetic tape storage to this syster,.620 6.1.1. Laser Technology Investigations of future technical feasibility of using laser devices for high speed data storage and/ Writing in 1965, Baker and Rugari have pointed or processing have thus been complemented by out that "a wide variety of lasers have been dis- exploration of possibilities for recording ontovery covered and developed since the first laser device large capacity storage media as also in develop- was operated five years ago." Lasers can be classed ments at Honeywell,' at the Itek Corporation,6.22 21 and at RCA,6.23 an IBM system designed for andnow 6.1.2. Photochromic Media and Technique in operation at the Army Electronics Commandas well as IBM developments in variable frequency Bydefinition,photochromic(orphototropic) lasers,6.24 and a recording system from Kodak that compoundsexhibitreversibleeffectsorcolor uses fine-grained photographic media, diffraction changes, resulting fromexposure to radiant energy grating patterns, and laser light sources,"26 Ho lo- in the visible or near visible portions of thespec- graphic techniques may also be appliedto the de- trum.642 Such media give excellent resolution and velopment of associative memories with possible reductioncharacteristics, and because of the analogies to human memory-recall systems.6.26 reversibility property, theycan theoretically be erased and rewritten repeatedly,6.43 althougha Kump and Chang (1966) describea thermo- continuing area of R & D concern is that of problems strictive recording mechanism effectedon uniaxial of fatigue.6.44 They also enable storage of images Permalloy films by the application ofa local stress with a wide range of gray scale.646 Such materials induced by either a laseror an electron beam, have been known for at leasta hundred years or promising large capacity memories of better than more (Smith, 1966). In fact, as Smith suggests, they 106 bits per square inch storage efficiency.6.27 Then may have provided the means for achieving the there are combined optical and film techniques for world's first "wrist-watch." 6.46 digital as well as imageor analog recording and storage. Specific examples include IBM photo-chip Tauber and Myers (1962) and Hanlonet al. (1965) developments 8.28 thermal recording developments offer summaries of NCR effortsto provide com- at the NCR research laboratories,6.26 Precision mercial applicability to photochromic recording Instrument's UNICON System 6.3o and, in general, techniques for large-capacity microimagestorage the area of photochromic storage technology. files.6.47 A British example of application is the Tech- nical Information on Microfilm Service.648 Areas of continuing R & Dconcern with respect A less favorable characteristic of the photo- tolasercommunicationspossibilitiesinclude chromic material appears in thecase of storage questionsofmodulationandtransmission1631 files the permanency of recording dependson acquisition and tracking problems,632 isolation from ambient temperature, ranging from onlya few atmosphericinterferrence conditions 8'33 and noun',at normal room temperatures to perhaps possibilities for controlled atmosphere systems.634 several years under rigid temperature controls.646 Vollmer (1967) notes that an experimental short- Therefore,for mass and archival storage, the range laser communication system has narrow beam NCR system involves transfer from the photo- width with significant advantagefor privacy. In chromic images to a high-resolution photographic particular,, "operation at 9020 A enhances this emulsion for permanent files." The remaining privacy by virtue of its invisibility." (p. 67.) Some advantages are two-fold. First, the reduction is examples of the experimentaluse of lasers for impressive: 1,400 pages of the text of the Bibleon communications purposes were given inan earlier an approximately 2" x 2" film chip is the widely report in this series,636 and it was noted that the demonstrated example (Fig.2). Secondly, 'spot' most successful ventures to 'date have been at erasure and rewriting provides an important in- opposite ends of the distance spectrum 8.36 However, spection and errorcorrectioncapability.Itis laser scanning techniques combined with other claimed that: "Instantaneous imagery followed by means of communication may offer important gains immediate inspection permits the production of in high-resolution facsimile transmission. Forex- essentially 'errorless' masters for the first time". ample, a system developed by CBS Laboratories (Hanlon et al., 1965, p. 10). uses a laser beam to scan photographic (llm, convert In the area of internalmemory and switching to video signals, and transmit, via satellite, military design, Reich and Dorion (1965) report of the photo- reconnaissancepicturesfromVietNamto chromictechniquesthat:"Thephotochromic Washington 6.37 medium has extremelylargestorage capacity Continuing requirements for further develoments latently available in physically small dimensions. in the application of lasers in displaysystems The basic photochromic switchesare the molecules involve, for example, efforts directed toward less themselves...Photochromic media can beem- expensive high quality semiconductor lasers 8'38 and ployed for many write-erase-rewrite cycles and toward solving problems of deflection, modulation, give almost nondestructive read...Appropriate and focusing." Kesselman suggests that results photochromic systems can retain stored data with- to date in terms of laser displays are inconclusive out power consumption...The memory can and that practical applicationsare not likely in the probably be designed to be stored for quitea long near future (1967, p. 167). As in the case of laser time." (p.572) 6.51 versus electron beam displays.ao the absence of A photochromic medium in the form oftrans- requirements for vacuum techniques favors the parent silicate glass containing silver halide particles eventual use of laser rather than electron beam has been suggested for such applicationsas eras- techniques in many high density datastorage able memories, displays for airtrafficcontrol recording applications 6.41 operations, and optical transmission systems." 22

tAilg.taliaidard.16,1A.L a

$1"'";:-

r ria 419F1, kisa * r"; 7.47i7v1" fa-L-17.4:1 -

FIGURE 2.Photochromic data reduction.

In addition, itis to be noted that photochromic needs for truly long-range R & D planning in many films may be activated by CRT phosphors for use in areas of computer and information sciences and information display systems (Dorion et al., 1966)6.53 technology, since it is by no means a recent area of and may also be used for real-time target tracking.654 investigation, the principles having been announced Recent developments suggest the use of photo- by Gabor as early as 1948.6.57 The basic holographic chromics for digital data storage.6.55 phenomena are described by Cutrona (1965, p. 89) Finally, the properties of photochromic materials as follows: "A hologram is produced by recording on might be used for improved performance of holo- photographic film the interference pattern, resulting graphicrecording,reconstructionanddisplay from the illumination of some object with a wave- systems.6.56 Thus, "the use of self-developing photo- front from the same source".6.58 chromic devices in the place of the photographic plate would enhance the value of wavefront recon- Leith et al. (1965, p. 151) point out further that struction microscope by permitting nearly real-time "by combining conventional wavefront reconstruc- operation and eliminating the chemical develop- tion techniques with interferometry, it has been possible to produce holograms from which high- ment process." (Leith et al., 1965, p. 156). quality reconstructions can be obtained. These 6.1.3. Holographic Techniques reconstructions bear close likeness to the original object, complete with three-dimensional charac- Holography isa new information processing teristics...The object can be a transparency, or technique, but it is, in fact, highly illustrative of itcan be a solid, three-dimensional object".6." 23 Armstrong (1965) emphasizesthat, in general,no lenses are required and the Electronics Command, Fort Monmouth.In partic- reconstructed image ular, "it is intended thata hologram of a binary data can be magnified or demagnifiedas desired.6.6° array would constitute the card-like removable As of early 1969,however, the question has been raised that holograms media. Upon insertion into thememory read unit, may be already outdated.6.6" the hologram would continuouslyfocus a real image A new wavefrontreconstruction devicethe kino- on the data onto a photodetector matrix. Such form is a computer-generateddevice intended to an provide a reconstructed arrangement can permit electronic randomaccess three-dimensional image to the information within thearray while eliminating of various objects withgreater efficiency than is the stringent optical requirements available with holographic procedures on the detectors 1969).6.62 (Lesem et al., involved." (Chapman and Fisher,1967, p. 372).6.72 Other R & D possibilities of In addition to theuse of holographic techniques interest includeex- perimentation with acoustic 6'73 andcomputer-gener- for three-dimensionalimage storage and recall ated holograms 8.74 The (including rotation 6'63)9 thesetechniques are also computer generation of being explored holographic or kinoform recordingsis thus another for bandwidth compressionin development in thisarea of advanced technology. pictorial data storage 6.64. theproduction of highly For example, digital hologramsmay be generated magnifiedimages,6.65and ,)ther novel applica- by computer simulation of tions.6.66 In particular, it hasbeen claimed that wave fronts that would holographic techniques offer emanate from particular optical elements arranged a new potential for in specific geometrical relationships(Hirsch et al., high-quality-image capture inregions of the electro- 1968).6.75 An interesting magnetic spectrum extending area of investigation is that beyond those that of computer synthesis ofholograms of three-dimen- have been achieved by opticalrecording techniques sional objects which donot, in fact, physically in the the region of visible light6.87 exist." Then it has been reportedthat "General Electric's Advanced DevelopmentLaboratories, Schenectady, build a laser holographreadera device capable of 6.1.4. Other OptoeleetronieConsiderations reading characters in severalways. It can detect a In general, it is emphasized single object out ofmany without scanning, or if that increasing in- scanning is used, terest has been evident in theuse of optoelectronic can recognize up to 100 different techniques for bothcomputer and memory design characters. The holographic readeris said to show for a wide variety of wide tolerance for variations reasons. Scarrott (1965) and in type font and is Chapman and Fisher (1967)point to the high den- expected to find applications in thecomputer field." sities (Veaner, 1966, p. 208.) achievablewithphotographicmedia.6.77 Laser and holographic techniques Reich and Dorion (1965)suggests a photochromic in combination film memory plane, 2"x 2", with 645 subarrays in- are also being investigated for high densitydigital dividually accessible and data storage, for example,at the Bell Telephone a total capacity, assuming Laboratories,6.68 at Carson only 50 percent utilization of thefilm area, of better IBm.676 Laboratories,666 and at than 12 million bits.6.76 Potentially,then, many of these techniques promisesignificant advances in Some specialareas where advanced optoelectronic data storage, in logic and techniques and improved materials processing circuitry, in or storage media alternative communicationsmeans, in computing continue to be needed include"certain operations, such as two-dimensional spatial or access speed, and in data collection withrespect filtering (that)can to two- and three-dimensional objectrepresenta- be readily accomplished, inprinciple, with coherent tion, including spatial filtering.6.76 light optics. Problems underconsideration ,include: the effect of film-grainnoise on the performance of Bonin and Baird (1965,p. 100) list other applica- a coherent optical system; the relation of film tions of optoelectronic techniquesfor tape and card thick- readers, position indicators, ness and exposure; techniques for the makingof and recognition equip- spatial filters; and theeffect on the reconstructed ment. In addition, importantnew areas will include picture of various operations (suchas sampling, use in communication and transmissionsystems. quantization, noise addition)upon the hologram." Here it is noted that opticaltechniques as applied (Quarterly Progress ReportNo. 80, Research to advanced communicationsystems planning relate Laboratory for Electronics, M.I.T.,221 (1966).) also to continuing theoreticalinvestigations. Thus, McCamy (1967) reports "preliminary studies of communicationsystems em- recent extension of pre- ploying optical frequencieshave indicated three vious R & D investigations into fadingand aging topics to which the blemishes in conventional microformsto the effect concepts and techniques of of formation of such blemishes modern communication theorymay most profitably on information stored by addressed. Theyare (i) the import of quantum by means of holograms. It isto be noted further that electrodynamics for the certain types of holograms havean important im- characteristics of efficient munity to dust and scratches."' communication systems, (ii) therelevant description of device noiseas it affects the performance of Possibilities fora holographic read-only storeare communication systems, and (iii) under investigationat IBM (Gamblin, 1968), RCA the statistical (Viklomerson et al., 1968), and characterization of the atmosphereas a propagation at the U.S. Army channel at optical frequencies."(Quarterly Progress 24

ysWy Report No. 80, Research Laboratory for Electronics, overcome present-day deficiencies of processing M.I.T. 178 (1966).) systems in such applications as pattern recogni- For use as computer logical elements, somewhat tion. less attention has been paid to date to the opto- In terms of relatively recent R & D literature, electronic techniques. However, Ring et al. (1965, Minnick (1967) provides a review of microcellular p. 33) point out that "it well may be that optical research, with emphasis upon techniques useful devices which do not appear at all suitable as binary for batch-fabricated circuit design; Bilous et al. computer elements may be very effective computing (1966) discuss IBM developments of large scale devices in the context of some other logic structure integration techniques to form monolithic circuit (e.g., majetrity logic, multivalued logics)." arrays, where on only nine chips it was possible to Reimann adds: "With the advent of the laser, replicate a specific System/360 computer model, efficientlight-emittingdiodes,andhigh-speed and, under RADC auspices, Savitt et al. (1967) have photodetectors, interest in the application of higher explored both language development and advanced speed opto-electronic circuits to digital logic has machine organization concepts in terms of large increased....We may in the future expect to scale integration (LSI) fabrication techniques 6'88 see opto-electronic circuits which will combine That is, in general, where integrated circuits based laser amplifiers with other high-speed semicon- on etched circuit board techniques had replaced ductor devices." (1965, p. 248). discrete components, the LSI techniques of fabri- Then we note that optoelectronic techniques may cation produce sheets of integrated logic compc also be used to attack some of the problems that nents as units.6.89 increasingly plague the circuit designer.68° Possi- To what extent do integrated fabrication tech- bilities for circumventing interconnection limitations niques hold promise for future developments in which become more severe as physical area per very large yet inexpensive memories? Rajchman component is reduced are also stressed by Reimann suggests that "the dominance of non-integrated (1965, p. 247). He states: "The possibility of signal memories is likely to be finally broken or at least connection between parts of the system without seriously challenged by integrated memories, of electrical or actual physical contacts arevery which the laminated-ferrite-diode and the super- attractive for integrated circuit techniques. With conductive-thin-sheet-cryotron memoriesare prom- optical signals, a totally new approach to the inter- ising examples." (Rajchman, 1965, p. 128.) And, connection of digital devices is possible." further, that"it appears certain that energetic Optoelectronic techniques as applied to the prob- efforts will continue to be devoted towards inte- lems of large, inexpensive memories are not only grated technologiesforlarger and less costly promising as such,681 they also may be used to memories, as this is still the single most important attack the noise problems still posed by integrated hardware improvement possible in the computer circuits.6.82 Thus Merryman says "one attractive art." (Rajchman, 1965, p.128.) Other advocates property of optoelectronic devices is their potential includeGross, Hudson '6'91Van Dam and for isolation; they can get rid of the noise that is Michener,692Pyke,693andConwayand generated when two subsystems are coupled. The Spandorfer.6.94 noise problem is even tougher in integrated circuit Hobbs says of silicon-on-sapphire circuits that systems, because the transformers used in tradi- their fabrication is suitable for large arrays and that tional methods of isolation are too bulky." (1965, they are indeed "promising, but presently being p. 52). pursued by only one company." (1966, p. 38.) Of active thin-film circuits, he concludes: "Potentially 6.2. Batch Fabrication and Integrated cheaper and easier to fabricate very large arrays. Circuits Feasibility is not proven and utilization much further away." (Hobbs, 1966, p. 38.) The same reviewer In very recent years, it has been claimed that continues: "Costs are expected to range between 3 integrated circuitry is the most significant advance and 5 cents per circuit in large interconnected in computer hardware technology since the introduc- circuit arrays...However, the ability to achieve tion of the transistor; 6.83 that it will bring important these costs is dependent upon the use of large inter- changes in the size, cost, reliability and speed of connected arrays of circuits and, hence, upon the system design components," and that advanced computer industry's ability to develop logical design high-speed techniques paradoxically also indicate and machine organization techniques permitting eventual lower costs.685 and utilizing such arrays." (Hobbs, 1966, p. 39.) Many potential advantages of increasedusage of Continuing R & D problems in terms of LSI LSI techniques are cited in the literature. These technology include those of packaging design 6.95 include, for example, applications in improved error detection and correction with respect to central processor unit speedor capacity perform- malfunctioning components; ".96 theproper balance ance, in system control and reliability, and in to be achieved between flexibility, redundancy, and content-addressable (associative) memory construc- maintenance or monitoring procedures, andques- tion and operation.6.86 Wilkes suggests that parallel- tions of segmentation or differentiation of functional ism achieved by use of these techniquesmay logic types.697 One example ofmany special prob- 25 lems is reported by Kohn as follows: "In all batch storage and recall concern first of all the problems fabricated memories, the problem of unrepairable of "main memory" that is, the preloaded, im- element failures is predominant...It is an open mediately accessible, information-recordingspace question how complex and expensive the additional allocated at any one time tonecessary system electronic circuits will be, which will disconnect the supervision and control, to user(s) programs and defective elements and connect thespare ones." data, and .o temporary workspace requirements. (Kohn, 1965, p. 132.) On the other hand, special It is to be noted that "this 'main'memory size is advantages of LSI techniques for self-diagnosis and related to the processing rate; the faster the arith- self-repair have been claimed.698 metic and logic units, the faster and larger thememory must be to keep the machine busy, or to enable itto solve problems without waiting for data." 6.3. Advanced Data Storage Developments (Hoagland, 1965, p. 53). In the area of advanced hardware, the prospects Further, "this incompatibility between logic and for much larger, much faster, and more versatilememory speeds has led to increased parallel opera- storage systems must of course be a major R & Dtion in processors and more complex instructionsas consideration. Current technological advances indi- an attempt to increase overall system capability." cate the desirability of increasing use of integrated (Pyke, 1967, p. 161). construction methods using ferrite aperture plates, As of current technology, main memoriesare still thin films, laminated-diode combinations, field-effect usually magnetic core, with typical capacities ofa transistors, and superconductive thin film systems, million bits and cycle times of aboutone micro- among other recent developments.699 For another second.61" One relatively recent exception is the example, possible applications of echoresonance NCR Rod Memory C -mputer, which is claimedto have "about the fastec mainmemory cycle time of techniques for microwave pulse delay lines that anycommercial would be suitable for high-speed memoriesare being computeryetdelivered 800 explored at the Lockheed Palo Alto Research Labor- nanoseconds." (Data Processing Mag. 7, No. 11, atory. (Kaplan and Kooi, 1966). 12 (Nov. 1965).) This is a thin-filmmemory, con- Advanced hardware developments for improvedstructed from beryllion-copper wires plated with data storage emphasize both higher speeds of access magnetic coating. 6.101 and readout and larger capacities at higher densities Petschauer lists the following trends whichmay of storage. There are the small capacity, ultra-high- be expected in magnetic memory developments in speed, memories of the read-only, scratchpad, and the near future: associative type. These typically supplement signifi- "1. Trend toward simple cell structures 2or 3 cantly larger capacity and slower speed "main wire arrays. memories". Next, there are continuing prospects "2. More automated assembly and conductor for high density, very large capacity stores. termination or batch-fabricated arrays. There is finally the question of R & D require- "3.More fully automated plane testing. ments inthe area where the development of "4. More standardization. "artificial" memories are designed to replicate,so "5.Extended use of integrated or hybrid circuits. far as possible, known neurophysiological phe- "6. Improved methods of packaging for stack and nomena. For example, Borsellino and his colleagues stack interface circuits to reduce packaging at the University of Genoa are studying physical- and assembly costs. chemical simulation, such as collagen "inemories", Reduced physical size." (Petschauer, 1967, in terms of possible mechanisms of axon action, p. 599). connectivity of pulses, and currents through mem- branes. (Stevens, 1968, p. 31). With respect to current prospects for much larger, We may thus conclude with Lickliderthat much faster main memories, Rajchman (1965) "insofar as memory media are concerned, current reviewspossibilitiesfor integrated construction research and development present many possibil- methods using ferrite aperture plates, thin films, ities. The most immediate prospects advanced for laminate-diode combinations, field-effect transistors, primary memories are thin magnetic films, coated and superconductive thinfilm cryotrons.6102It wires, and cryogenic films. For the next echelons, is noted further that "planar magnetic film memories there are magnetic disks and photographic films and offer many advantages for applicationsas main plates.Farther distant are thermoplastics andcomputer storage units in the capacity range of photosensitive crystals. Still farther awayalmost 200K to 5M bits." (Simkins, 1967,p. 593), and that whollyspeculativeareproteinmolecules and "perhaps the most significant system advantage otherquasi-living structures."(Licklider,1965, available to users of plated magnetic cylindrical pp. 63-64). thin film memory elements isa nondestructive readout capability. For mainmemory use, NDRO 6.3.1. Main Memories with equal Read-Write drive currents is most advantageous. It allows the greatest possible flexibility Questions of advanced tehcnological develop- of organization and operation. For maximumecon- ments related to data and program information omy, many memory words (or bytes) may be ac- 26 cessed by a single word drive line without need for macro instructions. Further, "Picoinstructions are more than one set of sense amplifiers and bit stored at constant radii upon a MYRA disk, in the current drivers. The set contains only the number of proper order to perform the desired task. The amplifiers needed to process the bits of one word advantages of the MYRA element are that the pico- (or byte) in parallel." (Fedde, 1967, p. 595). Simpson instructionsareautomaticallyaccessedinse- (1968) discusses the thin film memory developed at quence ..." 6.112 Holographic ROM possibilities Texas Instruments.6.1°3 are also under consideration.6.113 Nevertheless, the known number of storage In the area of associative, or content-addressable elements capable of matching' ultrafast processing memories ,6314 advanced hardware developments to and control cycle times (100-nanosecond or less) are date have largely been involved in processor design relatively few,6.1°4 and there are many difficulties to and provision of small-scale auxiliary or "scratch- be encountered in currently available advanced pad" memories rather than for massive selection- techniques.6106 Some specific R & D requirements retrieval and data bank management applications 6.115 indicated in the literature include materials research "Scratchpad" memories, also referred to as "slave" to lower the high voltages presently required for memories, e.g., by Wilkes (1965),6.116 are defined by light-switching inoptically addressed memories Gluck (1965) as "small uniform access memories (Kohn, 1965),63°6 attacks on noise problems in inte- with access and cycle times matched to the clock grated circuit techniques (Merryman, 1965),6.1°7 and of the logic." They are used for such purposesas the provision of built-in redundancy against element reducinginstruction-accesstime,formicro- failures encountered in batch fabrication techniques programming, for buffering of instructions or data (Kohn, 1965). In the case of cryotrons used for that are transferable in small blocks (as in the memory design, Rajchman (1965) notes that the "four-fetch" design of the B 8500),6.117 for storage of "cost and relative inconvenience of the necessary intermediate results, as table lookup devices,6.118 cooling equipment is justified only for extremely as index registers and, to a limited extent, for large storage capacities" (p. 126), such as those content addressing. 6.119 extending beyond 10 million bits, and Van Dam and Another example is the modified "interactive" Michener(1967)concur.6.1°8Considerationsof cell assembly design of content-addressable memory "break-even" economics with respect to cryogenic- where entries are to be retrieved by coincidence of element memories such as to balance high density a part of an input or query pattern with a part of storage and high speed access against the "cooling" stored reference patterns, including other variations costs has been assessed at a minimum random- on particular match operations (Gaines and Lee, access memory requirement of 107 bits.6109 As of 1965).6.12° In addition, we note developments with 1967-68, however, practical realizations of such respect to a solenoid array 6.121 and stacks of plastic techniques have been largely limited to small-scale, card resistor arrays,6.122 both usable for associative special-purpose auxiliary and content-addressable memory purposes; the GAP (Goodyear Associative memories, to be discussed next. Processor),6.123the APP(A ssoci ativ eParallel Processor) described by Fuller and Bird (1965),6.124 6.3.2. High-Speed, Special-Purpose, and Associative or the ASP (A ssoci ati on-Storing Processor) machine Content-Addressable Memories organization,6.126 and various approaches which Small,high-speed,special-purposememories compromise somewhat on speed, including bit- have been used as adjuncts to main memories in rather than word-parallel searching 6.126 or the use computer design for some years.6.11° One major of circulating memories such as glass delay lines.6.127 purpose is to provide increased speed of instruction Cryogenic approaches to the hardware realization access or address translation, or both. The "read- of associative memory concepts have been under only-stores" (ROS) in particular represent relatively investigation since at least the mid-1950's (Slade recent advances in "firmware," or built-in micro- and McMahon, 1957), while McDermid and Peterson programmingpii (1961) report work on a magnetic core techniqueas It is noted that "the mode of, implementing ROM's of 1960. However, the technology for developing spans the art, from capacitor and resistor arrays and high-speed reactivity in these special-purpose mem- magnetic core ropes and snakes to selectively ories has been advanced in the past few years. On deposited magnetic film arrays." (Nisenoff, 1966, the basis of experimental demonstration, at least, p.1826.) An Israeli entry involves a two-level there have been significant advances with respect memory system with a microprogrammed "Read toparallel-processing, associative-addressing, in- Only Store" having an access time of 400 nano- ternal but auxiliary techniques in the form of mem- seL 9nds. (Dreyer, 1968.) A variation for instruction- ories built-into some of the recently developed large access processes is the MYRA (MYRi Aperture) computer systems 6328 ferrite disk described by Briley (1965). This, when The actual incorporation of such devices, even accessed, produces pulses in sequential trains on if of somewhat limited scale, in operationalcom- 64 or more wires. A macro instruction is addressed puter system designs is of considerable interest, to an element in the MYRA memory which then whether of 25- or 250-nanosecond performance. For produces gating signals for the arithmetic unit and example, Ammon and Neitzert report RCA experi- signals for fetching both operands and the next ments that "show the feasibility of a 256-word 27

376-411 0 -'70 - 3

lEalLi,44411,1112-,2d1111(1, fWd44f 4 scratchpad memory withan access time of 30 nano- nanosecond cycle times 8.145 It is noted, however, seconds...The read/write cydle time, however, that "a considerable amount of development work is will still be limited by the amplifierrecovery so that still required to establish the handling, assembly, with the best transistors available itappears that and packaging techniques." (Abbas et al., 1967, 60 nanoseconds are required". (1965,p. 659). RCA p. 311). developments also includea sonic film memory in A plated wire randomaccess memory is under which thin magnetic films and scanning strainwaves development by UNIVAC for the Rome Air Develop- are combined for serial storage of digital informa- ment Center. "The basic memory module consists tion.6329 of 107 bits; the mechanical packagecan hold 10 Crawford et al. (1965) have claimed thatan IBM modules. The potential speed isa 1-to-2 micro- tunnel diode memory of 64 48-bit words anda read/ second word rate....Ease of fabrication has been restore or clear/write cycle time of less than 25emphasized in the memory stack design. These nanoseconds was "the first completememory sys- factors, together with the low plated wire element tem using any type of technology reported in this cost, make an inexpensive mass plated wire store a size and speed range". (p. 636).6330 Then there isan distinct possibility." (Chong et al., 1967,p. 363).6.146 IBM development ofa read-only, deposited magnetic RADC's interests in associative processingare also film memory, having high-speed read capability reflected in contracts with Goodyear Aerospace (i.e., 19ns access time) and promisingeconomics Corp., Akron, Ohio, for investigation and experi because the technique is amenableto batch fabrica- mental fabrication of associative memories and tion."' (Matick et al., 1966). processors. (See, for example, Gall, 1966). Catt and associates of Motorola describe "an integrated circuit memory containing 64 words of 8 bits per word, which is compatible in 6.3.3. High-Density Data Recording and Storage respect to Techniques both speed and signal level with high-speedcurrent- mode gates. The memory hasa nondestructive read Another importantfield of investigation with cycle of 17 nanoseconds anda write cycle of 10 nano- respect to advanced data recording, processing, and seconds without cycle overlap." (Cattet al., 1966, storage techniques is that of further development of p. 315).6.132 Anacker et al. (1966) discuss 1,000-bit high-density data recording media and methods and film memories with 30 nanosecondaccess times.6.133 bulk storage techniques, including block-oriented Kohn et al. (1967) have investigateda 140,000 bit, randomaccessmemories.6.147Magnetictech- nondestructiveread-outmagneticfilm memory niquescores, tapJs, and cardscontinue to be that can be read witha 20-nanosecond read cycle pushed toward multimillion bit capacities.6.148 A time, a 30-nanosecond access time, anda 65-nano- single-wall domain magneticmemory system has second write time. More recently, IBM hasan- recently been patented by Bell Telephone Labora- nounced a semi-conductormemory with 40 nano- tories.6.143 In terms of R & D requirements for these second access.6.134 techniques, further development of magnetic heads, Memories of this type thatare of somewhat larger recording media, and means for track location has capacity but somewhat less speed (in the 100-500 been indicated,6.13° as is also the case for electron nanosecond range) are exemplified by suchcom- or laser beam recording techniques.6.131 Videotape mercially-announced developmentsasthose of developments are also to be noted.6.132 Electronic Memories,6333 Computer Control Com- In addition to the examples of laser, holographic, pany,6336 and IBM 6.13' Thus, Werneret al. (1967) and photochromic technologies applied to high describe a 110-nanosecond ferritecore memory with density data recording previouslygiven, we may a word capacity of 8,192 words,6.138 while Pugh et al. note some of the other advanced techniques that (1967) report other nig developments involvinga are being developed for large-capacity, compact 120-nanosecond film memory of 600,000-bit capacity. storage. These developments include the materials McCallister and Chong (1966) describean experi- and media as well as techniques for recording with mental plated wire memory system of 150,000-bit light, heat, electrons, and laser beams. In particular, capacity with a 500-nanosecond cycle time anda 300 - "a tremendous amount of research work is being nanosecond access time, developedat UNIVAC.6.139 undertaken in the area of photosensitive materials. Another UNIVAC development involves planar thin Part of this has been sparked by the acute shortage fiims.o.140 A16,384-word, 52-bit, planar filmmemory of silver for conventional films andpapers. In with half-microsecondor less, (350 nanosecond) October, more than 800 people attenda sym- cycle time, under development at Burroughs lab- posium in Washington, D.C., on Unconventional oratories for some years, has been described by Photographic Systems. Progresswas described in a Bittman (1964).6.141 Other recent developments have number ofareas, including deformablefilms, been discussed by Seitzer (1967) 6.142 and Raffel electrophotography, photochromic systems,uncon- et al. (1968),6.143 among others. ventionalsilversystems,and photopolymers." For million-bit and higher capacities, recent IBM (Hartsuch, 1968, p. 56). investigations have been directed toward theuse of Examples include the General Electric Photo. "chain magnetic film storage elements" 6.144in charge,6.133 the IBM Photo-Digital system,4the both DRO and NDRO storage systems with 500- UNICON mass memory,6.133a system announced 28 by Foto-Mem InC.6.156 and the use of thin dielectric information density with adequate redundancy, but filmsat Hughes Research Laboratories.6"87 At the subject is one that attracts considerable re- Stanford Research Institute,a program for the U.S. search interest." (Smith, 1966, p. 1298), Army Electronics Command is concerned with in- Materials and media for storage are also sub- vestigations of high-density arrays of micron-size jects of continuing R & D concern in both the storage elements, which are addressed by electron achievement of higher packing densities with fast beam. The goal is a digital storage density of 108 direct access and in the exploration of prospects bits per square centimeter.0.188 for storage of multivalued data at a single physical Still another development is the NCR heat-mode location. For example: "A frontal attack on new recording technique. (Carlson and Ives,1968). materials for storage is, crucial if we are to use the This involves the use of relatively low power CW inherent capability of the transducers now at our lasers to achieve real-time, high-resolution (150: 1) disposal to write and read more than 1 bit of data recording on a variety of thin films on suitable sub- at 1 location.. . strates."88 In particular, microimage recordings can "One novel approach for a multilevel photo- be achieved directly from electronic character- graphic store now being studied is the use of color generation devices.6.'60 Newberry of General Elec- photography techniques to achieve multibit storage tric has described an electron optical data storage at each physical memory location. .Color film technique involving a 'fly's eye' lens system for can store multilevels at the same point because both which a "a packing density of 108 bits per square intensity and frequency can be detected." (Hoagland, inch has already been demonstrated with 1 micron 1965, p. 58). beam diameter." (1966, p. 727-728). "An experimental device which changes the color Then there is a new recording-coding system, of a laser beam at electronic speeds has been developed...IBM scientistsbelieveitcould from Kodak, that uses fine-grained photOgraphic lead to the development of color-coded media, diffraction grating patterns, and laser light computer sources.8.181 As a final example of recent record- memories with up to a hundred million bits of infor- ing developments we note that Gross (1967) has mation stored on one square inch of photographic described a variety of investigations at Ampex, film." (Commun. ACM 9, 707 (1966).) including color video recordings on magnetic film Such components and materials would have plated discs, silver halide film for both digital and extremely high density, high resolution character- analog recordings, and use of magneto-optic effects istics. One example of intriguing technical possi- for reading digital recordings.8"82 bilities is reported by Fleisher et al. (1965) in terms of a standing-wave, read-only memory wheren color Areas where continuing R & D efforts appear to be sources might provide n information bits, one for indicated include questions of read-out from highly each color, at each storage location.6,'66 These compact data storage,6'163 of vacuum equipment in authors claim that an apparently unique feature of the case of electron beam recordin 0.164 and of this memory would be a capability for storing both noise in some of the reversible media.8.'" Then it is digital and analog (video) information 0167 and that noted that "at present it is not at all clear whatcom- parallel word selection, accomplished by fiber-optic promises between direct image recording and holo- light splitting or othermeans, would be useful in graphic image recording will bestpreserve high associative selection and retrieva1.8.188

7. Debugging, On-Line Diagnosis, Instrumentation, and Problemsof Simulation

Beyond the problems of initial design of informa- Special, continuing, R & D requirementsare raised tion processing systems are those involved in the in the situations, first, of checking out very large provision of suitable and effective debugging, self- programs, and secondly, of carrying out checkout monitoring, self-diagnosis, and self-repair facilities operations under multiple-access, effectuallyon- in such systems. Overall system design R & D line, conditions." In particular, the checkout of requirements are, finally, epitomized in increased very large programs presents special problems." concern over the needs for o -line instrumentation, simulation, and formal modelling of information 7.1. Debugging Problems flows and information handlingprocesses, and with the difficulties so far encountered in achieving Program checkout and debuggingare also prob- solutions to these problems. In turn,many of these lems of increasing severity in terms of multiple- problems are precisely involved in questions of access systems. Head states that "testing of many systems evaluation. non-real-time systemseven large _oneshas all It has been cogently suggested that thearea of too often been ill-planned and haphazard with aids to debugging "has been givenmore lip service numerous errors discovered only after cutover. ... and less attention than any other" 7.1 in considera- In most real-time systems, the prevalence oferrors tions of information processing systems design. after cutover, any one of which could force the 29 system to go down, is intolerable." (1963, p. 41.) over when a hardware error occurs. During time- Bernstein and Owens (1968) suggest that conven- sharing, the error must be analyzed, corrected if tional debugging tools are almost worthless in the possible, and the user or users affected must be time-sharing situation and propose requirements notified. For all those users not affected, no sig- for an_improyed debugging support system.7.4 nificant interruption should take place.". On-line debugging provides particular challenges to the user, the programmer and the system de- 7.2. On-Line Diagnosis and Instrumentation signer.7.5 It is important that the console provide versatile means of accomplishing system and pro- Self-diagnosis is an important area of R & D con- gram self-diagnosis, to determine what instruction cern with respect both to the design and the utiliza- caused a hang-up, to inspect appropriate registers tion of computer systems.7.11 In terms of potentials In a conflict situation, and to display anticipated for automatic machine -self- repair, it is noted that results of a next instruction before it is executed. "a self-diagnosable computer is a computer which A major consideration is the ability to provide inter- has the capabilities of automatically detecting and pretation and substitution of instructions, with isolating a fault (within itself) to a small number traps, from the console. A recent system for on-line of replaceable modules." (Forbes et al., 1465, p. debugging, EXDAMS (EXtendable Debugging and 1073).7.12 To what extent can the machine itself be Monitoring System), is described by Balzer (1969).7.° used to generate its own programs and procedures? Aids to debugging and performance evaluation Forbes et al. suggest that: "If the theory of self- provided by a specific system design should therefore diagnosing computers is to become practical for a include versatile features for address traps, in- family of machines, further study and development struction traps, and other traps specified by the of machine generation of diagnostic procedures is programmer. For example, if SIMSCRIPT programs necessary." (1965, p. 1085). are to be run, a serious debugging problem arises Several different on-line instrumentation* tech- because of the dynamic storage allocation situation niques have been experimentally investigated by where the clients needs to find out where he is and Estrin and associates (1967), by Hoffman (1965), provide dynamic dumping, e.g., by panel interrupt Scherr (1965) and by Sutherland (1965), among without halting the machine. Programmers checking others.7.13 Monitoring systems for hardware, soft- out a complex program need an interrupt-and-trap- ware, or both are described, for example, by to-a-fixed location system, the ability to bounce out Aviiiensis (1967, 1968),7.14 Jacoby (1959),7.15 and of a conflict without being trapped in a halt, to jump Wetherfield (1966),7.1° while a monitoring system for if a program accesses a particular address, to take the multiplexing of slow-speed peripheral equipment special action if a data channel is .tied up for ex- at the Commonwealth Scientific and Industrial pected input not yet received, or to jump somewhere Research Organization in Australia is described by else on a given breakpoint and then come back to Abraham et al. (1966). Moulton and Muller (1967) scheduled address, e.g., on emergence of an over- describe DITRAN (Diagnostic FORTRAN), a com- flow condition.7.7 piler with extensive error checking capabilities that Problems ofeffectivedebugging,diagnostic, can be applied both at compilation time and during and simulation languages are necessarily raised.7.8 program execution, and Whiteman (1966) discusses For example, McCarthy etal.report: "In our "computer hypochondria".7.17 opinion the reduction in debugging time made pos- Fine et al. (1966) have developed an interpreter sible by good typewriter debugging languages and program to analyze running' programs with respect to adequate access to the machine is comparable to determining sequences of instructions between page that provided by the use of ALGOL type languages calls, page demands by time intervals, and page for numerical calculation." (McCarthy et al., 1963, demands by programs. In relatively early work in p. 55). Still another debugging and diagnostic R & D this area, Licklider and Clark report that "Program requirement is raised with respect to reconfigura- Graph and Memory Course are but two of many tions of available installations and tentative evalua- possible schemes for displaying the internal proc- tions of the likely success of the substitution of one esses of the computer. We are working on others configuration for another.7.° that combine graphical presentation with symbolic In at least one case, a combined hardware-soft- representation...By combining graphical with ware approach has been used to tackle anothersymbolic presentation, and putting the mode of special problem of time-shared, multiple-user sys- combination under the operator's control via light tems, that of machine maintenance with minimum pen, we hope to achieve both good speed and good interferencetoongoingclientprograms. The discrimination of detailed information." (1962, p. STROBES technique(for Shared-time-repair of 120). Howlver, Sutherland comments that: "The big electronic systems) has been developed at the information processing industry is uniquely wanting Computation Center of the Carnegie Institute of in good instrumentation; every other industry has Technology.7.10 This type of development is of sig- meters, gauges, magnifiersinstruments to measure nificance because as Schwartz and his co-authors *Instrumentation" in this context means diagnostic and monitoring procedures which are applied to operating programa in a "subject" computer as they are being report (1965, p. 16): "Unlike more traditional sys- executed in order to assemble records of workload, system utilisation, and other tems, a time-sharing system cannot stop and start siormilar data. 30 and record the performance of machines appropriate techniques as an alternative, "Because of the large to that industry." (Sutherland, 1965, p. 12). More number of random variablesmany of which are effective on-line instrumentation techniques are thus interdependentthat must be taken into account in urgently required, especially for the multiple-access a completely general treatment of time-sharing processing system. operation, one cannot expect to proceed very far Huskey supports the contentions of Sutherland with analyses of the above nature. Thus, itseems and of Amdahl that; "Much more instrumentation of clear that simulation must also be used to study on-line systems is needed so that we know what is time-shared computer operation." (Schwartz et al., going on, what the typical user does, and what the 1965, p. 21). A 1967 review by Borko reaches similar variations are from the norms. It is only with this conclusions,7'20 information that systems can be 'trimmed' so as to optimizeusefulnesstothecustomerarray." 7.3. Simulation (Huskey, 1965, p. 141). Sutherland in particular points out that plots of The on-going analysis and evaluation of informa- times spent by the program in doing various sub- tion processing systems will clearly require the tasks, can tighten up frequently used program and further development of more sophisticated andmore sub-routine loops and thus save significant amounts accurate simulation models than are available of processor running-time costs." He also refers today."' Special difficulties are to be noted in the to a system developed by Kinslow in which a pic- case of models of multiple access system where torial representation of "which parts of memory "the addition of pre-emptive scheduling complicates were 'occupied' as a function of time for his time- the mathematics beyond the point where models sharing system. The result shows clearly the small can even be formulated" (Scherr, 1965, p. 32) and spaces which develop in memory and must remain in that of information selection and retrieval applica- unused because no program is short enough to fit tions where, as has been frequently charged, "no into them." (Sutherland, 1965, p. 13). In general, it accurate models exist". (Hayes, 1963, p. 284). is hoped that such on-line instrumentation tech- In these and other areas, then, a major factor is niques will bring about better understanding of the the inadequacy of present-day mathematical tech- interactions of programs and data within the proc- niques.2.22 In particular, Scherr asserts that "simula- essing system.7.'" tion models are required because the level of detail Improved techniques for the systematic analysis necessary to handle some of the features studied is of multiple-access systems are also needed. As beyond thescope of mathematicallytractable Brown points out: "The feasibility of time-sharing models." ( Scherr, 1965,p. 32). The importance of depends quite strongly upon not only the time- continuing R & D efforts in this area, even if they sharing procedures, but also upon...the following should have only negative results, has, however properties, characteristic of each program when it been emphasized by workers in the field.7.23 is run alone: Thus, for example, at the 1966 ACM-SUNY Conference, "Professor C. West Churchman... (1) The percentage of time actually required for pointed to the very large [computer] models that execution of the program.. can now be built, and the very much larger models (2) The spectrum of delay times during which the that we will soon be able to build, and stated that program awaits a human response... the models are not realistic because the quality (3) A spectrum of program execution burst of information is not adequate and because the right lengths... questions remain unasked. Yet he strongly favored A direct measurement of these properties is diffi- the building of models, and suggested that much cult; a reasonable estimate of them is important, information could be obtained from attempts to however, in determining the time-sharing feasibility build several different and perhaps inconsistent of any given program." (1965, p. 82). However, most models of the same system.' (Commun. ACM 9, of the analyses implied are significantly lacking to 645 (1966).) date, although some examples of benefits to be We are led next, then, to problems of simulation. anticipated are given by Cantrell and Ellison (1968) There are obvious problems in this area also. First and by Campbell and Heffner (1968). there is the difficulty of "determining and building Schwartz et al. emphasize that "another research- meaningful models" (Davis, 1965, p. 82), especially able area of importance to proper design is the where a high degree of selectivity must be imposed mathematical analysis of time-shared computer upon the collection of data appropriately representa- operation. The object in such an analysis is to pro- tive of the highly complex real-life environments and vide solutions to problems of determining the user processes that are to be simulated.7.24 capacity of a given system, the optimum values for Beyond the questions of adequate selectivity in the scheduling parameters (such as -:uantum size) simulation-representation of the phenomena,op:ma- to be used by the executive systen-.nd, in general, tions,andpossiblesystemcapabilitiesbeing the most efficient techniques for sequencing the modelled are those of the adequacy of the simulation object programs." (Schwartz et al., 1965,p. 21). languages as discussed by Scherr, Steel, and Continuing, they point to the use of simulation others.7.22 Teichroew and Lubin presenta compre- 31 hensivesurvey of computer simulation and applications, languages separate operations are carriedout concurrently or with tables of comparativecharac- inparallel, and that the teristics,as of 1966.7." In addition,IBM has simulation technique provideda bibliography on simulation, also requires a serial sequencingof these operations. 1966. as of Dependingupon the choice of whichone of the theoretically concurrentoperations is processed Again, as in thearea of graphic input manipula- first in the sequentializing tion and output, the fieldof effective simulation procedure, the results specific R & D has of the simulationmay be significantly different in requirements for improvedand one case than in another.7.3° more versatile machine models andprogramming languages. Clancy andFineberg suggest that "the For example, the SL/1language being developed very number and diversity of languagessuggests at the University of Pisa underCaracciolo di Forino that the field [of digitalsimulation languages] suffers (1965) is based inpart on SOL (Simulation-Oriented from a lack of perspectiveand direction." (1965, Language,see Knuth and McNeley, 1964) andin p. 23). part on SIMULA (the ALGOLextension 'developed The area of improvedsimulation languages isone by 0. J. Dahl, of theNorwegian Computing Center, that has a multipleinteraction between softwareand Oslo).7,31 A secondversion, SL/2,now under hardware, especially wherea computer is to be used development, will provideself-adapting featuresto to simulate anothercomputer, perhaps one whose optimize the system. Caraccioloemphasizes that, design is notyet complete 7.27 or to simulatemany for any set of deterministicprocesses that are to be different scheduling,queuing and storage allocation applied simultaneously,but where problems of alternativesintime-sharedsystems(see,for incompatibilitymay arise, the problemscan be example, Blunt 1965). Suchproblems are also dis- reduced toa set of probabilisticprocesses. Other- cussed by Scherr (1965)and by Larsen and Mano wise, if one sequentializesparallel, concurrent (1965),among others, while Parnas (1966) processes actually dependent a describes upon the order of modificationof ALGOL (SFD-ALGOL,for sequentialization, then hiddenproblems of incom- "System Function Description") applicableto the patibility may vitiate theresults obtained. simulation of synchronous systems. Despite difficulties, however, However, thereare difficult current problems in progress has been and is being made. Thuscomputer simulation has that languages suchas SIMSCRIPT do not take been investigated advantage of the modularityof many processing as a means of system simulation for determination ofprobable costs and benefits systems, that conditional schedulingof sequences in advance of major of events is extremelydifficult 7.28 and that "weare investments in equipmentor still plagued by procedures.7.32 Then,as reported by Gibson (1967), our inability to program for simul- simulation studies have taneous action, even for thescheduling of large been used to determine units in a computing system." that block transfersof 4 to 16 words willfacilitate (corn, 1966,p. 232). reduction of effective internal In addition, for simulationand similar applications, access times to a few heuristic or ad hoc nanoseconds. Otherprograms to simulate digital programming facilitiesmay be data processing,time-shared system required. Thus, "acomputer program which isto and the like, performance, serve as a model must be ableto have well-organized are discussed by Larsen and Mano (1965) and by Scherr(1965). Simulation studies yet manipulatable datastorage, easily augmentable terms of multiprocessor in and modifiable. Theprogram must be self-modifying systems are represented by in a similarly organized Lindquist et al. (1966)7.33 and by Huesmannand way. It should be able to Goldberg (1967).7.34 handle large blocksof data orprogram routines by specification ofmerely a name." (Strum,1965, Other advantages fromresearch and development p. 114.) efforts to be anticipatedfrom computer simulation For simulationsor testings with controls, and with- experiments are those oftransfer of applications out discernible interruptionor reallocation of normal from a givencomputer to another not yet installedor servicing of clientprocessing requests, compilers available,7" advancementsin techniques of pic- must be availablethatwill transformqueries torial data processing andtransmission,7.36 advance expressed inone or more commonly available appraisals of performance oftime-shared systems,737 customer languages to the language(s)most effec- and investigations ofprobable performanceof tively used by thesubstituted experimentalsystem adaptive recognitionsystems.7.38 and to the format(s)available ina master data base. Finally, we note Then there prospects for system simulation are problems in the development ofan as a means of evaluation and ofredesign, including appropriate "scenario",or sequence of events to be thealteration of schedulingprioritiesto meet simulated.7.29 Burdick andNaylor (1966) providea changing requirementsof the system's clientele. survey account of the problemsof design and Three examples from theliterature areas follows: analysis of computersimulation experiments. The problems of effective (1) "Use ofa simulator permits the installationto simulation of complex, continue running its interdependentprocessesareanother area of programs as reprogram- increasing ming proceedson a reasonable schedule." concern. Suppose, for example, thatwe (Trimble, 1965, are seeking to simulatea process in which p. 18). many (2) "Effectiveresponse time simulationcan be 32 easily modified to provide operatingcosts of be developed: "The benefits and drawbacks of retrieval." (Blunt, 1965,p. 9). empirical data gatheringvs. simulation vs. mathe- (3) "When large systemsare being developed matical analysis are well documented. Whatwe another set of programs is involvedto perform would really like to be ableto do is a little of all a function not required for simpler situations. three, back and forth, untilour gradually increasing These are the simulation and analysispro- comprehension of the problem becomes the desired grams for system evaluation and for semi- solution." (Greenberger, 1966,p. 347). Similarly, it automated systems having a human com- may be claimed that simulation models "... are ponentsystem training." (Steel, 1965, p. 232). often cumbersome and difficultto adapt to new On the other hand, as Daviswarns: "It is obvious configurations, with results of somewhat uncertain that there is some threshold beyond whichthe real interpretation due to statistical sampling variability. environment is too complex to permit meaningful Ideally, simulation and analytic techniquesshould simulation." (1965, p, 82). For the future, therefore, supplement each other, for each approach hasits a system of multiple-working-hypotheses might well advantages." (Gayer, 1967,p. 423).

8. Conclusions

As we have seen, major trends in input/output, (4) "Today, and toan even greater extent storage, processor, and programming design relate tomorrow, the use of multiple functional units to multiple access, multiprogrammed, and multi- within the information processingsystem, the processor systems. On-line simulation, instrumenta- multiplexing of input and outputmessages, and tion, and performance evaluation capabilitiesare the increased' use of softwareto permit multi- necessary in order to effectively measure and test programming will requiremore subtle meas- proposed techniques, systems, and networksof uresto evaluate a particular system's perform- broad future significance to improved utilizationof ance." (Nisenoff, 1966,p. 1828.) automatic data processing techniques. (5) "Broad areas for further researchare indicated...Comparative experimental studies We may therefore close this reporton overall of computer facility performance, suchas system design considerations with the following online, offline, and hybrid installations,sys- quotations: tematically permuted against broad classes of program languages (machine-oriented, pro- (1) "In rating the completeness, clarity, and cedure-oriented, and problem-oriented lan- simplicity of the system diagnostics,com- guages),andrepresentativeclassesof mand language and keyboard procedures,we programming tasks." (Sackmanet al., 1968, found their 'goodness'was inversely related p. 10), and to the running efficiency of the system... (6) "Improved methods of simulation,optimizing Systemdevelopersshouldexaminethis techniques, scheduling algorithms, methods condition to determine whether inefficient of dealing with stochastic variables, t'heseare execution is an inherent feature of system[s] the important developments thatare pushing supplying complete and easily understood back the limits ofour ability to deal with very diagnostics, or a function of the specific large systems." (Harder, 1968, interests and prejudices of the developers." p. 233.) (O'Sullivan, 1967, p. 170). Finally we note that the problems of theinformation processing system designer, then,are today (2) "An engineer who wishes toconcern himself aggrevated not only by networking, time-sharing, with performance criteria in the synthesis of time-slicing, multiprocessor and multiprogramming new systems is frustrated by the weakness potentialities, but also by critical questions involving of measurement of computer system behavior." the values and the costs of maintaining theintegrity (Estrin et al., 1967, p. 645.) of privileged files. By the terminology "privileged files", we suggest the interpretation ofall data (3) "The setting up of criteria of evaluation... stored in a machine-usefulsystem that may have demands user participation and providesan varying degrees of privacy,confidentiality,or indication of whether theuser understands the securityrestrictionsplaced upon unauthorized reason for the system, the role of the system access. Some of the background considerations and his responsibilitiesas a prospective sys- affecting both policy and design factorswill be tem user." (Davis, 1965, p. 82.) discussed in the next report in this series.

33 Appendix A. BackgroundNotes on Overall SystemDesign Requirements

In this Appendix we present further discussionand background material intendedto highlight currently identifiable research and developmentrequirements in the broad field of thecomputer and information sciences, with emphasisupon overall system design considerations withrespect to information processing systems. A number of illustrativeexamples, pertinent quotations from theliterature, and references to current R and D effortshave been assembled. These backgroundnotes have been referenced, as appropriate, in thesummary text. 1. Introduction 1.1 There are certain obvious difficultieswith In general, controversial opinionsexpressed or respect to the organization of material fora series of implied in any of thereports in this series are the reports on research and development requirements sole responsibility of the author(s) ofthat report and in the computer and information sciencesand tech- are not intended in any way to represent the official nologies. These problemsstem from the overlaps policies of the Center for Computer between functional Sciences and areas in which man-machine Technology, the National Bureau of Standards,or interactions of both communication and control the Department of Commerce. However,every are sought; the techniques, tools, and instrumenta- effort has been made to buttress potentiallycontro- tion available to achieve such interactions, andthe versialstatethentsorimplications wide variety of application eitherwith areas involved. direct quotations or with illustrative examplesfrom The material that has been collected andreviewed the pertinent literature in the field. to date is so multifaceted andso extensive as to It isespecially to be noted that the refer- requireorganizationintoreasonablytractable ences and quotations included in the text of this (but arbitrary)subdivisions. Having considered report, in the corroborative background notes,or in some of the R and D requirements affecting specific the bibliography,are necessarily highly selective. Boxes shown in Figure 1 (p. 2) in previousreports, Neither inclusionnor citation is intended in any we will discuss here some of the overall system way to represent an endorsement ofany specific design considerations affectingmore than one of commercially available deviceor system, of any the processes or functions shown in Figure1. particular investigator's results withrespect to those Other topics to be covered inseparate reports in of others, or of the objectives ofprojects that are this series will include specific problemsof informa- mentioned. Conversely, omissionsare often in- tion storage, selection and retrievalsystems and the advertent and are inno sense intended to imply questions of maintaining the integrity ofprivileged adverse evaluations of products, materialsand files (i.e., some of the backgroundconsiderations media, equipment, systems, projectgoals and with respect to the issues of privacy,confidentiality, project results, or of bibliographic referencesnot and/or security in thecase of multiply-accessed, included. machine-based files, data banks, andcomputer- There will be quite obviousobjections to this communication networks). necessary selectivity from readers whoare also In general, the plan of attack in eachindividual R & D workers in the fields involved report in the series will be to outline in relatively as to the repre- sentativeness of cited contributions fromtheir own short discursive text the topics ofconcern, sup- work or that of others. Suchcriticisms are almost plemented by backgroundnotes and quotations and inevitable. by an appendix giving the bibliographic Nevertheless, these reportsare not citations of intended to be state-of-the-artreviews as such, but, quoted references. It is planned, however,that there rather, they are intendedto provide provocative will be a comprehensivesummary, bibliography, suggestions for further R & D efforts. and index for the seriesas a whole. Selectivity must also relate to a necessarily arbitrarycut-off Since problems of organization,terminology, and date in terms of the literature covered. coverage have all been difficult in the preparation of These reports, subjectto the foregoing caveats, this series of reports, certain disclaimersand obser- are offered as possible contributions to the under- vations with respect to thepurpose and scope of this standing of the generalstate of the art, especially report, its necessary selectivity, and the problems with respect to long-range research of organization and emphasis possibilities ina are to be noted. Ob- variety of disciplines thatare potentially applicable viously, the reviewer's interests andlimitations will to information processing problems. The emerge at least indirectly in terms of the selectivity reports that has been applied. are therefore directed to a varied audienceamong whom are those who plan, conduct,and support 35

71i T1:aLlutl resew in these varied disciplines. Theyare also interruptibility, and interlacings of computerpro- addressed to applications specialists whomay hope grams.Itis assumed, further, that the overall eventually to profit from the results of currentre- scheme involves hierarchies of systems, devices search efforts.Inevitably, there must besome and procedures, that processing involves multistep repetitions of the obvious or over-simplifications of operations, and that multimode operation ispos- certain topics for some readers, and there must also sible, depending on job requirements, prioror be some too brief or inadequately explained discus. tentative results, accessibility, costs, and the like. sions on other topics for these and other readers. It should be noted, next, that techniquessug- What is at best tutorial forone may be difficult for gested for a specific systemmay apply to more than another to follow. It is hoped, however, that the one operational box or function shown in the notesandbibliographiccitationswillprovide generalized diagram of Figure 1. Similarly, ina sufficient clues for further follow-upas desired. The specific system, the various operationsor processes literature survey upon which this report is based may occur in different sequences (including itera- generally covered the period from mid-1962to mid- tions) and several different onesmay be combined 1968, although a few earlier anda few later refer- in various ways. Thus, for example, questions of ences have also been included as appropriate. remote console design may affect original item input, 1.2Certain features of the information flow and input of processing service requests, output, and and process schema of Figure 1are to be noted. It is entry of feedback information from the user or the assumed, first, that the generalized information system client. The specific solutions adopted may be processing system should provide for automatic implemented in each of these operationalareas, or access from and to many users at many locations. combined into one, e.g., by requiring all imputs This implies multiple inputs in parallel,system and outputs to flow through thesame hardware.

2. Requirements and Resources Analysis

2.1"The single information flow concept... their level and for controlling the libraryone level is input-oriented. The system is organizedso that down, and so forth. The flexibility ofa distributed essential data are inserted into acommon reservoir system is an outgrowth of the ability of each of the through point-of-origin input/output devices. User lower executive programs to organize itsprogram requirements are then satisfied from this reservoir on the basis of separate inputs reaching it directly." of fundamental data about transactions. (Bennett, 1964,pp. 104-106). "Thus, the single information flow concept is "By a distributed implementation ofan information characterized by random entry of data, directaccess service system we mean that the dataprocess- to data in the system, and complete real-time ing activity is carried out by severalor many installa- processing...fast response, a high degree of tions...The data base is now distributedamong reliability,and aneasilyexpansiblesystem." the installations makingup the information network (Moravec, 1965, p. 173). for this service system... 2.2"In a highly distributed system, however, "The distributed information network should offer information on inputs to the organization flow considerable advantage in reducing thecost of directly to relatively low-levelway stations where terminal communications by permitting installations all possible processing is done and all actionsare to be located near concenrations of terminals." taken that are allowed by the protocol governing (Dennis, 1968, p. 373). that level. In addition to the direct actions that 2.3"A large number of factors (user communi- it takes, the lowest, or reflexive, level of information ties,, document forms, subject disciplines, desired processing ordinarily generates two classesof services, to name but a few) compete for theatten- information. These are, first, summaries of actions tion of the designer of information servicesystems. taken or anticipated and, second, summaries of A methodology for the careful organization ofthese information inputs that, because of their type,factors and the orderlyconsideration of their salience, or criticality, fall outside therange of relationships is essential if intelligent decisionsare action that policy has established as appropriate for to be made." (Sparks et al., 1965, pp. 1-2). that level.. . "The lack of recognition of thenature and even, "In computer terms, a highly distributed system in some cases, the existence of the problems facing involves a primary executive program that adds and the information systems designer hasmeant that subtracts subroutines to various primary libraries there has been little or no orderly development of from which alternative subroutines are to be drawn generallyagreeduponsystemmethodology." and combined. Secondary executiveprograms, (Hughes Dynamics, 1964, p. 1-7). responding to separate inputs and conditions, select "To the best of our knowledge,no one has yet and organize subroutines from each of these primary developed a completely satisfactory theory of inlibraries and add and subtract subroutines to various formation processing. Because there isno strong secondary libraries from which tertiary executive theoretical basis for the field,we must rely on intui- programs select alternative subroutines for use at tion, experience and the application of heuristic 36 notions each time we attempt to solve a new informa- network operations, available trained man- tion processing problem." (Clapp, 1967, p. 4). power, and ability to respond to change?" 2.4Additional examples are as follows: (Sayer, 1965, pp..144-145). "Preliminary data support the previous indica- "Some of the details the user must determine are tions (Werner, Trueswell, et al.) that the introduc- the number and location of remote points, frequency tion of new services is not followed by an im- of use, response time required, volume of data to be mediately high level use of them. The state-of-the- communicated, on line storage requirements, and art of equipment, personnel, and documentation the like." (Jones, 1965, p. 66). still offers continuing problems. Medical researchers 2.7"Neglect of 'WHERE the system is to be in the study do not seem to look upon the system as used' is the most frequent cause of inadequate sys- being an essential source of information for their tem designs." (Davis, 1964, p. 21). work, but as a convenient ancillary activity." 2.8Thus Sayer points out the need for "popula- (Rath and Werner, 1967, p. 62). tion figures describing the user community in detail, "A major study recently conducted by Auerbach its interest in subject disciplines, and the effect of Corporation into the manpower trendsinthe this interest on the effective demand on the system engineeringsupportfunctionsconcernedwith from both initiative and responsive demands." information ... which involvedinvestigations (Sayer, 1965, p. 140). of a large number of company and government Sparks et al. raise the following considerations: operations, was both surprising and disconcerting "There are certain basic dimensions of an informa- because it showed that there are large areas of tion service system which it is appropriate to recog- both government and industry in which there is nize in a formal way. One of these is the spectrum of very little concern about, or work underway toward, selected disciplines which are to be represented in solvingtheinformationflowandutilization the information processed by the system. Another problem." (Sayer, 1965, p. 24). of these is the geographical area to be served by the 2.5"There are seven properties of a system that system and in which the user population will be can be stated explicitly by the organization request- distributed... ing the system design: WHAT the system should "The number of user communities into which the be, WHERE the system is to be used, and WHERE, user population is divided determines (or is deter- WHEN, WITH WHAT, FOR WHOM, and WITH mined by) the number of direct-service information WHOM the system is to be designed." (Davis, centers in the system. Thus, it has a major effect 1964, p. 20). on system size and structure." (Sparks et al., 1965, 2.6Consider also the following: pp. 2-6, 2-7). "Consequently, it appears that two earlyareas 2.9"In structuring shiny, new information sys- of required investigation are those of determining: tems, we must be careful to allow for resistance to 1) who are the potential users of science and/or change long before the push buttons are installed, engineering information systems, where are they especially when the users of the systems have not located, what is their sphere of activity? and 2) been convinced that there is a real need for change." What is the real nature and volume of material (Aines, 1965, p. 5.) that will flow through a national information sys- "Examine the various systems characteristics tem?... such as: user/network interface; network usage "In undertaking a program to establish informa- patterns;training requirements;trafficcontrol; tion service networks itis necessary to know: service and organization requirements; response effectiveness;cost determinations; and network 1. Who are the users? capacity." (Hoffman, 1965, p. 90-91.) 2. What are the user information needs? "As an appendage to a prototype network, some 3. Where are these users? experimental retraining programs would be well 4. How many users anduser groups are there, advised... and how, do their needs differ? "A massive effort directed at retraining large 5. What information products and services will numbers of personnel now functioning in libraries meet these needs? will be required to produce the manpower necessary 6. What production operationsare necessary to for a real-time network ever to reach a fully opera- producetheseinformationproductsand tional status." (Brown et al., 1967, p. 68). services? "Where do experimental studies of user perform- 7. Which of these products and servicesare ance fit into burgeoning information services? The really being produced now; by whom and where answer is inescapable: the extent of experimental and how well is an ultimate purpose already activity will effectively determine the level of ex- being achieved? cellence, in method and in substantive findings, 8. How will any new system best integrate with with which key problems regarding user perform- existing practices? ance will be met. If experimental studies in man- 9. What are the operations best performed from computer communication continue to be virtually a standpoint of quality and timeliness of nonexistent, the gap in verified knowledge of user service to users, economy of costs and overall behavior will continue to be dominated by. immediate 37 cost and narrow technicalconsiderations rather mental conditions, were only indirectly concerned than by the users' long range interests. Everyonewill with man-computer communications, dealing largely be a loser. Neither the managers of computer utili- with knobs, buttons and dials rather than with the ties, or the manufacturers, or the designers of cen- interactiveproblem-solving of theuser.Inall tral systems will have tested, reliable knowledgeof fairness, there were some exceptions to this rule, what the user needs, how he behaves, how long it but they were too few and too sporadic to make a takes him to master new services, or how well he significant and lasting impact on the mainstream performs. In turn, the user will not have reliable, of user development. Since there was, in effect, validated guidance to plan, select, and become an applied scientific vacuumsurrounding man- skilled in harnessing the information services best computer communication, it is not atall surprising suited to his needs, his time, and his resources. that there does not exist today a significant, cumula- Since he is last, the user loses most." (Sackman, tive experimental tradition for testing andevaluating 1968, p. 351). computer-userperformance."(Sackman,1968, 2.10"Everyone talks about the computer user, p. 349). but virtually no one has studied him in a systematic, "The problem is, of course, to get the right scientific manner. There is a growingexperimental information to the right man at the right time and lag between verified knowledge about usersand at his work station and withminimum effort on rapidly expanding applicationsfor them. This his part. What all this may well be saying is that the experimental lag has always existed in computerinformation problem that existsis considerably technology. Technological innovation and aggres- more subtle and complexthan has been set forth ... sive marketing of computer wareshave consist- The study for development of a Methodology for ently outpaced established knowledgeof userAnalysis of Information Systems Networks arrives, performanceabiasincomputertechnology both directly and by implication at the same con- largely attributable to current management outlook clusion as have a number of other recent studies. and practice. With the advent oftime-sharing That conclusion is that much more has to be known systems, and with the imminenceof the much- about the user and his functions, and much more heralded information utility, the magnitude of this has to be known about what the process of RDT & E scientific lag may have reached a criticalpoint. actually is and how can information, as raw material If unchecked, particularly in the crucial areaof input to the process, flow most efficiently and most software management, it may become acrippling effectively." (Sayer, 1965, p. 146). humanistic lag a situation in which boththe "The recurrent theme in general review articles private and the public use of computerswould be concerned with man-computer communication is characterized by overriding concern forimmediate the glaring experimental lag. Innovation and un- machine efficiency and economy, and by an en- verified applications outrace experimentalevalua- trenchedneglectof human needs,individual tion on all sides. capabilities, and long-range social responsibilities." "In a review of man-computer communication, (Sackman, 1968, p. 349). Ruth Davis points out that virtually no experimental "Quite often the most important parameter inwork has been done on user effectiveness.She a system'sperformance is the behavior of the characterizes the status of user statistics as inade- average user. Thisinformation is very rarely known quate and 'primitive', and she urgesthe specifica- in advance, and can only be obtainedby gathering tion and development of extensive measures of statistics.Itis important to know, for example, user performance. ... how long a typical user stays on atime-sharing "Pollack and Gildner reviewed the literature on system during an averagesession, how many lan- user performance withmanual input devices for guage processors he uses,how much computing man-computercommunication.Their extensive power he requiresduring each 'interaction' with survey covering largenumbers and varieties of the system, and so forth. Modeling andsimulation switches, pushbuttons, keyboards and encoders can be of greathelp in pre-determining this infor- revealed 'inadequate research data establishing mation if the environment is known, butin many performance for various devices and device char- commercial or University time-sharing systems acteristics, and incomplete specification of operator there is little control over or priorknowledge of input tasks in existing systems.' There was a vast the characteristics of the users."(Yourdon, 1969, experimental gap between the literally hundreds p. 124). of manual input devices surveyed and the very "The lag in user studies is a heritage which stems small subset of such devices certified by some form mainly from the professional mix thatoriginally of user validation. They recommended an initial developed and used the technology of man-computer program of research onleading types of task/device communications.Fortwocritical,formative couplings, and on newer and more natural modes of decades, the 1940's and the 1950'scomprising manual inputs such as speech and handwriting." the birth and development of electronicdigital (Sackman, 1968, p. 350). computerssocialscientists,human engineers 2.11"Information control at input can be used and human factors specialists, theprofessionals to achieve improved system efficiencyin several trained to work with human subjects underexperi- different ways. First, a reduction in the total 38 volume of infOrmation units or reports to be received, raphy, or by project staffin transcription processed, or storedcan be gained through theuse from source to of filtering proceduresto reduce the possible paper tape. In any case, error re- is a factor in reducing thepossibility of dundancies between itemsreceived. (Timingcon- siderations are important in such identity of duplicates. procedures, as "2. Variationsamong bibliographiesbOthin noted elsewhere, becausewe won% want a delayed style and content. A bibliographical and incorrectmessage to 'update' its owncorrec- citation tion notice.) gives several different kinds ofinformation; that is, it conteins several'elements,' such "Secondly, input filtering proceduresserve to as author of item, title, publication data, reduce the total bulk of informationto be processed re- or stored both by elimination of duplicate views and annotations. Eachsource bibliog- items raphy more or less consistentlyemploys one as such and by the compression of thequantitative style for expressing amount of recording used torepresent the original information, but each information unit style differs fromevery other in some or all or riaessage within the system. of the followingways: "A third technique ofinformation control at input is directed to the control ofredundancy within a. number of elements a single unit or report. Conversely, inputfiltering b. sequence of elements procedures of this typecan be used to enhance the c. typographical details" (1965,p. 96). value of information to be stored. For example, in 2.14 pictorial data processing,automatic boundarycon- "File integritycan often be a significant trast enhancements motivation for mechanization. Toinsure file integ- or `skeletonizations' may im- rity in airline maintenance prove both subsequent human patternperception records, files have been and system storage efficiency.Another example is republished monthly in cartridgeroll-microfilm natural text processing, wheresystematic elimina- form, since mechanics wouldnot properly insert tion of the 'little', 'common',and 'non-informing' update sheets in maintenancemanuals. Freemont words can significantly reducethe amount of text Rider's originalconcept for the microcard, which to be manipulated by the machine."(Davis, 1967, was a combination of a catalog card anddocument p. 49). in one record, failed inpart because of the lack of file integrity. Every librarian 2.12 knows that if there In this area, R & Drequirements for the wasn't a rod through the holein the catalog card they future include thevery severe problems of sifting would not be able to maintain and filtering the integrity of the enormous masses of remotely collected card catalog." (Tauber, 1966,p. 277). data. For example, "our abilityto acquire data is so 2.15 far ahead ofour ability to interpret and manage it "Retirement of outmoded datais the only that there is long-rangeeffectivemeans of maintaining an some question as to just how farwe efficient system." (Milleret al., 1960, p. 54). can go toward realizing the promise of much of this With respect to maintenance remote sensing. Probably 90% of the data processes involving gathered the deletion of obsolete items,there are substantial to date have not been utilized, and, withlarge multi- fact finding research sensor programs in the offing, we face the danger of requirements for large-scale ending up prostrate beneath documentary item systems interms of establishing a mountain of utterly efficient but realistic criteria for useless films, tapes, and charts."(Parker and Wolff, "purging". Kessler 1965, p. 31). comments on this point as follows: "It isnot just a matter of throwing away 'bad'papers as 'good' 2.13 "Purging because ofredundancy is ex- ones come along. The scientific literatureis unique tremely difficult to accomplish bycomputer program in that its best examplesmay have a rather short except in the case of 100% duplication. Redundancy life of utility. A worker inthe field of photoelec- purging success is keyedto practices of standardiza- tricity need not ordinarily be referredto Einstein's tion, normalization, field formatting,abbreviation original paperon the subject. The purging of the conventions and the like. Asa case in point, docu- system must be basedon criteria of operational ment handling systems universally haveproblems relevanceratherthanintrinsicvalue.These with respect to bibliographiccitation conventions, criteria are largely unknownto us and represent transliterations ofproper names, periodical title another basicarea in need of research and inven- abbreviations, corporate author listingpractices and tion." (1960;pp. 9-10). the like." (Davis, 1967,p. 20). See also Ebersole (1965),Penner (1965), and "Chronological cutoff is thatdevice attempted most frequently in automated informationsystems. Sawin (1965) who pointsto some of the difficulties It is employed successfully with respect toa bibliographic collection or file,as in real-time systems follows: such as aircraftor satellite tracking or airline reservations systems where theinformation is "1. Actualerrors, such as incorrect spelling of useless aftervery short time intervals and where words, incorrect report ofpagination, in one it is so voluminousas to be prohibitive for future or more of the duplicates. Theerror may be analyses... mechanically or humanly generated; the error "That purging which is doneis primarily replace- may have been made in thesource bibliog. ment.Data managementorfilemanagement 39 systems are generally programmed so that upon sion, Wright-Patterson AFB had itsfile system proper identification of an item during the manual organized on this basis by IBM in the early 1960's. input process it may replace an item already in It was found from surveys of manual translators the system data bank. The purpose of replacement that the majority of vocabulary references were as a purging device is not volume control. It is to less than one thousand words. These were isolated for purposes of accuracy, reliability or timeliness and located in the fastest-access memory: the rest controls." (Davis, 1967, p. 15). of the dictionary was then relegated to lower priority "The reluctance to purge has been a leading locations..." (Davis, 1967, pp. 18-19). reason for accentuating file storage hierarchy con- "The network might show publications being siderations. Multi-level deactivation of information permanently retained at a particular location. This is substituted for purging. Deactivation proceeds would allow others in the network to dispose of through allocating the material so specified first little-used materials and still have access to a to slower random-access storage devices and then copy if the unexpected need arose ... tosequentially-accessedstoragedeviceswith 'Such an 'archival' copy could, ofcourse, be decreasing rates of access all on-line with the relocated to a relatively low-cost warehousearea computer. As the last step of deactivation the for the mutual benefit of those agencies in the net- information is stored in off-line stores... work. Statistics on frequency of usage might be "Automatic purging algorithms have been written very helpful in identifying inactive materials, and for at least one military information system and for the network could also fill this need." (Brown et al., SDC's time-sharing system...In the military 1967, p. 66). system ...the purging program written allowed "Periodic reports to users on file activitymay all dated units of information to be scanned and reveal possible misuse or tampering." (Petersen those prior to a prescribed date to be deleted and and Turn, 1967, p. 293). transcribed onto a magnetic tape for printing. The 2.17"Accessibility.For a system output a information thus nominated for purging was re- measure of how readily the proper information viewed manually. If the programmed purge decision was made available to the requesting user on the was overridden by a manual decision the falsely desired medium." (Davis, 1964, p. 469). purged data then had to be re-entered into system 2.18Consider also the following: files as would any newly received data." (Davis, "The system study will consider that the docu- 1967, pp. 16-18). ment-retrievalproblemliesprimarilywithin "Automatic purging algorithms have been ex- the parameters of file integrity; activity and activity plored for the past three years. The current scheme distribution; man-file interaction; the size, nature attempts to dynamically maintain a 10 percent and organization of the file; its location and work- disc vacancy factor by automatically deleting the place layout; whether it is centralizedor decentral- oldest files first. User options are provided which ized; access cycle time; and cost. Contributing permit automatic dumping of files on a backup, factors are purging and update; archival considera- inactive file tape...prior to deletion." (Schwartz tions;indexing;typeofresponse;peak-hour, and Weissman, 1967, p. 267). peak-minuteactivity;permissable-errorrates; "The newer time-sharing systems contemplate and publishing urgency." (Tauber, 1966,p. 274). a hierarchyoffilestorage,with 'percolation' Then there are questions of sequential decision- algorithmsreplacingpurgingalgorithms.Files making and of time considerations generally. "Time will be in constant motion, some moving 'down' considerationisexplicitly,althoughinformally, intohigher-volume,slower-speedbulkstore, introduced by van Wijngaarden as 'the value of a while others move 'up' into lower-volume, higher- text so far read'. Apart from other merits of van speed memory all as a function of age and refer- Wijngaarden's approach, and his stressing the inter- ence frequency." (Schwartz and Weissman, 1967, action between syntax and semantics, we would like p. 267). to draw attention to the concept of 'value at time t', 2.16"Some computer-orientedstatisticsare which seems to be a really basic concept in pro- provided to assist in monitoring the system with gramming theory." (Caracciolo di Forino, 1965, p. minimum cost or time. Such statistics are tape 226). We note further that "T as the time a fact length and length of record, checks on dictionary assertion is reported must be distinguished from the code number assignment, frequency of additions time of the fact history referred to by the assertion." or deletions to the dictionary, and checks to see (Travis, 1963, p. 334). that the correct inverted file was updated." (Smith Avram et al., point more prosaically to practical and Jones, 1966, p. 190). problems in mechanized bibliographic reference "Usage statistics as obsolescence criteria are data handling, as in the case of different types of commonly employed inscientific and technical searches on date: The ease of requesting all works information systems and reference data systems... on, say, genetics, written since 1960 as against that "Usage statistics are also used in the deactivation of all works on genetics published since 1960 with process to organize file data in terms of its refer- respect to post-1960 reprints of pre-1960 original ence frequency. The Russian-to-English automated texts. translation system at the Foreign Technology Divi- For the future, moreover, "In some instances, the 40 search request would haveto take into account,sistencies we eliminate, for example, the large which data has been used in thefixed field. For errors that would result when something has been example, shouldone want a display of all the books improperly reported in pounds instead of in thou- in Hebrew published during,a specific time frame,sands of pounds. Perhaps one-thirdto one -.half of an adjustment would have to be made to the date the time our UNIVACs devoteto processing these in the search request tocompensate for the adjust. Censuses will be spent checking for such ment made to the data at input time," (Avram incon- et al., sistencies and eliminating them.... 1965, p. 42). "Similar checking procedures 2.19 "Here you run into the phenomenon are applied to the of the approximately 7,000 product lines for whichwe `elastic ruler', At the time when certaindata were have reports. Ina like manner we check to see accumulated, the measurementswere made with whether such relationshipsas annual man hours a standard inch or standard meter...whether and number of production workers,or value of researchers were usingan inch standardized before shipments and cost of labor and materials, a certain date, or one adopted later." (Birch, 1966, are p. 165). within reasonable limts for the industryand area involved.... 2.20 "Large libraries face the problem ofcon- "For example, thecomputer might determine verting records that exist inmany languages. The for an establishment classified most complete discussion of this problem as a jewelry repair to date is shopohat employees' salaries amountedto less than by Cain & Jolliffe of the British Museum.They sug- 10 percent of total receipts. For thiskind of service gest methods for encoding different languages and trade, expenditures for labor usuallyrepresent the speculate on the extent to which certaintranslitera- major item of tions could be done by machine. The expenses and less than 10 percent possibility of for salariesis uncommonly low. Ourcomputer storing certain exotic languageson videotapes is would list thiscase for inspection, and a review suggested as a way of handling the printingproblem. At the Brasenose Conference of the report might result ina change in classifi- at which this paper cation from 'jewelry repair shopto retail jewelry was presented, the authors analyzed the difficulties store', for example." (Hansen and McPherson, in bibliographic searching caused by transliteration 1956, pp. 59-60). of languages (this is the schememost generally suggested by those in the data processingfield)." 2.23"The use of logicalsystems for error (Markuson, 1967, p. 268). control is in beginning primitivestages. Question- 2.21"The question of integrity of information answering systems and inference-derivation within anautomated,system is infrequently ad- programs may find their most valueas error control dressed." (Davis, 1967,p. 13). procedures rather thanas query programs or "No adequate reference service existsthat would problem-solving programs." (Davis, 1967,p. 47). allow users to determine easily whetheror not rec- "A theoretically significant result'of introducing ords have the characteristics of qualityand com-' source indicators and reliability indicatorsto be patibility that are appropriateto their analytical carried along with fact assertionsin an SFQA requirements." (Dunn, 1967,p. 22). [question-answering] system is that theyprovide 2.22"Controlsthrough 'common sense'or a basis for applying purifyingprograms to the fact logical checks...include the use of allowable assertions stored in thesystem i.e., for resolving numerical bounds suchas checking bearings by contradictionsamongdifferentassertions,for assuming them to be bounded by 0°as a minimum culling out unreliable assertions,etc.... and 360° as a maximum. They include consistency "Reliabilityinformationmight indicatesuch checks using redundant information fieldssuch as things as: S's degree of confidencein his own social security number matched againstaircraft report if S is a person; S's probableerror if S is type and aircraft speed. They also includecurrent ameasuring instrument;S's awareness checks such as matches of diplomat by dependabilityas determined by whetherlaterexperiencecon- name against reported location by city against firmed S's earlierreports; conditions under which known itinerary against known politicalviews." S made its report, etc." (Travis, (Davis, 1967, p. 36). 1963, p. 333). 2.24"Another interesting distinctioncan be "A quite different kind of workis involved in made between fileson the basis of, their accuracy. examing for internal consistency thereports from A clean file isa collection of entries, each of which the more than 3 million establishmentscovered was precisely correct at the time of its inclusion in the 1954 Censuses of Manufacturersand Busi- in the file. On the other hand,a dirty file is a file ness. If these reports were all complete and self- that contains a significant portionof errors. Are- consigtent and if wewere smart enough to foresee circulating file is purged and cleansedas it cycles a all the problems involved in classifyingthem, and if we made utility-company billing file is of thisnature. After no errors in our office work. the job of the File'settles down,' the proportion oferrors getting out the Census reports would be laborious imbedded in the file isa function of the new activity but straightforward. Unfortunately,some of the applied to the file. The reports do contain omissions, error rate is normalized errors, and evidence with respect to the business cycle."(Patrick and of misunderstanding. By checking forsuch incon- Black, 1964, p. 39). 41 "When messages are a major source of the error detection only, error detection and correction information entering the system corrections 'to a with a non-constant speed of end-to-end data previously transmitted original message can be transfer (during the correction cycle transmission received before the original message itself.If stops), and error detection and correction with a entered on an earlier update cycle the correction constant speed of end-to-end data transfer (during data can actually be 'corrected' during a laterthecorrectioncycletransmissioncontinues)." update cycle by the original incorrect message." (Menkhaus, 1967, p. 35). (Davis, 1967, p. 24). 2.25"Errors will occur in every data collection "There are two other potential 'error injectors' system, so itis important to detect and correct which should be given close attention, since more as many of the errors as possible." (Hillegass and control can be exercised over these areas. They Me lick, 1967, p. 56). are: the data collection, conversion and input de- "The primary purpose of a data communications vices, and the human being, or beings, who collect system is to transmit useful information from one the data (or program a machine to do it) at the location to another. To be useful, the received copy source. Bell estimates that the human will commit of the transmitted data must constitute an accurate an average of 1,000 errors per million characters representation of the original input data, within handled, the mechanical device will commit 100per the accuracy limits dictated by the application million, and the electronic component, 10per requirements and the necessary economic tradeoffs. million. .. Errors will occur in every data communications "Error detection and correction capability is a system. This basic truth must be kept in mind `must' in the Met Life system and this is provided throughout the design of every system. Important in several ways. The input documents have Honey- criteria for evaluating the performance of any com- well's Orthocode format, which uses five rows of munications system are its degree of freedom from bar codes and several columns of correction codes data errors, its probability of detecting the errors that make defacement or incorrect reading virtually that do occur, and its efficiency in overcoming the impossible; the control codes also help regenerate effects of these errors." (Reagan, 1966, p.26). partially obliterated data.... "The form of the control established, as a result of the investigation, should be decided only after "Transmission errors are detected by using a dual considering each situation in the light of the three pulse code that, in effect, transmits the sic,ils for a control concepts mentioned earlier. Procedures, message and also the components of those signals, such as key verification, batch totals, sight verifica- providing a double check on accuracy. The paper tion, or printed listings should be used only when tape reader, used to transmit data, is bi-directional; they meet the criteria of reasonableness, in light if a message contains a large number of errors, due of the degree of control required and the cost of possibly to transmission, noise, the equipment in the providing control in relation to the importance and head office detects those errors and automatically volume of data involved. The objective is to estab- tells the transmitting machine to 'back up and start lish appropriate control procedures. The manner over'." (Menkhaus, 1967, p. 35). in which thisis donei.e., the particular com- 2.28"Inputinterlockschecks which verify bination of control techniques used in a given set that the correct types and amounts of data have been circumstanceswill be up to the ingenuity of the inserted, in the correct sequence, for each trans- individual systems.designer." (Baker and Kane, action. Such checks can detect many procedural 1966, pp. 99-100). errors committed by persons entering input data 2.26"Two basic types of codes are found into the system." (Hillegass and Melick, 1967, p. 56). suitable for the burst type errors. The first is the 2.29"Parityaddition of either a 'zero' or 'one' forward-actingHagelbarger code which allows bit to each character code so that the total number fairly simple data encoding and decoding with of 'one' bits in every transmitted character code provisions for various degrees of error size cor- will be either odd or even. Character parity checking rection and error size detection. These codes, can detect most single-bit transmission errors, but however, involve up to 50 percent redundancy it will not detect the loss of two bits or of an entire in the transmitted information. The second code character." (Hillegass and Melick, 1967, p. 56). type is the cyclic code of the Bose-Chauduri type "Two of the most popular error detection andcor- which again is fairly simple to encode and can detect rection devices on the marketTally's System 311 various error burst sizes with relatively low redun- and Digitronics' D500 Seriesuse retransmission dancy. This code type is relatively simple to decode as a correction device. Both transmit blocks of for error detection but is too expensive to decode characters and make appropriate checks for valid for error correction, and makes retransmission the parity. If the parity generated at the transmitter only alternative." (Hickey, 1966, p.182). checks with that which has been created from the 2.27"Research devoted tofinding ways to received message by the receiver, the transmission further reduce the possibility of errors is progress- continues. If the parity check fails, the last block is ing on many fronts. Bell Telephone Laboratories retransmitted and checked again for parity. This is approaching the problem from three angles: method avoids the disadvantages of transmitting 42 the entire message twice and of having tocompare check digit. If one of his characters isnot clear, the second message with the first for validity." the machine looks at the check digit, carriesout (Davenport, 1966, p. 31). its arithmetic on the legible characters, and sub- "Full error detection and correction is provided. tracts the result from the result that would give The telephone linecan be severed and reattached the check digit to establish the character in doubt. hours later without loss of data ..Error detection It then rebuilds this character." (Rothery, 1967, is accomplished by a horizontal and vertical parity p. 59.) bit scheme similar to that employedon magnetic 2.31"A hash total works in the followingway. tape." (Lynch, 1966, p. 119). Most of our larger computerscan consider alphabetic "A technique that has proven highly successful information as data. These dataare added up, is to group the eight-level characters into blocks just as if they were numeric information, and of eighty-four characters. One of the eighty-four meaningless total produced. Since the high-speed characters represents a parity character, assuring electronics are very reliable, they should produce that the summation of each of the 84 bits at each the same meaningless numberevery time the same of eight levels is either always oddor always even. data fields are summed. The transfer of information For the block, there is now a vertical parity check within the computer and to and from the various (the character parity) and a horizontal parity check input/output units can be checked by recomputing (theblockparitycharacter).This dualparity this sum after every transmission and checking check will be invalidated only whenan even against the previous total... number of characters within the block havean "Some computers have special instructions built even number of hits, each at the same level. The into them to facilitate this check, whereas others probability of such anoccurrence is so minute accomplishitthroughprogramming. Thefile that we can state that the probability ofan unde- designer considers the hash totalas a form of built- tected error is negligible. In an 84-character block, in audit. Whenever the file is updated, the hash constituting 672 bits, 83 + 8 = 91 bitsare redundant. totals are also updated. Whenevera tape is read, Thus, at the expense of adding redundancy of 13.5 the totals are reconstitutedas an error check. per cent, we have assured error-free transmission. Whenever an error is found, the operation isre- At least we know thatwe can detect errors with peated to determine if a randomerror has occurred. certainty. Now, let us see how we can utilize this If the information is erroneous,an alarm is sounded knowledgetoprovideerror-freedataathigh and machine repair is scheduled. If information has transmission rates. One of the most straightforward been actually lost, then human assistance is usually techniques isto transmit data in blocks, auto- required to reconstitute the file to itscorrect con- matically checking each block for horizontal and tent. Through a combination of hardware andpro- vertical parity at the receiving terminal. If the gramming the validity of large reference filescan block parities check, the receiving terminal de- be maintained even though the file is subjectto livers the block and an acknowledgment character repeated usage." (Patrick and Black, 1964,pp. (ACK) is automatically transmitted backto the 46-47). sending terminal. This releases the next block and 2.32"Message lengthchecks which involve the procedure is repeated. If the block parities do a comparison of the number of characters as spec- not check, the receiving terminal discards the ified for that particular type of transaction. Message block Pnd a nonacknowledgment character (NACK) length checks can detectmany errors arising from is returned to the sender. Then, thesame block is both improper data entry and equipmentor line retransmitted. This procedure requires thatstorage malfunctions." (Hillegass and Milick, 1967,p. 56). capacity for a minimum ofone data block be 2.33"In general, many standard techniques provided at both sending and receiving terminals." such as check digits, hash totals, and format checks (Rider, 1967, p. 134). can be used to verify correct input and trans- 2.30"What then can we say that will summarize mission.These checksareperformedatthe the position of the check digit? Wecan say that it computer site. The nature and extent of the checks is useful for control fields that is, those fields will depend on the capabilities of thecomputer we access by and sort on, customer number, em- associated with the response unit. One effective ployee number, etc. We cango further and say technique is to have the unit respond witha verbal thatitreally matters only with certain control repetition of the input data." (Melick, 1966,p. 60). fields, notall. With control fields, the keys by 2.34"Philco has a contract for building what is which we find andaccess records, it is essential called a Spelling-Corrector...It reads text and that they be correct ifwe are to find the correct matches it against the dictionary to findout whether record. If they are incorrect through juxtaposition the words are spelled correctly." (Gibbs and Mac- or other errors in transcription, we will 1) not find Phail, 1964, p. 102). the record, and 2) find andprocess the wrong "Following keypunching, the information retrieval record.... technician processes the data usinga 1401 com- "One of the most novel uses of the check digit puter. The computer performs sequence checking, can be seen in the IBM 1287 optical scanner. The editing, autoproofing (each word of input is checked writer enters his control field followed by the against a master list of correctly spelled words 43

376-411 0 - 70 - 4 to determine accuracya mismatch is printed out in Morse Code transmission." (Vossler and Brans- for human analysis since it is either a misspelled ton, 1964, p. D2.4-1). or a new word), and checking for illegitimate char- "Harmon, in addition to using digram frequencies acters. The data is now on tape; any necessary to detect errors, made use of a confusion matrix to correctionchanges or updating can be made determine the probabilities of various letter sub- directly." (Magnin°, 1965, R. 204), stitutions as an aid to correcting these errors." "Prior to constructing the name file, a 'legitimate (Vossler and Branston, 1964, pp. D2.4-1 D2.4-2). name' list and a 'common error' name list are tabu- "An interesting program written by McElwain lated ..The latter list is formed by taking char- and Evens was able to correct about 70% of the acter error information compiled by the instrumenta- garbles in a message transmitted by Morse Code, tion system and thresholding it so only errors with when the received message contained garbling in significant probabilities remain;i.e.,`e' for 'a'. 0-10% of the characters." (Vossler and Branston, These are then substituted one character at a time 1964, p. D2.4-1). in the names of the 'legitimate name' list to create a "The design of the spoken speech output mo- `common error' name list. Knowing the probability dality for the reading machine of the Cognitive of error and the frequency of occurrence of the Information Processing Group already calls for OW `legitimate name' permits the frequency of occur- a large, disc-stored dictionary ...The possibility rence for the 'common error' name to be calculated." of a dual use of this dictionary for both correct (Hennis, 1967, pp. 12-13). spelling and correct pronunciation prompted this 2.35"When a character recognition device errs study." (Cornew, 1968, p. 79). in the course of reading meaningful English words it "Our technique was first evaluated by a test will usually result in a letter sequence that is itself performed on the 1000 most frequent words of not a valid word; i.e., a 'misspelling'," (Cornew, English which, by usage, comprise 78% of the 1968, p. 79). written language...For this, a computer pro- 2.36"Several possibilities exist for using the gram was written which first introduced into each information the additional constraints provide. A of these words one randomly-selected, randomly- particularly obvious one is to use special purpose placed letter substitution error, then applied this dictionaries, one for physics texts, one for chem- technique to correct it. This resulted in the fol- istry, one for novels, etc., with appropriate word lowing overall statistics 739 correct recoveries of lists and probabilities in each...." the original word prior to any other; 241 incorrect "Because of the tremendous amount of storage recoveries in which another word appeared sooner; which would be required by such a 'word digram' 20 cases where the misspelling created another method, an alternative might be to associate with valid word." (Cornew, 1968, p. 83). each word its one or more parts of speech, and make "In operation, the word consisting of all first use of conditional probabilities for the transiC.,n choice characters is looked up. If found, it is as- from one part of speech to another." (Vossler and sumed correct; if not, the second choice characters Branston, 1964, p. D2.4-7). are substituted one at a time until a matching word 2.37"In determining whether or not to adopt an is found in the dictionary or until all second choice EDC system, the costliness and consequences of substitutions have been tried. In the latter case any error must be weighed against the cost of in- a multiple error has occurred (or the word read stalling the error detection system. For example, in correctly is not in the dictionary)." (Andrews, 1962, a simple telegram or teleprinter message, in which p. 302). all the information appears in word form, an error 2.39"There are a number of different tech- in one or two letters usually does not prevent aniques for handling spelling problems having to reader from understanding the message. With train- do with names in general and names that are homo- ing, the human mind can become an effective error nyms. Present solutions to the handling of name detection and correction system;it can readily files are far from perfect." (Rothman, 1966, p. 13). identify the letter in error and make corrections. Of 2.40"The chief problem associated with... course, the more unrelated the content of the mes- large name files rests with the misspellingor mis- sage, the more difficult it is to detect a random misunderstanding of names at time of input and with take. In a list of unrelated numbers, for example, it possible variations in spelling at the time of search. is almost impossible to tell if one is incorrect."In order to overcome such difficulties, various (Gentle, 1965, p. 70). coding systems have been devised to permit filing 2.38In addition to examples cited in a previous and searching of large groups of names phonetically report in this series, we note the following: as well as alphabetically ..A Remington Rand "In the scheme used by McElwain and Evens, Univac computer program capable of performing urAisturbed digramsor trigrams in the garbledthe phonetic coding of input names has been message were used to locate a list of candidate prepared." (Becker and Hayes, 1963, p. 143). words each containing the digram or trigram. These "A particulartechniqueusedinthe MGH were then matched against the garbled sequence [Massachusetts General Hospital] system is probably taking into account various possible errors, such as worth mentioning; this is the technique for phonetic a missing or extra dash, which might have occurred indexing reported by Bolt et al. The use described 44 involves recognition of drugnames that have been It is thereforenecessary to develop the data and typed in, moreor less phonetically, by doctorsor nurses; in the longer view this standardize before furtherprocessingcommences. one aspect of a large "It is then possibleto process the data against the effort that must be expendedto free the man- existing file information machine interface from the . .The objective of the need for letter-perfect processing is to categorize theinformation with information representation bythe man. People respect to all other information within the just don't work thatway, and systems must be files... developed that To categorize the information,a substantial amount can tolerate normal human impre- of retrieval and association ofdata is often required cision without disaster." (Mills,1967, p. 243). . Many [data] contradictionsare resolvable by 2.41"...The object of the studyis to determine if we the system." (Gurk and Minker,1961, pp. 263-264). can replace garbled characters in 2.44 "A number ofnew developments are based names. The basic plan was to develop the empirical on the need for serving clustered environments. A frequency ofoccurrence of sets of characteres in names and cluster is defined asa geographic area of about three use these statistics to replacea miles in diameter. The basicconcept is that within missing character." (Carlson, 1966,p. 189). a cluster of stations and computers, it is possible "The specific effecton error reduction is impressive. If a to provide communication capabilitiesat low cost. scanner gives a 5% charactererror Further, it is possibleto provide communication rate, the trigram replacement techniquecan correct paths between clusters, approximately 95% of these as well as inputs to and errors. The remaining outputs from other arrangementsas optional fea- error is thus...0.25% overall.... tures, and still maintain economies withineach "A technique like thismay, indeed, reduce the cluster. This leads cost of verifying the mass of data input to a very adaptable system. It is coming from expected to find wide applicationon university scanners ...[and] reduce thecost of verifying massive data conversion campuses, in hospitals, within industrial complexes, coming fromconven- etc." (Simms, 1968,p. 23). tional data input devices likekeyboards, remote terminals, etc." (Carlson, 1966,p. 191.) 2.45"Among the key findingsare the following: 2.42"The rules established forcoding struc- Relativecost-effectivenessbetweentime- tures are integrated in theprogram so that the sharing and batch processingis very sensitive computer is able to take a fairly sophisticatedlook to and varies widely with the preciseman- at the chemist's coding and the keypunchoperator's machine conditions under whichexperimental work. It will not allowany atom to have too many or comparisons are made. too few bonds, nor is a '7' bond codepermissible Time-sharing showsa tendency toward fewer with atoms for which ionicbonds are not 'legal'. man-hours andmore computer time for experi- Improper atom and bondcodes and misplaced characters are recognized by mental tasks than batch processing. the computer,as The controversy is showingsigns of narrowing are various other types of errors." (Waldoand down to a competition between De Backer, 1959,p. 720). conversation- 2.43 ally interactive time-sharingversus fast-turn- "Extensive automatic verificationof the around batch systems. file datawas achieved by a variety of techniques. As an example,extracts were made of principal Individual differences inuser performance are lines plus thesequence number of the record: generally much larger andare probably more specifically,all corporatename lines were ex- economicallyimportantthantime-sharing/ tracted and sorted;any variations on a given name batch-processing system differences. were altered to conform to the standard. Similarly, Users consistently andincreasingly prefer all law firm citationswere checked against each interactive time-sharingor fast turnaround other. All city-and-state fieldsare uniform. A zip- batch over conventional batchsystems. code-and-place-name abstractwas made, with the Very little is known aboutindividual perform- resultant file being sorted by zipcode: errorswere ance differences, user learning, and human easy to sort and correct,as with Des Moines decision-making, the key elementsunderlying appearing inthe Philadelphialisting." (North, the general behavioral dynamicsof man-com- 1968, p. 110). puter communication. Then there is theeven more sophisticated case Virtually no normativedata are availableon where ". . An important inputcharacteristic data-processing problems andtasks, nor on is that the data is not entirelydeveloped forproc- empirical use ofcomputer languages andsys- essing or retrievalpurposes. It is thus necessary tem support facilities the kind of dataneces- first to standardize and developthe data before sary to permit representative samplingof manipulating it. Thus, to mentionone descriptor, problems, facilities andsubjects for crucial `location', the desired machineinput might be experiments thatwarrant generalizablere- `coordinate', 'city', and 'state',if a city ismen- sults." (Sackman, 1968,p. 350). tioned; and 'state' alone when no city is noted. However, on at leastsome occasions, some clients However, inputs to thesystem might contain a of a multiple-access, coordinate and city without time-shared systemmay be mention of a state. satisfied with,or actually prefer, operation ina 45 batch or job-shop mode to extensive use of the con- Second, the job to be done always liesembedded versational mode. withinsome formalorganizationalstructure." "Critics (see Patrick 1963, Emerson 1965, and (Bennett, 1964, p. 98). MacDonald 1965) claim that the efficiency of time- "Formal organizingprotocolexistsrelatively sharing systems is questionable when compared to independently of an organization's purposes, origins, modern closed-shop methods, or with economical or methods. Theseestablished operating procedures small computers." (Sackman et al., 1968, p. 4), of an organization impose constraints upon the Schatzoff et al. (1967) report on experimental available range of alternatives for individual be- comparisons of time-sharing operations(specifically, havior. In addition to such constraints upon the MIT's CTSS system) with batch processing as em- degrees of freedon within an organization as re- ployed on IBM's IBSYS system. strictions upon mode of dress, conduct, rangeof ..One must consider the total spectrum of mobility, andstyle of performance, there are tasks to which a system will be applied, andtheir protocolconstraintsupontheformat,mode, relative importance to the total computing load." pattern, and sequence ofinformation processing (Orchard-Hays, 1965, p. 239). and information flow. It is this orderlyconstraint 44. processingandinformation .A major factor to be considered in the uponinformation design of an operating system is the expected job flow that we call, for simplicity, the information mix." (Morris et al., 1967, p. 74). system of an organization.The term 'system' "In practice, a multiple system may containboth implies little more than procedural restrictionand types of operation: a group of processorsfed from orderliness. By 'information processing' we mean a single queue, and many queuesdifferentiated some actual change in the natureof data or docu- by the type of request being serviced by theattached ment,. By 'information flow' we indicate a similar processor group ..." (Scherr, 1965, p. 17). change in the location of these data or documents. 2.46"Normalization is a necessary preface to Thus we may define an information system as the merge or integration of our data. By merge, or simply that set of constraining specificationsfor integration, as I use the term here to represent the thecollection,storage,reduction,alteration, last stage in our processes, I am referring to a transfer,anddisplayoforganizationalfacts, complex interfiling of segments of our data the opinions, and associated documentationwhich is entries. In this` interfiling,' we produce, for each established in order to manage, command if you article or book in our file, an entry which is a com- will, and control the ultimate performanceof an posite of information from our various sources.If organization.... one of our sources omitsthe name of the publisher "With this in mind, it is possible to recognize of a book, but another gives it, the final entrywill the dangers associated with prematurelystandardiz- contain the publisher's name. If one sourcegives ing the information-processing tools, theforms, the the volume of a journal in which an article appears, data codes, the message layouts, theprocedures but not the month, and another gives themonth, for message sequencing, the file structures,the but not the volume, our final entry will contain calculations,andespeciallythe data-summary 1965, forms essential for automation. Standardization both volume and month. And so on." (Sawin, simple p. 95). of these details of a system is relatively "Normalize. Each individualprintedsource, and can be accomplished by almost anyonefamiliar which has been copied letter by letter, hasfeatures with the design of automatic procedures.However, of typographical format and style, someof which if the precise nature of the job and itsorganiza- are of no significance,others of which are the tional implications are not understood indetail, means by which a personconsulting the work it is not possible to know the exactinfluence that distinguishes the several 'elements' of the item. these standards will have on the performanceof The family of programs for normalizing theseveral the system." (Bennett, 1964, pp. 99, 103). files of data will insert appropriateinformation 2.48"There is a need for design verification. separators to distinguish andidentify the elements That is, it is necessary to have some method for of each item and rearrange it according to aselected ensuring that the design is under control and that canonical style, which for the Pilot Study is one the nature of the resulting system can bepredicted which conforms generally to that of theModern before the end of the design process. In command- Language Association." (Crosby, 1965, p. 43). and-control systems, the design cycle lasts from evolving from a simple 2.47"Some degree of standardized processing two to five years, the design and communication is at the heart of anyinforma- idea into complex organizations of hardware, soft- tion system, whether the system is thebasis for ware, computer programs,displays, human opera- mounting a major military effort, retrievingdocu- tions, training, and so forth. At all times during this ments from a central library,updating the clerical cycle the design controller must be able to specify and accounting records in a bank, assigningairline the status of the design, the impact that changes reservations, or maintaining a logistic inventory. in the design will have on the command, and the There are two reasons for this. First, allinformation probability that certain components of the system systems are formal schemesfor handling the infor- will work. Design verification is the process that mational aspects of a formally specified venture. gives the designer this control. The methods that make up the design-verificationprocess range from units may be added, removedor replaced in accord- analysis and simulationon paper tofull-scale ance with changing requirements." (Dennis and system testing." (Jacobs, 1964, p. 44). Van Horn, 1965, p. 4). See alsonotes 5.83, 5.84. 2.49"Measurement of the systemwas a major 2.53"The actual execution of datamovement area which was not initially recognized. It was commands should be asynchronous with themain necessary to develop the tools to gather data and processing operation. It should bean excellent use introduce program changes togenerate counts and of parallel processing capability." (Opler,1965, p. parameters of importance. Future systems designers 276). should give, this areamore attention in the design 2.54"Work currently inprogress [at Western phase to permitmore efficient data collection." Data Processing Center, UCLA] includes:investi- (Evans, 1967, p. 83.) gations of intra-job parallel processing which will 2.50 "[The user] is given several controlstatis- attempt to produce quantititative evaluations of tics which tell him the amount of dispersion in each component utilization; the increase in complexity category, the amount of overlap of each category of the task of programming; and the feasibilityof with every other category, and the discriminating compilers which perform the analysisnecessary to power of the variables...These statistics are convert sequential programs into parallel-pathpro- based on the sample of documents that heassigns grams." (Dig. Computer Newsletter 16, No. 4, 21 to each category...Various users of an identical (1964).) set of documents can thus derive their own structure 2.55"The motivation for encouraging theuse of of subjects from their individual points of view." parallelism in a computation is notso much to make (Williams, 1965, p. 219). a particular computation run more efficientlyas it 2.51"We will probably seea trend toward the is to relax constraintson the order in which parts of concept of a computer as a collection of memories, a computation are carried out. A multi-program buses and processors with distributed control of scheduling algorithm should then be ableto take their assignments on a dynamic basis." (Clippinger, advantage of this extra freedom to allocatesystem 1965, p. 209). resources with greater efficiency." (Dennis and Van "Both Dr. Gilbert C. McCann of Cal. Tech and Horn, 1965,pp. 19-20). Dr. Edward E. David, Jr., of Bell TelephoneLabo- 2.56Amdahl remarks that "the principalmoti- ratories stressed the need for hierarchies ofcom- vations for multiplicity ofcomponents functioning puters interconnected in large systems to perform in an on-line systemare to provide increased the many tasks of a time-sharing system." (Commun. capacity or increased availabilityor both." (1965, ACM 9, 645 (Aug. 1966).) p.38). He notes further that "bypooling, the 2.52"Every part of the system should consist of number of components provided neednot be large a pool of functionally identical units (memories, enoughtoaccommodatepeakrequirements processors and so on) that can operate independently occurring concurrently in eachcomputer, but may and can be used interchangeablyor simultaneously instead accommodatea peak in one occurring at . . at all times - the same time asan average requirement in the "Moreover, the availability of duplicate units other." (Amdahl, 1965,pp. 38-39). would simplify the problem of queuing and the 2.57"No large system isa static entityit must allocation of time and space to users." (Fano and be capable of expansion of capacityand alteration CorbatO, 1966, pp. 134-135). of function to meetnew and unforeseen require- "Time-sharing demands high system reliability ments." (Dennis and Glaser, 1965,p. 5). and maintainability, encourages redundant, modu- "Changing objectives, increaseddemands for lar, system design, and emphasizes high-volume use, added functions, improved algorithms andnew storage (both core and auxiliary) with highly parallel technologies all call for flexibleevolution of the system operation." (Gallenson and Weissman, 1965, system, both as a configuration of equipment and p. 14). as a collection of programs." (Dennis and Van Horn, "A properly organized multipleprocessor system 1965, p. 4). provides great reliability (and theprospect of con- "A design problem ofa slightly different char- tinuous operation) since aprocessor may be trivially acter, but one that deserves considerable emphasis, added to or removed from thesystem. A processor is the development ofa system that is 'open-ended'; undergoingrepairorpreventivemaintenance i.e., one that is capable of expansionto handle merely lowers the capacity of thesystem, rather new plants or offices, higher volumes of traffic, than rendering the system useless." (Saltzer,1966, new applications, and other difficult-to-foresee de- p. 2). velopments associated with the growthof the busi- "Greater modularity of the systems willmean ness. The design and implementation ofa data easier, quicker diagnosis and replacement of faulty communications system isa major investment; parts." (Pyke, 1967,p. 162). proper planning at design time to provide for future "To meet the requirements of flexibility ofcapac- growth will safeguard this investment."(Reagan, ity and of reliability, the most natural form...is 1966, p. 24). as a modular multiprocessor system arrangedso 2.58"Reconfigurationis used for two prime that processors, memory modules and filestorage purposes: to remove a unit from the system for 47

,I,...v.t.fbVEtil$kiiktgrgtiaakfri service or because of malfunction, or to reconfigure function isfail-safe. Such a system requires at the system either because ofthe malfunction of leastone moreunit of each type of system one of the units or to 'partition'the system so as component,withtheinterconnectioncircuitry to have two or more independent systems.In this to permit it to replace any of its type in any con- last case, partitioning would be used either to figuration... debug a new system supervisor or perhaps to aid "A multi-computer system which can perform a in the diagnostic analysis of a hardware malfunction satisfactory subset of its tasks in the presence of where more than a single system component were a malfunction is fail-soft. The set of taskswhich needed." (Glaser et al., 1965, p. 202.) must still be performed to provide a satisfactory "Often, failure of a portion of the system to through degraded level of operation, determines provide services can entail serious consequences the minimum number of each component required to the system users. Thus severereliability stand- after a failure of one of its type." (Amdahl, 1965, ards are placed on the system hardware. Many of p. 39). these systems must be capable of providing service "Systems are designed to provide either full to a range in the number of users and mustbe able service or graceful degradation in the face of failures to grow as the system finds more users.Thus, one that would normally cause operations to cease. A finds the need for modularity to meet these demands. standby computer, extra mass storage devices, Finally, as these systems are used, they must be auxiliary power sources to protect against public capable of change so that they can be adapted to utilityfailure, and extra peripherals and com- the ever changing and wide variety of requirements, munication lines are sometimes used. Manual or problems, formats, codes and other characteristics automatic switching of spare peripherals between of their users. As a result general-purpose stored processors may also be provided."(Bonn, 1966, program computers shouldbe used wherever p. 1865). possible." (Cohler and Rubenstein, 1964, p. 175). 2.60 "A third main feature of the communica- 2.59"On-line systems are still in their early tion system being described is high reliability. development stage, but now that systems are The emphasis here is not just on dependable hard- beginning to work, I think that it is obvious that ware but on techniques to preservethe integrity more attention should be paid to thefail safe of the data as it moves from entry device, through aspects of the problem." (Huskey, 1965, p. 141). the temporary storage and data modes, over the "From our experience we have concluded that transmissionlines and eventually to computer system reliability ... mustprovide for several tape or hard copy printer." (Hickey, 1966, p. 181.) levels of failure leading to the term 'fail-soft' 2.61In addition to the examples cited in the dis- rather than `fail- safe'." (Baruch, 1967, p. 147). cussion of client and system protection in the pre- Related terms are "graceful degradation" and vious report in this series (on processing, storage, "high availability ", as follows: and output requirements, Section 2.2.4), we note "The military is becoming increasingly interested the following: in multiprocessors organized to exhibit the property "The primary objective of an evolving special- of graceful degradation. This means that when purpose time-sharing system is to provide areal one of them fails, the others canrecognize this service for people who are generally not computer and pick up the work load of the one that failed, programmers and furthermore depend on the system continuing this process until all of them have failed." to perform their duties. Therefore the biggest opera- (Clippinger, 1965, p. 210). tional problem is reliability. Because the data at- "The term 'high availability' (like its synonym tached to special-purpose system are important and `fail safe') has now become a cliche, and lacks any also must be maintained for a long time, reliability precise meaning. It connotes a system characteris doubly crucial, since errors affecting the data base istic which permits recovery from all hardware cannot only interrupt users' current procedures but errors.Specifically,itappears to promise that als-, jeopardize past work." (Castleman, 1967,p. 17). critical system and user data will not be destroyed, u he system is designed to handle both special- that system and job restarts will be minimized and 1,, :posefunctions and programming development, that critical jobs can most surely be executed, then why is reliability a problem? It is a problem despite failing hardware. If this is so, then multi- because in a real operating environment some new processing per se aids in only one of the three char- `dangerous' programs cannot be tested on the sys- acteristics of high availability." (Witt, 1968, p. 699). tem at the same time that service is in effect. As a "The structureof a multi-computer system result, new software must be checked out during planned for high availability is principally deter- offhours, with two consequences. First, the system mined by the permissible reconfiguration time and is not subjected to its usual daytime load during the ability to fail safely or softly. The multiplicity checkout time. It is a characteristic of time-shared and modularity of system components should be programs that different 'bugs' may appeardepend- chosen to provide the most economical realization ing on the conditions of the overall system activity. of these requirements... For example, the 'time-sharing bug' of a program "A multi-computer system which can perform the manipulating data incorrectly because another pro- full set of tasks in the presence of a single mal- gram processes the same data at virtually the same 48 time would be unlikely on a lightly loaded system. center, 3) protection of the center and its gear from Second, programmers must simulate at night their fire and other hazards, 4) insist that separate facili- counterparts of laymen users. Unfortunately, these ties separate routes and used to connect locations two types of people tend to use application programs on the MIS network, and 5) build extra capacity differently and to make different types of errors; so into the MIS hardware system." (Dantine, 1966, program debugging is again limited. Therefore, be- p. 409). cause the same system is used for both service and "It is far better to have the system running at development, programs checked as rigorously as half speed 5% of the time with no 100% failures than possible can still cause system failures when they to have the system down 21/2% of the time." are installed during actual service hours." (Castle- (Dantine, 1966, p. 409). man, 1967, p. 17). "Whenever possible, the two systems run in "Protection of a disk system requires that no user parallel under the supervision of the automatic be able to modify the system, purposely or inad- recovery program. The operational system per- vertently, thus preserving the integrity of the soft- formsallrequired functions and monitors the ware. Also, a user must not be able to gain access back-up system. The back-up system constantly to, or modify any other user's program or data. repeats a series of diagnostic tests on the computer, Protection in tape systems is accomplished: (1) by memory and other modules available toit and making the tape units holding the system records in- monitors the operational system. These tests are accessible to the user, (2) by making the input and designed to maintain a high level of confidence in output streams one-way (e.g., the input file cannot these modules so that should a respective counter- be backspaced), and (3) by placing a mark in the part in the operational system fail, the back-up input stream which only the system can cross. In unit can be safely substituted. The back-up system order to accomplish this, rather elaborate schemes also has the capability of receiving instructions to have been devised both in hardware and software perform tests on any of its elements and to execute to prevent the user from accomplishirg certain these tests while continuing to monitor the opera- input-output manipulations. For example, in some tionalsystemtoconfirm thatthe operational hardware, unauthorized attempts at I/O manipula- system has not hung up." (Armstrong et al., 1967, tion will interrupt the computer. p. 409). "In disk-based systems, comparable protection 2.63"The large number ofpapers on vote- devices must be employed. Since many different taking redundancy can be traced back to the kinds of records (e.g., system input, user scratch fundamental paper of Von Neuman where multiple- area, translators, etc.) can exist in the same physical line redundancy was first established as a mathe- disk file, integrity protection requires that certain maticalrealityforthe provision of arbitrarily tracks, and not tape units, must be removed from reliable systems." (Short, 1968, p. 4). the realm of user access and control. This is usually accomplished by partitioning schemes and central 2.64 "A computer system contains protective I/O software systems similar to those used in tape- redundancy if faults can be tolerated because of based systems. The designer must be careful to the use of additional components or programs, preserve flexibility while guaranteeing protection." or the use of more time for the computational (Rosin, 1966, p. 242). tasks.... 2.62"Duplex computers are specified with the "In the massive (masking) redundancy approach spare and active computers sharing I/O devices the effect of a faulty component, circuit, signal, and key data in storage, so that the spare computer subsystem, or system is masked instantaneously can take over the job on demand." (Aron, 1967, by permanently connected and concurrently operat- P. 54). ing replicas of the faulty element. The level at "The second channel operates in parallel with which replication occurs ranges from individual the main channel, and the results of the two chan- circuitcomponents to entireself-contained nels are compared. Both channels must independ- systems." (Aviiienis, 1967, p. 733-734). ently arrive at the same answer or operation cannot 2.65"An increase in the reliability of systems proceed. The duplication philosophy provides for is frequently obtained in the conventional manner two independent access arms on the Disk Storage by replicating the important parts several (usually Unit, two corebuffersand redundant power three) times, and a majority vote...A technique supplies." (Bowers et al., 1962, p. 109). of diagnosis performed by nonbinary matrices... "Considerableefforthasbeencontinuously require, for the same effect, only one duplicated directed toward practical use of massive triple part. This effect is achieved by connecting the modular redundancy (TMR) in which logic signals described circuit in a periodically changing way are handled in three identical channels and faults to the duplicated part. If one part is disturbed the are masked by vote-taking elements distributed circuit gives an alarm, localizes the failure and throughout the system." (Ayiiienis, 1967, p. 735). simultaneously switches to the remaining part, "He must give consideration to 1) back-up power so that a fast repair under operating conditions supplies that include the communications gear, (and without additional measuring instruments) is 2) dual or split communication cables into his data possible." (Steinbuch and Piske, 1963, p. 859). 49 2.66"Parameters of the model are as follows: that any item of information must be stored in at n = total number of modules in the least two independent places, and that the updating system of queue tables and other auxiliary data must be m =number of unfailed modules needed carefully synchronized so that operation can con- for system survival tinue smoothly after correction of a malfunction. If P f =Probability of failure of each module it cannot be determined exactly where a transmis- some timedu,ringthemission. sion was interrupted, procedures should lean toward This parameter thus includes both pessimism. Repetition of a part of a message is less themissiondurationandthe grievous than a loss of part of it." (Shafritz, 1964, p. module MTBF, N2.3-3). Pnd=probabilityofnotdetecting an occurred module failure "Reference copies are kept on magnetic tapes for PSprobabilityofsystemsurvival protective accountability of each message. Random throughout the mission requests for retransmission are met by a computer Pf=1"-'11.S =--probability ofsystem failure search of the tape, withdrawal of the requiredmes- sages and automatic reintroduction of the message during the mission. nlm =redundancy factor in initialsys- into the communications system." (Jacobellis, 1964, tem. . p. N2.1-2). "Depending upon the attainable Pf and Pad, "Every evening, the complete disc file inventory the theoretical reliability of a multi-module com- is pruned and saved on tape to be reloaded the fol- puting system may be degraded by adding more lowing day. This gives a 24-hour 'rollback' capability than a minimal amount of redundancy. For ex- forcatastrophicdiscfailures." (Schwartz and ample, P f= 0.025...it is more reliable to have Weissman, 1967, p. 267). only one spare module rather than two or four, for a typical current-day Pnd such as 0.075. Even "It is necessary to provide means whereby the for a Pnd as, low as 0.03 (a very difficult Pnd to achieve contents of the disc can be reinstated after they in a computer), the improvement obtained in system have been damaged by system failure. The most reliability by adding a second spare unit to the straightforward way of doing this is for the disc to system- is minor." (Wyle and Burnett, 1967,pp. 746, be copied on to magnetic tapeonce or twice a day; 748). re-writing the disc then puts the clock back, but users at least know where they are. Unfortunately, "The probability of system failure... is the copying of a large discconsumes a lot of com- k=n puter time, and it seems essential to develop PF= P! (prk(1_pf)n -k methods whereby files are copiedon to magnetic k=(n.m+1)(nk)!k! tape only when they are created or modified. It k=(nm) n! would be nice to be able to consider the archive Pk(1 Pf)nk[1 (1 P me] and recovery problems as independent, butreasons (n_ k!)k! of efficiency demand thatan attempt should be made to develop a satisfactory common system. (Wyle and Burnett, 1967, p. 746). We have, unfortunately, little experience in this 2.67 'tine of the system design considerations isarea as yet, and are still groping our way." (Wilkes, the determination of the optimum number of redun- 1967, p. 7). dant units by means of which the required systemre. liability is to be reached. It will beseen that Pnd as "Our requirements, therefore,were threefold: well as Pf must be considered in determining thesecurity, retrieval, and storage. We investigated most economical design." (Wyle and Burnett, 1967, various means by which we could meet these p. 748). requirements; and we decided on the use of micro- "One of the prime requisites for a reliable, de- film, for two reasons. First, photographic copies pendable communications data processing system of records,including those on microfilm,are is that it employ features for insuringmessage pro- acceptable as legal representations of documents. tection and for knowing the disposition of every We could photograph our notebooks, store the message in the system (message accountability) in film in a safe place, and destroy the booksor, at case of equipment failures. The degree of message least, move them to a larger storagearea. Second, protection and accountability will vary from applica- we found on the market equipment with which tion to application." (Probst, 1968, p. 21). we could film the books and then, with a suitable "Elaborate measures are called for to guarantee indexing system, obtain quick retrieval of infor- message protection. At any given moment, a switch- mation from that film" (Murrill, 1966,p. 52). ing center may be in the middle of processing "The file system is designed with thepresump- many different messages in both directions. If a tion that there will be mishaps,so that an auto- malfunction occurs in any storageor processing matic file backup mechanism is provided. The device, there must be enough information stored backup procedures must be prepared forcon- elsewhere in the center to analyze the situation, and tingencies ranging from a dropped biton a magnetic to repeat whatever steps are necessary. This means tape to a fire in the computer room. 50 "Specifically,the following contingenciesare processed data. Also consider the possiblecosts provided for: oftransportingclassifieddataelsewherefor "1. A user may discover that he has accidentally computation: express, courier,messenger, Brink's deleted a recent file and may wish torecover service." (Bush, 1956, p. 110). it. "For companies in the middlerange, the com- "2. There may be a specific system mishap mercial underground vaults offer excellent facilities which causes a particular file to beno longer at low cost. Installations of this type are available readable for some 'inexplicable'reason. in a number of states, including New York, Pennsyl- "3. There may be a total mishap. For example, vania, Kansas, Missouri and California. In addition the disk-memory read heads may irreversibly to maximum security, they providepre-attack score the magnetic surfaces so that all disk- clericalservicesandpost-attackconversion stored information is destroyed. facilities. The usual storage chargeranges from $2 to $5 a cubic foot annually, depending "The general backup mechanism is provided by on whether community or private storage is desired. .. the system rather than the individualuser, for the "The instructions should detail procedure for more reliable the system becomes, the more the converting each vitalrecord to useable form, user is unable to justify the overhead (or bother) as well as for utilizing the converted data to per- of trying to arrange for the unlikely contingency form the desiredemergency functions. The lan- of a mishap. Thus an individualuser needs insur- guage should be as simple as possible and free 11, ance, and, in fact, this is what is provided." (Corbato of 'shop' terms, since inexperienced personnel and Vyssotsky, 1965, p. 193). will probably use the instructions in thepost- "Program roll-back for corrective action must be attack." (Butler, 1962,pp. 65, 67.) routine or function oriented since it is impractical 2.70"The trend away from supporting records from a storage requirement point of view to provide is a recent development that hasnot yet gained corrective action for each instruction. The roll- widespread acceptance. There is ample evidence, back must be to a point where initial conditions however, that their use will decline rapidly, if are available from sensors, prestored, or recon- the cold war gets uncomfortably hot. Except stitutable.Even anintermittent memory mal- for isolated areas in their operations, function during access becomesa persistent error an increasing since it number of companiesare electing to take a cal- is immediately rewritten inerror. Thus, culated risk in safeguarding basic recordsbut critical routines or high iteration rate real-time not the supporting changes. For example,some routines (for example, those which perform inte- of the insurance companies microfilm the basic gration with respect to time) should be stored in-force policy records annually and forego the redundantly so that in the event of malfunction changes that occur between duplicating cycles. the redundantly stored routine is usedto preclude This is a good business risk for tworeasons: (1) routine malfunction or error buildup with time." supporting records are impractical for mostemer (Bujnoski, 1968, p. 33). gency operations, and (2) a maximum one-year lag in 2.68"Restart procedures should be designed the microfilm record would not seriously hamper into the system from the beginning, and theneces- emergency operations." (Butler, 1962, p. 62.) sity for the system to spend time in copying vital "Mass storage devices hold valuable records, information from one place to another should and backup is needed in the event of destruction be cheerfully accepted.... or nonreadability of a record(s). Usually the entire "Redundant informationcan be included in file is copied periodically, anda journal of trans- supervisor communicationor data areas in order actions is kept. If necessary, the filecan be recon- to enable errors caused by system failure to be structed from an earliercopy plus the journal to corrected. Even a partial application of this idea date." (Bonn, 1966, p. 1865). could lead to important improvements inrestart capability. A system will be judged 2.71'The life and stability of the [storage] as much as medium under environmental conditionsare other by the efficiencies of its restart proceduresas by considerations to which a great deal of attention the facilities that it provides.... "Making it possible for the systemto be restarted must be paid. How long will the medium last? after a failure withas little loss as possible should How stable will it be under heat and humidity changes?" (Becker and Hayes, 1963,p. 284). be the constant preoccupation of the software It must be noted that, in thepresent state of designer." (Wilkes and Needham, 1968,p. 320). magnetic tape technology, the "Procedures must also be presCribed for work average accurate with the archive collection life of tape recoreq isa matter of a few months to prevent loss or only. The active master filesare typically rewritten contaminationof the master records by tapeon new tapes regularly, as a part of normal up- erasure,statisticaladjustment,aggregationor datingand reclassification." (Glaser et al., 1967, maintenanceprocedures.Special p. 19). precautions must be undertaken, however,to 2.69"Standby equipment costs should receive assure the same for duplicate master tapes, wher- some consideration,particularlyin a cold war ever located. situation:duplicatetapes, raw dataorsemi- "Security should also be considered in another 51 sense. Paper must be protected against fire and the reels are replaced periodically in updating flooding, magnetic tapes against exposure to electro- cycles. Long-term storage of microfilm, however, magnetic fields and related hazards. No special willrequireproper temperature and humidity precartion isnecessary for microfilm, provided control in the storage area." (Butler, 1962, p. 64.)

3. Problems of System Networking

3.1As noted in a previous report in this series: such as those at M.I.T. and the Systems Develop- "Information processing systems are but one ment Corporation are linked by digital transmission facet of an evolving field of intellectual activity channels. The difficulties involved in establishing called communication sciences. This is a generic computer networks appear not to be difficulties of term which is applied to those areas of study hardware design or even of hardware availability. in which the interest centers on the properties Rather, they appear to be difficulties of social and of a system or the properties of arrays of symbols softwareorganization,of conventions, formats, which come from their organization or structure and standards, of programming, interaction, and rather than from their physical properties; that is, communication languages. It is a situation in which the study of what one M.I.T. colleague calls 'the there now exist all the component hardware facilities 1.zoblemsof organizedcomplexity'."(Wiesner, that can be said to be required, yet in which there 1958, p. 268). do not now exist any general-purpose networks Theterminologyapparentlyoriginatedwith capableof supportingstage-threeinteraction." Warren Weaver. Weaver (1948) noted first that the (Licklider, 1967, pp. 5-6). areas typically taCkled in scientific research and "The state of affairs at the end of 1966 can be development efforts up to the twentieth century summarized as follows. Multiaccess-system applica- were largely concerned with two-variable prob- tion techniques and user-oriented subsystems have lems of simplicity; then from about 1900 on, power- been developed to only a relatively primitive level. ful techniques such as those of probability theory Far too many people, particularly those with a andstatisticalmechanicswere developedto scientific-applicationbent,stillhold the short- deal with problems of disorganized complexity sighted view that the real value of large, multiaccess (that is, those in which the number of variables systemsliesintheirabilityto simultaneously isvery large, the individual behavior of each present a powerful 'private computer' to each of of the many variables is erratic or unknown, but several tens or hundreds of programmers. Nearly the system asa whole has analyzable average all of the systems actually in operation are used n properties). Finally, he points to an intermediate this way. Application-oriented systems that free the region "which science has as yet little explored user of most, or all, of his concern with details of or conquered" (1948, p. 539), where by contrast computer-system characteristics and conventional to those disorganized or random situations with programming are coming much more slowly. Their which thestatistical techniques can cope, the development has been inhibited by among other problems of organized complexity require dealing things, the temporary plateau that has been reached simultaneously with a considerable number of in basic multiaccess system technology." (Mills, variables that are interrelated in accordance with 1967, p. 247). organizational factors. 3.3"The analytical tools are simply not avail- 3.2"Organizational generalityiss an attribute able..." (Baran, 1964, p. 27). of underrated importance. The correct functioning "The essence of rational benefit-cost analysis is of on-line systems imposes requirements that have the tracing of indirect as well as direct effects of been met ad hoc by current designs. Future system programs and the evaluation and summing of these designs must acknowledge the basic nature of the effects. Typically, the methodology for tracing all problems and provide general approaches to their but the most obvious linkages is entirely lacking or resolution." (Dennis and Glaser, 1965, p. 5). fails to use the relevant information." (Glaser et al., "Diversity of needs and divisibility of computer 1967, p. 15), resources demand a much more sophisticated multi- "The problem associated with providing the inter- plexing strategy than the simple communication connection of a network of processors is a major case where all users are treated alike." (David, 1966, one." (Estrin and Kleinrock, 1967, p. 92). p. 40). "Solving the data base management problems of a "As we turn toward stage three, the stage char- distributed network has been beyond the state of acterized by the netting of geographically distrib- the art." (Dennis, 1968, p. 373). uted computers, we find ourselves with a significant "Although techniques for multiplex communica- base of experience with special-purpose computer tions are well developed, we are only beginning to networks, but with essentially no experience with learn how to multiplex computers." (David, 1966, general-purpose computer networks of the kind that p. 40). will come into being when multiple- access systems "The formalism of hardward /software system 52 management is just beginning to take shape. SAGE controlling the use of such systems. These prob- is a landmark because it worked in spite of its lems have evidenced themselvesin computer immense size and complexity." (Aron, 1967, p. 50). scheduling, program capability constraints, and "The system engineer presently lacks sufficient the allotment of auxiliary storage." (Linde and tools to efficiently design, modify or evaluate com- Chaney, 1966, p. 149). plex information systems." (Blunt, 1965, p. 69). "A network has to consider with great care 3.4"The design and analysis problems associ- the many possibilities of user access which approach ated with large communications networks are fre- more and more the vast possibilities and intri- quently not solvable by analytic means and it is cacies of direct human communication." (Cain therefore necessary to turn to simulation techniques. and Pizer, 1967, p. 262). Even with networks which are not particularly large "Increased attention needs to be placed on the the computational difficulties encountered when problem of techniques for scheduling the many other than very restrictive and simple models are userswiththeirdifferentpriorities."(Bauer, to be considered preclude analysis. It has become 1965, p. 23). clear that the study of network characteristics and 3.6"Much of the design effort in a message- traffic handling procedures must progress beyond switchingtypecommunicationssystemgoes the half-dozen switching center problem to consider into the network which links the terminals and networks of dozens of nodes with hundreds or even nodal points together. The distribution of terminals thousands of trunks so that those features unique to can be shown, the current message density is these large networks can be detern,ned and used known, and programs exist to help lay out the in the design of communications systems. Here it is network. With most interactive systems thisis evident that simulation is the major study tool." not the case." (Stephenson, 1968, p. 56). (Weber and Gimpelson, 1964, p. 233). 3.7"In looking toward computer-based, com- "The time and costs involved make it almost puter-linkedlibrarysystems,thathavebeen mandatory to 'prove' the 'workability' and feasi- proposed as a national technical information net- bility of the potential solutions via pilot systems or work, studies of perceived needs among users by implementation in organizations or associations are likely to be of very little use. Instead it would which have some of the characteristics of the seem to be more appropriate to initiate small- national system and which would therefore serve as scaleexperiments designedtoproduce, on a a model or microcosm of the National Macrocosm." limited basis, the effects of a larger -scale system (Ebersole, 1966, p. 34). in order to determine whether such, experiments "A report by Churchill et al., specifically recog- produce the expected benefits." (Schon, 1965, nizes the need for theoretical research in order to p. 34). build an adequate foundation on which to base 3.8"Transmittingdatacollectionsystems systems analysis procedures. They point out that can assume a wide variety of equipment con- recent computer developments and particularly figurations, ranging from a single input unit with large computer systems have increased the need for cable-connected recorded to a farflung network research and the body of data that research can with multiple input units transmitting data to provide in such areas as data coding and file multiplerecorders or computers by means of organization." (Borko, 1967, p. 37). both common-carrier facilities and direct cable 3.5"The coming importance of networks of connections. Probably the most important parameter computers creates another source of applications inplanning the equipment configurationof a for...multiple-queue disciplines. Computer net- system is the maximum number of input stations work disciplines will also have to be dependent on that can be connected to a single central recording transmission delays of service requests and jobs unit." (Hillegass and Melick, 1967, pp. 50-51). or parts of jobssfrom one computer to another as well 3.9Licklider stresses the importance of "co- as on the possible incompatibilities of various types herence through networking" and emphasizes: "On betweendifferentcomputers.Thesynthesis the average, each of n cooperative users can draw and analysis of multiprocessor and multiple proc- n-1 programs from the files for each one he puts essornetworkprioritydisciplinesremainsa into the public files. That fact becomes so obvi- fertile area of research whose development awaits ously significant as n increases that I can conclude broad multiprocessor application and an enlighten- by concluding that the most important factors in ing experience with the characteristics of these software economics are n, the number of netted disciplines." (Coffman and Kleinrock, 1968, p. 20). users, and c, the coefficient of contributive coopera- "We still are plagued by our inability to pro- tiveness that measures the value to his colleagues gram for simultaneous action, even forthe sched- of each user's creative effort." (Licklider, 1967, uling of large units in a computing system." (Gorn, p. 13). 1966, p. 232). "The circumstances which appear to call for the "As computer time-sharing systems have evolved establishment of physical networks (as opposed to from a research activity to an operational activity, logical networks) are generally: and have incre..,sed in size and complexity, it has "1. The existence ofspecial data banks or become clear that significant problems occur in special collections of information located at a 53 single institution but useful toan audience c, all operations may be publicized." (Brown geographically dispersed. et al., 1967, p. 209). "2. The inadequacy of general data banksor general collections of informationto meet 3.12"The operator's chargesto clients must be local needs where remoteresources can be in fair proportion to theusage made of installation used in a complementary fashionto fulfill the resources (processing, time, storageoccupancy, need. etc.). Therefore adequate recordsmust be kept of 443. resource use." (Dennis, 1968, p. 375). The centralization of programming services, "A principal [individual processing capabilities or scientificresources or group of individuals] with a geographically dispersed need. is charged for resources consumed bycomputations 4.The need for interpersonal (including inter- running on his behalf. A principal is alsocharged group) direct communication. This includes for retention in the system ofa set of computing teleconferencing and educational activities. entitiescalled retained objects, whichmay be 5.A justification on economic, securityor social program and data segments. .." (Dennis and groundsfordistributionof responsibility Van Horn, 1965,p. 8). for load sharing among organizationsor geo- "The equitable allocation ofspace and time by graphiCal regions." (Davis, 1968,pp. 1-2). administrative fiat. This is probablyan overwhelm- "In certain areas, such ing problem in a network since the predictingof as law enforcement, communicationpathsand computingfacilities medicine, social security, and education, there isa required by anyuser would be quite unwieldy.... need for joint Federal-State computer communica- "Thus a more elaborate schemeseems to be tions networks which can applynew technology to necessary one whose rates are proportional to improving the management of major national pro- the value of the service. This would bemodified grams in these areas." (Johnson, 1967, p. 5), in a network because therate for the same kind of "It has been suggested thata principal advantage service may vary among the installations."(Brown to be gained from computer networks is the ability et al., 1967, p. 212). to distribute work evenly over the available installa- "Built-in accounting and analysis ofsystem logs tions or to perform certain computationsat installa- are used to provide a history of system performance tions particularly suited to the nature of the job," as well as establish a basis for charging users." (Dennis, 1968, p. 374). (Estrin et al., 1967,k). 645). "The time-sharing computer systemcan unite a group of investigators in a cooperative search for 3.13"Questions of technicalfeasibilityand the solution to a common problem,or it can serve economic value are not the sole determinants of the as a community pool of knowledge and skill on which computer utility. The development of the computer anyone can draw according to his needs." (Fano and utility may be influenced bynorms, or lack of norms, CorbatO, 1966, p. 129). about the confidentiality of data. At themoment "Within a computer network,a user of anythere do not seem to beany clear standards of cooperating installationwould have accessto good practice; perhaps therewas less need before programs running at other cooperating installations, technologygreatlyincreasedcapabilitiesfor even though the programs were written in different handling data." (Jones, 1967,p. 555). languages for different computers. This forms the "It should be easy and convenient fora user to principal motivation for considering the implementa- allow controlledaccess to any of his segments, with tion of a network." (Marill and Roberts, 1966,p. different access privilegesfor different users." 426). (Graham, 1968, p. 367). 3.10"The establishment of a networkmay lead 3.14"The designer must decide whether he will to a certain amount of specialization among the provide the high-speed service toall users, to cooperating installations. Ifa given installation, X, provide the service that the majorityrequest, and by reason of special softwareor hardware, is leave the minority to fend for themselves,or to particularly adept at matrix inversion, for example, provide the degree of speed needed in eachcase, one may expect that users at other installations in but no more. Ideally, he should know theentire the network will exploit this capability by inverting distribution ofresponse time requirements. It is their matrices at X in preferenceto doing so on even desirable to know how the arrival of these their own computers." (Marill and Roberts,1966, queries will be distributed in time throughoutthe p. 426). day. In attempting to meet requirements, hemust "An interconnected network would makeit consider what is actually being retrievedin any possible for the top specialists inany field to in- stated response interval. Does theuser want hard struct anyone within the reach of a TV receiver." copy or will he be satisfied with citations or index (Brown et al., 1967, p. 74). records? The engineering problems associatedwith 3.11"For initial network trials the advantages high-speed retrieval of hardcopy from files can be of an open systemare: formidable. If a conversational,or browsing, mode of search is used in which the searcheruses a a. ease of programming, succession of queries, dowe aim to minimize his b. services for allusers, total search time,or only to give him. immediate 54 response to each single query?" (Meadow, 1967, oriented problems, four major system equipment p. 191). performance factors must be evaluated: "Precedence is computed as a composite function 1. Real time processing capability of central of: processor(s) 1) the ability of the network to accept additional 2. Core memory size provided in central proc. traffic; essor(s) 2) the 'importance' of each user and the 'utility' 3. Bulk storage size provided of his traffic; 4. Limitations onrealtime accesstobulk 3) thedatarate of each input transmission storage." (Birmingham, 1964, p. 38.) medium or the transducer used; 3.16"It seems imperative that EDUCOM... 4) the tolerable delay time for delivery of the establish certain technical standards andopera- traffic." (Baran, 1964, p. v). tion procedures which each stateor regional group "Many separate low-data-rate devices time-shared must meet before they can be interconnected. or concentrated into a single high-data-rate link These standards should apply todigital trans- permit better averaging, as compared to a few mission,telephoniccommunications, andtele- correspondingly-higher-data-rateusers.But,as vision..." (Brown et al., 1967, p. 54). many of the high-data-rate lasers 'get in' and 'get "One final thought about integration. Integration out' fast, they have a short holding time. This helps is facilitated by the standardization of equipment, the averaging process. To be precise in this compu- processes, and languages. However, standardiza- tation, a better understanding of the number of tionincommand-and-control systems must be users, their use statistics, and the network charac- considered in the light of the evolutionary nature teristics appears mandatory. of thesesyste.tns.First,standardization should .The mixed requirement that, whilewe be based upon those elements which are mission- wish togivepriority treatment to the higher- independent; that is, the elements standardized precedence traffic of equal network loading,we should be general-purpose in nature. Secondly, must also satisfy the goal that we preserve a mini- the standardization of system elements should mum transmission capability for the lower-prece- be modular; that is, it should be possible to add dence traffic...Thus, instead of a blanket rule that other elements to them in order to modifyor all traffic of a given precedence grade will be trans- increase the capabilities of the system. If system mitted before handling the next lower precedence elements are standardized at too low a level of grade, we choose to use the time ratios of these aggregation, the system's speed of response is precedence categoriestoactasa preference increased, but its flexibility is reduced. If,on the weighting factor." (Baran, 1964, pp. 30, 33). other hand, elements are standardized atthe higher levels of aggregation, provided theseare "An added complication may be introduced not higher than the level of the designer's prob- in the form of a hierarchy of precedence classifica- lem, flexibility is increased, but there isan accom- tions. This can be a very useful feature of the com- panying reduction in the system's speed ofre- munications system, allowing important messages to sponse.Itisthistrade-off betweenflexibility avoid delay by by-passing a string of messages of and speed of response that makes the standardiza-. relatively low urgency. But it adds an extra dimen- tion problem such a difficult one." (Jacobs, 1964, sion to the message queue, requiring separate list- pp. 41-42). ings for each precedence. This system can go beyond governing of the order of transmission, and can "In general the standards of distributed-control allow high-priority messages to interrupt others systems are standards built around each class of duringtheirtransmission.Insuch a system, job for each level of job for each unique function of message switching has an advantage over circuit the system. Procedures and languages need not switching, in that an interrupted message can be be standardized across job levels or across functions. automatically retransmitted as soonas possible, Minimum standardization does not, however, imply with no further action by the sender. But the pos- the complete freedom of each functional unit to sibility of interruption necessitates that the entire selectidiosyncraticcommunicationcodesor contents of a message be retained in storage bizarre formats. Such matters as codes, formats, until its last transmi &sion is completed." (Shaf- file structures, vocabularies, and message syntax ritz, 1964, p. N2.3-3). are all aspects of perforMance programs, and the library of these program building blocks, from which "The difference between control strength and any information-processing job can be built, is priority is that control strength is used for defining bounded from above. Executive control over the interrupt classes (an interrupt classis the set limits of the library establishes the boundaries of of all requests with the same control strength), the range of alternatives available at any organiza- while priority is used for ordering requests within tional level. This is standardization of a sort, but it the same interrupt class." (Dahm etal., 1967, allows considerably more flexibility than the stand- p. 774). ardization generated by a rigid set of specifications 3.15"Inrealtimedatacommunications to be applied across functions and up and down the 55 hierarchy of information-processing jobs." (Bennett, "The least expensive method of organizing a 1964, p. 107), science information system networkappears to be 3.17"A built-in system for user feedback would on a regional basis with the centralization of acqui- be essential in determining near-future needs and sition input processes being undertaken ina central current inadequacies of the network." (Brown et al., clearing house." (Sayer, 1965, p. 142.) 1967, p. 216). 320 "In the network concept, then, the tech- "It was considered that it might be useful to nical information centers would be linked by the have all users of materials feed back their evalua- traffic routing centers. Each would become de- tions, which could be analyzed statistically for pendent upon the other with both respondingto the consideration of the next user." (Brown et al., 1967, law of supply and demand, service andcustomer p. 63). satisfaction, and continued viability basedupon "Provisionallywe characterize a network by: justificationof existence through performance." A. Remote and rapid services regarding selec- (Vlannes, 1965, p. 5). tion,acquisition, organization, storage, re- "Other choices in the spectrum may include that trieval, and processing of information and of a network of information centers in which each procedures in current files... community performs and contributes to the ad- B. Feedback to the vancement of knowledge in accordance with its 1. Originator or organizer of the informa- capabilities. Of course, a network must imposea tion (hence there would be a Community series of constraints in order to operate, but it also of users improving a common store of allows for the flexibility thata rigidly structured materials and procedures). system cannot accommodate. A network also fosters 2. Supervisorofthenetworkservices a sense of competition in which each community (hence the systeM would be adaptive to must ever strive to re-orient itself in order to survive the needs of the users)." (Brown et al., and progressinitschanging environment. In 1967, pp. 49-50). addition, each must become sensitive to the changes in the other communities in order that it may react, "In designing a priority handling system,we re-evaluate and adapt to the net set of goals that should never permit ourselves to believe that we are inevitable." (Vlannes, 1965, p. have more (or less) usable communications capa- "In order to gain control over the accountability bility than we really have. This implies network data, a telephone switchboardwas added to the status control feedback loops." (Baran, 1964, pp. system...With the formalization of the terminal 17-18). network, the concept of operation changed fromE 3.18"To facilitate system scaling, reliability, central computer with satellite terminals to the and modularity, many multi-processor operating concept of a central terminal network with satellite systems are designed to treat the processors as .~computers." (O'Sullivan, 1907,_p. 169). homogeneous system resources. Hence, there is no 3.21"Many of the largersystems must also `supervisor' processor, each schedules and controls take into account the requirements for providing itself. To prevent critical races and inconsistent machine-readable output for use in a decentralized results, only one processor at a time is permitted network of search centers. The designer must to alter or examine certain shared system data remember that other users will place constraintson bases; all other processors attempting simultaneous the parent system. It must be remembered thata access are locked-out. This phenomenon is not change to the central system has multiple effects strictly limited to homogeneous processor systems, on the various members of the decentralized net- similar requirements apply to any multi-processor work. Good system documentation will be essential scheme utilizing shared data bases." (Madnick, in provieing programs to the local search centers. A 1968, p. 19). constant training requirement will also be imposed 3.19"Some general observations may be ofupon the central system, and technical liaison must interest. There are indications that the cost of be maintained with all users in the network. Effec- operating an information system network, organized tive file maintenance procedures must be developed along subject lines, varies little with change of well in advance of implementation of the decentral- process allocation within the system. Whether all ized system. Changes and updatings to the central acquisition and input processes are carriedon in a file will occur frequently, andan adequate mecha- center clearing house or distributed in some logical nism must be available for insuring that thesesame_ manner among the service centers does not appear changes are made to all files in the field." (Austin, to make a significant difference in cost. On the other 1966, p. 245). hand, centralization in the regionally organized "Locatingthepointof minimumsufficient system becomes imperative if excess operating costs centralization for a system may call fora some are to be avoided. In a system organized to serve what atypicalphilosophy of systemdesign,a users on a project basis, there is an indication of philosophy not commonly held by theoreticians some economy of operation being achieved by on the subject but often implicit in the daily design completedecentralization."(Sayer,1965,pp. practices of the engineers and logicians involved 141-142.) in the actual specification of system details. That 56 is, a sysiem should have standardized procedures Number of for only the smallest job units that can be formally days to be specified. These fixed subroutines can then be Estimated monthly message volume studied combined to form larger routines suitable for Under 1000 20 performing larger segments of the overall activity. 1000 to 2000.. 10 At any point where two jobs are dissimilar, this 2000 to 5000 5 dissimilarity can be reflected not only in different 5000 to 10,000 3 flow diagrams but also in different formats, codes, 10,000 and over 2 sequencing procedures, indexing methods, displays, and so forth. To this extent, the system is neither Ideally the working days to be studied 'should be tailor-made nor ready -made. It is not uneconomi- chosen at random, but if for any reason a series of cally designed sithat every job is unique, nor is consecutive days must be selected, care should be it standard but ill-fitting because dissimilar jobs taken to avoid days immediately preceding or are forced into standardization. Rather, like a following holidays. In addition, the count must be made -to- measure suit, the system is built around made at each location from which information is small standardizedparts, each designed tofit sent and at which information is received." (Gentle, a small part of the overall job. The larger portions 1965, pp. 58-59.) of the system are not standardized as total units, "Calculate the Total Transmission Time. The butareuniqueconfigurationsof standardized third step, after calculating the average number of smaller parts." (Bennett, 1964, p. 108). characters per message, is to determine the average daily total transmission time. At this point a transmission speed must be assumed. This transmission 3.22Baran(1964)considers,forexample, time can be calculated by dividing the average "four separate techniques that can be used singly number of characters per message by the assumed or in combination to achieve, through automation, speed of the system. If the average message has `best' use of a seriously, degraded and overloaded 2,500 characters, for example, and the assumed communicationsplant,withintheframework transmission speed is 10 characters per second, the of a rapidly changing organizational structure." average transmission time per message will be 250 (p. v). seconds. To this figure, however, must be added some operating time for dialing the call, waiting for 3.23"Calulation of the average daily volume the connection to be established and, in some cases, and the peak volume of information to be handled coordinating the forthcoming transaction with the in the system consists of four steps: personnel at the receiving end. Operating time should be calculated from a study of a sample of 1. Calculate the average daily volume of mes- calls,butif thisisimpracticable, the system sages presently flowing in the system. designer may use 100 seconds as an overall average 2. Calculate the average number of charactersfor the operating time on each dialed-up data in each message. communications call.... "The amount of .delay to be expected during the 3. Calculate the average daily total transmission busy hour depends upon the holding time of the time. circuitatthe receiving location and the total 4. Calculate the peak volumes. number of minutes in the busy hour during which information will be received. Data communication The communicationsdesignermustplanthe planners refer to a series of charts which indicate system to handle the peak traffic loads with accept- the expected delay in transmissions when holding able delay as well as the total traffic load." (Gentle, time, circuit use, and number of circuits in the 1965, p. 58.) group are known factors. The number of incoming "Calculate Call Voitune. The first step in cal- circuits affects the probability that a calling part culating the volume of informationthat must will receive a busy. signal." (Gentle, 1965, pp. 63, 65). be handled by the data communications system is "The intervals at which messages are trans- todetermine _the number of messages (called mitted.Are these intervalsfixedor random? `traffic') handled in an average day. This is done What are the peak rates, and at what times of for traffic to and from every point in the system. day will they occur?" (Reagan, 1966,p. 23). The .volumeiscalculated by taking a sample "To determine the proper size of a dial system of severaldays'trafficand actually counting required, it is necessary to study the company's the namber of messages handled each day at busy hour, calculate the average message length, each location. The number of days to be included determine the total number of call seconds in- in the study is based upon the estimated number volved, and then consult the hundred-call-second of messages that are handled in a month. An (CCS) tables developed for telephone trunk loading. estimate of the monthly volume should be made, On the CCS tables is a listing for the number of and the following table may be used as a guide trunk lines required for a Oven loading and grade of in determining the number of days to be studied. service desired. If you want only one lost (busy) 57 call in every hundred, the tables will show how together onto one high-speed line." (Stephenson, many trunk lines are required. If you can tolerate 1968, p. 55). 10 lost calls per hundred, the tables show that "It is vital to have some knowledge of theaverage you can get by with fewer lines. In this manner, mix of message types and message lengths in the you choose the grade of service you require to system." (Stephenson, 1968,p. 55). handleyourparticulardatacommunications "The results obtained from the numerical solution problem." (Birmingham, 1964, p. 37). of a model give a precise and comprehensive de- "Wheneveritisnecessary to have a large scription of the statistical effects of high traffic. The number of stations communicate amonga large value of such precise and extensive data forcom- number of potential addresses, itis a practical puter systems may not be obvious, especially since necessity to use some form of switching. There `worst ea 3e' examples and physical reasoningcan is always a very wide variety of potential groupings establish much of the qualitative behavior of the and possible network configurations. The shape system. The two prominent facts which warrant and complexity of the resulting network isvery such analysis are the large scale ofmost multi- much dependent upon the economiesone wishes console systems, and the critical nature of machine to make in circuit groupings. The choice of these response in their conception. Because multi-con- groupingsin turn depends upon thestatistics sole systems are of such large scale, itcan be worth- of the expectedtraffic.If the trafficstatistics while economically to thoroughly evaluate proposed are known very accurately, large savings in cost designs, seeking to achieve maximum capacity. of selection of routes and assignment of channels Such design evaluation requiresa rather accurate can be realized." (Baran, 1964, p. 15). knowledge of the traffic in the various parts of the system. Because much of the effectiveness of a "Complex data communicationssystems that multi-Console systemcan be rapidly dissipated by terminate many lines in a central facility usually poor response characteristics, an accurate statistical use either a multi-line communications controller description of response is also needed. The exist- inconjunctionwithgeneral-purposecomputer ence of a capability for rapidly solving general or a specialized, stored-program communications queueing models makes this approacha much processor. These units are capable of buffering needed alternative to Monte Carlo simulationsor and controlling simultaneous input/outputtrans- experiments in traffic studies." (Fife and Rosenberg, missions on many different lines. Again,a wide 1964, pp. H1-6). variety of equipment is now availableto perform thesefunctions.Theavailabledevicesdiffer 3.24"The unequal and intermittent loading of in the number and speed of lines theycan ter- net-type channels in present communications results minate and in their potential for performing auxiliary in inefficient utilization of radio frequencies. If or independent data processing. Examples include channels were made availableto other users during the three multi-line communications controllers periods of idleness,more communications could available for use with the general-purpose IBM be handled within a given frequency band. A solu- System/360computers andtheCollinsData tion to the problem of frequency congestioncan be Centralsystem,a computer systemdesigned found in giving eachuser access to a group of especiallyfor message switching applications." channels througha system which selects an idle (Reagan, 1966, p. 26). channel for each call and releases the channelas soon as the call is terminated. Such a system may "Data rate alone, however, does not providea be called a random accesssystem, since an idle completemeasureofnetworkloading;some channel is selected at random from thegroup each devices have a short duty cycle, suchas one time a user wishes to place a call. Aftera channel is computer sending the contents of its core toa selected, a means is neededto direct the call to the remote computer. While such devices placea intended group or individual without dsturbing heavy peak demand for service, theyare highly users to whom the call is not directed. This process intermittent. On the other hand,a pulse-coded is called discrete addressing and is appliedin the telephone call places a lower peak demand load, form of tone signalling tomany systems in use today. but ties up network capacity fora longer period The combination of theterms Random Access and results in heavieraverage loading. Therefore; Discrete Address describesa class of communica- we should include an expected message-duration tion systems employing these principles andis or holding-time factor in the network-load weighting frequently referred to by theacronym, `RADA'." table." (Baran, 1964, p. 30). (Home et al., 1967, pp. 115-116). "It is necessary to havesome idea of the types "Adaptive channel communicationsystems pro- of messages that the system will be handlingso that vide efficient bandwidth utilization by allowing time estimates of transmission rate requirementscan be sharing of a small number of channels bya large made. This calculation is intimately tied in with the group of users with low duty rates. Unlike fixed distribution of terminals from the centralcomputer frequency netted systems, wherea call to a non- system. It may be economically justifiable, even a busy subscriber cannot be made if his assigned necessity, to multiplex several of the terminals frequency is in use by another of thenet members, 58 ,t)

the adaptive system will allow' completion of all 3.26"If empirical data were available in the calls provided the number of simultaneous calls is information field we could use an input-output less than or equal to the number of system channels. matrix. On the input side the various international The adaptive system is thus similar to the telephone and national information services should be in- system where each user has a private line, but the cluded. Many of these services act also as users of number of trunks connecting the lines is 1e'ss than the input from other services, and they process it the number of lines." (Home et al., 1967, p. 120). for their own purpose. Thus the output side will also contain most of the same services. The content 3.25"In the distributed network routing scheme of the matrix should be information just as an ...if the preferred path is busy, the in-transit economic matrix will contain employment, goods or Message Block is not stored, but rather sent out over services. The economist expresses his flows in a less efficient, but non-busy link. This rapid passing dollar values, since money is the conventional sub- around of messages without delay, even if a secon- stitute. A convention measuring the information dary route must be chosen, is called a 'hot-potato' flow of documents has to be developed." (Tell, 1966, routing doctrine.... p. 119). "With such a doctrine, only enough storage at 3.27"Based upon information taken in these each node need be provided to permit retransmitting [Lockheed] studies, it is quite clear that long term messages if an acknowledgement of correct receipt information exchange requirements will include is not received from the adjacent station within a wide band data and image transmission and that prescribed time interval. Message storage capacity data entry points nay ultimately involve more than is modest.... 10,000 terminals throughout the nation." (Johnson, "A dynamically-updated routing table stored at 1967, p. 7). each node indicates the best route for each Message "The issues of communication in the sense of the Block to take to reach its directed end terminal electrical transmission of data have come to the station...When two messages seek the same forefront during 1966. Most computer-system im- preferred line, a random choice is used to select plementers and users are encountering the problems which message is sent out over the best path... of communication engineering for the first time. Simulation has shown that this use of secondary Many have found disquieting the fact that the ele- paths in lieu of storage is a surprisingly effective ment of system cost arising from the necessary data- doctrine, enabling transmission of about 30-50 per communication support of large, multiaccess sys- cent of the peak theoretical capacity possible in a tems is surprisingly large often of the same order store-and-forward system having infinite storage as that of the central computer facility." (Mills, 1967, and infinite delays." (Baran, 1964, pp. 9, 12). p. 244).

4. Input-Output, Terminal Design, and Character Sets

4.1"The display-computer interface is a gen- networking via the IMP." (System Development eralized requirement of all displays and provides Corp., 1968, p. 1-13). computer buffering for the display system. In addi- 4.3"A method must be devised to develop the tion, some systems may require logic level and/or storage adress of a record from a key in the record word length changes from computer to display. itself. This is usually referred to as a randomizing These operations are also performed by the inte- formula. What is implied is an arithmetical operation face." (Mahan, 1968, p. 9). on a key in the record to develop from this key an 4.2Plans for implementation of an experimental actual storage address for the record. Study is networkforthe Advanced Research Planning required to ascertain which technique or formula Agency (ARPA) have been reported as follows: provides a good file utilization, with the least number "The ARPA contractors' network is...in the of common addresses for different keys to keep over- planning stage. As one of the nodes of this network, flow_records down." (DePais,_1965,r pp. 30-31). SDC would receive a small computer (of the PDP-8 "The logical address of a data item defines the class) as an interface message processor (IMP). All relative position of the item within the structure of other nodes in the network would likewise have a the data base. The logical address is coded so that a similar IMP. The IMP would be two-faced: one view unique code may be created for each item in the facing the local contractor's time-sharing system; data base. The logical code is a numerical repre- the other facing other node IMPs. In this fashion, sentation of the nodes in the multi-list tree structure network 'protocol would be standardized at the of the data base, and is called the Item Position IMP, while still maintaining the flexibility of per- Code." (Barnum, 1965, p. 50). mitting dissimilar local time-sharing systems to be "The technique of hash addressing by randomiz- included as nodes. An IMP will support up to two ing the input word was used to generate an address local time-sharing systems, thereby permitting local for the dictionary look-up. This method results in 59

376-411 0 - 70 - 5 the address of the first element of a chain of words in "[A] CRT display console...[should meet storage, each of which yieldedthe same random at least]three on-line capability criteria. First, address. An examination of the chain would pro- it is directly tieable to a data processing system. ceed in sequence until the word was found or until Second,ithas abilitytoinitiate messages or the last element of the chain was compared."control signals from a data entry keyboard or (Baker and Triest, 1966, p. 3-13). switches for transmission to the computer. Finally "Every word encountered in the scan of an input it has ability to receive digital messages or control text, i.e., during the actual operationof ELIZA, is signals." (Frank, 1965, p. 50). randomized by the same hashing algorithm as was "Desiretodisplaydatarapidly. ..places originally applied to the incoming keywords, hence a premium ontheefficiency of the graphical yields an integer which points to the only possiblelanguage used at the display level." (Dertouzos, list structure which could potentially containthat 1967, p. 203). word as a keyword." (Weizenbaum, 1966, p.38). "The CAFE system permits definitionof, or 4.4"The concept of on-line information control selection from, a library of pictorial elements implies the ability of such users of the system to (static and dynamic), formation of complex pictures change the performance of the system to meettheir from simpler ones, and parameter control of their own changing needs orwishes. With adequate con- individual display characteristics as well as their trol, they can experiment with the display ofalter- synchronizationintoa sequence of composite native data formats or configurations, withalternative displays.Once apictorial elementisdefined sequencesofdataretrieval,withalternative by the user-editor, it is readily available by reference formulae for summarizing, processing, or analyzing to a name supplied atthe time of its definition." data." (Bennett et al., 1965, p. 436). "Other sugges- (Nolan and Yarbrough, 1968, p. C103). tions indicate the need to reconcile data formatsof "High-levelmethodsforexpressingscope considerable variety." (Brown et al., 1967, p. 54). output and console input operationshave pro- 4.5"Special provisions (perps in the software) duced a great deal of display programming activity. may be required to preventoverlap of symbology, The obvious advantages over assembly coding adjacent aircraft tracks." of clarity, brevity, and fewer mistakes are a strong as might result from incentive for users. The compitler-compiler system (Israel, 1967, p. 208). made relatively easy the imple- 4.6"Time-shared computers need an accurate,on TX-2 has versatile clock to schedule programs, subroutines mentation and evolution of an extendedhigh- or problems, and tosynchronize the computers with level language based on ALGOL. The language, a desired timebase.Large computer systemscalled LEAP (Language for Expressing Associa- usually have their own built-in clock, but smallertive Procedures) has associative data structuring units used in a time-sharing mode must bemodified operations, reserved procedure forms for display by the addition of an external clock." (Electronics or input manipulation,and real time variables such as the clock time and tablet stylus coordinates. 38, No. 23, 194 (1965). Direct means for invoking the symbol recognizer's 4.7The question of color on output is reflected servicesare evenincorporated. A 'Recognize' first in the more conventional documentation orstatement gets a symbol fromthe tablet just as library situation. Thus it is to be noted that "the symbol from the console use of color in printing isof increasing importance", a 'Read' statement gets a but that there are questions of whether copiesin keyboard. Writing interactive programs which use color, such as the color film versions of valuable the display is straightforward, and experimentation Library or and modification can be rapid. manuscripts supplied by the Bodleian "Having LEAP availableas a programming the French Bibliotheque Nationale (Gunther, 1962, tool has facilitated the evolutionary development p. 8), can bepreserved over extended periods ofof application programs for graphical program- time. (Applebaum, 1965, p. 493). ming, data analysis,logic diagram input, and 4.8"The term 'graphical communication' pre-integrated circuit mask layout. The largest effort supposes a graphicallanguage in which pictorialhas been on circuit mask programs. A circuit information is transmitted between the designer and designer controls the mask layout program with the computer." (Lang et al., 1965, p. 1). freehand figures sketched on the Sylvania tablet. 4.9"M.StaffordofWestboro,Mass.... The computer recognizeshis rough marks as discussed the ways in which graphic communica- commands to create, move, group, and delete tions systems are currently used, withemphasis various integrated circuit components. Once a on their interfacebetween computers and infor- circuit design is complete, output tapes for each mation storage centers. Some examples oftheir of the mask levels required can be punchedfor use, he noted, are tosend signature samples later use by a precision patterns making machine. and information on accounts between a main Individual variations among designers in drawing bank and its branches, to send weather mapsstyle are accommodated easily by the trainable and technical drawings across the country, and to recognizer." (Sutherland et al., 196%. p. 632). send pages of newspaper copy between cities." "First of all, language must be concise and (LC Info. Bull. 25, App., 288 (1966).) easily learned. It should permit the user to specify 60. the various features of a thawing in the natural duced, but also the associations whichmay exist order in which they occur to him and in a continuous between parts of the picture. A display routine stream rather than in segmented form in separate within the executive threads itsway through this statements. For publication purposes, it must give ring structure transmitting the data if finds in the the user direct and ultimate control ofevery line structure to the display generator." (Forgie, 1955, drawn if he so desires. Yet, where applicable, the p. 606). user should be able to cause a particular version of 4.11"A three dimensional windowing subsystem a whole superstructure to be generated by the sys- is available for the AGT [Adage Graphics Terminal] tem merely by specifying a few simple options. in which upper and lower boundscan be placed Toward this end, the language should include the (in digital registers) on x, y, andz. The vector facility to construct higher level statements from the generator then blanks whenever the beam goes basic language statements. It is envisioned thata beyond one of the bounds, and it also tells the set of such 'user defined' statements could be de- program which bound was exceeded. This device veloped by an experienced programmer fora par- finds use in a number of applications including ticular application. Once defined, such statements uncluttering pictures, testing the dimensions and could then be used by non-programmers without intersections of solids, and splitting the CRT screen knowledge of their genesis. Preferably, the language up into rectangles allocated to different pictures, should meet the needs of users of widely varying which can then move beyond the 'edge' without computer experience. At one end of the scale it encroaching upon its neighbor's display space." should appeal to a user essentially untrained in (Hagan et al.1968, p. 753). computer programming for the simple transcription 4.12"Early graphics systems, such as the GM of drawings from a rough draft. At the other end of DAC system and Sketchpad,were little better than the scale it should satisfy a user desiring to generate automated drafting boards. This statement isnot pictures controlled by algorithm at execution time. intended in any way to belittle their efforts, but Drawing on a conditional basis is particularly attrac- merely to underline the fact that therewas very tive for applications such as circuit drawings and little that could be done with a pictureonce it had the production of musical scorei:. Finally, the im- been generated. Certain of Sutherland's illustrations plementation of this language should readilyaccom- are quite startling in their apparent sophistication, modate minor changes in syntax dictated byuser but generally return to the use of constraints (which experience. In addition, it should be designed to run were satisfied using least squares fit, which is an easily on a variety of computers, and hopefullyon a energy constraint in engineering). In endeavoring variety of terminal CRT systems, suchas the to ascribe meaning to pictures, later investigators Stromberg Carlson 4060 or the RCA Videocomp." were forced to use data structures in a more (Frank, 1968, p. 179). sophisticated manner, and it became obvious that "For the past five years, hardware has existed associations should be much more complex than the that allows a computer user to enter drawn, printed original ring structures, etc. The CORAL language, or written information directly into a computer as APL, and AL, the language described by Feldman, easily as writing with pencil on paper. The missing are all outgrowths of the need to ascribe extra asso- element that would make such devices viable ciations and meanings to a picture. Manyare now entities in computer systems has been theappro- working on this problem but information in technical priate software support. At SDC, we are developing literatureisrelatively sparse. To illustrate the the necessary programs to allow a user scientist to techniques being developed at the University of hand-print,on-line,two-dimensionalstructures Michigan, and to show the power of the associative required in the statement and solution of his prob- language, a detailed example willnow be given." lem. These programs include an on-line, real-time (Sibley et al., 1968, p. 553). character recognition program that is independent of 4.13Some other examples of experimental both position and size of the input; editing programs capabilities in this area are as follows: "GRIN that can deal with two-dimensional entities (as [(GRaphical INput) language] is particularly suit- opposed to linear strings); and contextual parsing able for use in problems requiring the extensive programs that re-structure the recognized, edited real-time manipulation of graphical information at input for subsequent processing. This work is being the console. It takes full advantage of the incre- done within the constraints of the SDC Q-32 Time- mental display structure of the scope...Thus, Sharing System.. if a display part is composed only ofa sequence "The editing facilities are simple and straight- of incremental words, its positionon the display forward, requiring a minimum of efforton the part scope can easily be changed by changing only the of the user while providing repositioning,erasure, initial absolute entry point. Also ifa part is repre- replacement and insertion for such diverse notations sented only incrementally, itcan be called up using as mathematics and organic chemical structures." the display part subroutine linkage at many places (Bernstein and Williams, 1968, p. C84). on the scope face. The part has to exist in storage 4.10"The display information is stored ina only once, however." (Ninke, 1965,p. 845). ring-type list structure, which reflects not only the "The GRIN-2 (GRaphical INteraction) language order in which parts of the display are to bepro- is a high-level graphical programming language that 61 permits the generation and manipulation of the programmer. As an example of a high-level system, graphical data structure, and providesstatements we can mention the formal language LEAP, in for controlling real-time man-machine interaction. which the programmer can easily manipulate the The interaction portion of the language is used with logical elements of his model, and the structuring the GRAPHIC -2 graphical terminal. Therest of of the information (in the form of hash-coded tables) the language pertains to thecommon data structure is performed automatically by the language system. used byall, graphical devices and terminals." At the other extreme we havea language like L6 (Christenson and Pinson, 1967,p. 705). which isa macro language useful in creating "The PLOT operator is used to createa picture arbitrary list structures. The difference between on the oscilloscope corresponding to the graphical these two 'graphical languages' isso great that one language statements ..It builds a list of 'con- could easily conceive of implementing the LEAP sole commands' (plot a point,or a line, reposition language using the L6 language. An excellent the beam etc.) known as the display file, whichare review of this subject is given by Gray." (Laurance, interpreted by the display console hardwareso 1968, p. 387). producing the desired picture." (Lang et al., 1965, "Onlytwoothercompiler-compilersystems p. 39). which cater for graphics with Computer Aided "VITAL (VariablyInitializedTranslatorfor Design in mind are known to the author. The Algorithmic Langauges), a general purpose trans- first of these, AED, due to Douglas T. Ross is lator for the Lincoln Laboratory TX-2 computer, a very long-standing and general systen of great is currently being adapted for use as a graphical interest. GULP only attemptsa small part of this control language translator." (Roberts, 1966, p. 173). generality; as AED has been justly called 'a system "The MAD language offers a powerful facility to of systems for building systems'. AEDprocesses programmers to tailor their compiler to fit their graphic language using a macro-processor ina application,viz.,the MAD operator-definition different way from the character definition of facility. Using this part of the MAD compiler, the GULP. AED is also able to deal withcontext- programmer (or programming staff) can extend the dependentlanguages,andwith more general compiler by introducing new operators and new types of precedence besides including an ALGOL- operand mode definitions into the language. In the like compiler and many special-purpose packages formulation of any graphical language, the MAD for design; e.g., POLYFACE." (Pankhurst, 1968, operator-definition facility should play a large role. p. 416). The syntax extensions we are describing in this "Itis of vital importance that the language paper are those which cannot be encompassed facilityfor the Computer-Aided Design System withintheoperator-definitionfacility. The list- includenot onlyflexibledescriptive and pro- processing facilities described in the last section gramming languages in word form, but a generalized resulted from the examination of the L6 syntax, and capability for graphical communicationas well. the identification of these elements which were There are many aspects of design in almostany beyond the scope of the operator-definition ability. field, for which the naturalmeans of expression The rest we leave for development at the MAD is in terms of pictures or diagrams, andany attempt programming level. to convey equivalent information in verbal form "In a similar fashion, we have tried to identify wouldbeextremelyunnaturaland awkward, those elements of graphical language which were and would defeat the basic principle that the beyond the scope of the MAD syntax for their designer-user be able to operate ina manner possible incorporation into the compiler. The one which isnatural to him." (Ross and Feldman, example of a graphic language which encompasses 1964, p. 15). most of the desirable elements is the LEAP language 4.14"Even within equipment classes there is of Feldman and Rovner. This language is an ALGOL- a wide variation in keyboard arrangements. Though type language which includes elements of set opera- thealphanumericsgenerallycorrespond,the tions as well as Feldman's own method of represent- availabilityand locationof specialcharacters ing graphical relations. LEAP is predicated on a isby no means standard. Functional controls highly elaborate but efficient method of data storage are even more varied, and in the case of devices involving hash coding, but the details of the imple- using complex editing features (the alphanumeric mentation do not concern us here. What does con- display device) the number, type, function, and cern us, however, is the language syntax, insofar as placement of functional controlsiscompletely itis incompatible with a MAD representation." dissimilarbetweenmanufacturers."(Auerbach (Laurance, 1968, p. 392). Corp., SDA, 1967, p. 2-10). "The subject of data structures has received a "For theinquiry/displayconsole, the major great deal of attention in the past few years, espe- problem is that of determining theproper functions cially in relation to computer-aided design. Pro- to implement for the user." (Hobbs, 1966, p. 44). gramming systems used for creating data structures "Other operator input devicesareavailable (sometimes dignified by the name 'graphical lan- onvariousconsoles.Alphanumeric keyboards guages') vary greatly in the rigidity of their repre- and function keys are used. Some function keys sentation and the types of facilities offered to the use plastic overlays for additional coding. Track 62 balls and joy sticks are preferred by some users. availableprogramminglanguages."(Newman, The Rand Tablet, which provides an easy method 1968, p. 47). for graphic input,is available as an accessory "The major problemisthe development of in several systems." (Machover, 1967, p. 158). sufficientlyversatile programming languages di- "About a half of the available buttons are left rectedatgraphic, interactive devices." (Flynn, unprogrammed, sothatevery user cantailor 1966, p. 99). the system to his own particular needs by con- "Unlike conventionalcomputer languages, structing just those operators which are useful graphical languages have received little study, and to him at a given moment. A paper overlay is used their formal properties have not been examined in to mark the labels under the buttons...The depth. The lack of precise ways to formulate and systemprovidesoperatorswhichalloweach represent graphical language fundamentals impedes user to keep his own private programmed push- the use of graphical techniques in many problem buttons and functions in a deck of cards. The areas." (Sutherland, 1967, p. 29). same programmable buttons may then be utilized 4.16"The types of phosphors used in current by different users for col,pletely different purposes alphanumeric display terminals are of relatively low simply by reading in a small deck of cards at the persistence. To present a display that is suitable start and punching out a new deck when someone for viewing and free from annoying flickering, the else wants to use the system." (Clem, 1966, p. 136). display must be continually regenerated. Buffer "Registration...is important in systems using storage is provided within the central controller or multiple display projectors or sources.Display within each display unit to store data entered locally overlays are a prime example where static informa- or received from the remote computer. Logic tion is projected over the dynamic display. It is circuitry within the controller or display unit utilizes very important that the images are properly regis- the buffer Storage to regenerate the display, usually teredwith respect to each other.Ingeneral, 30 to 40 times per second." (Reagan, 1967, p. 33). registration accuracy should meet or exceed the "Continual regeneration of a short persistence resolution of the display in order that misregistration CRT for flicker-free viewing demands high data not be detected." (Mahan, 1968, p. 5). transfer rates, or considerable buffer storage; but in

5. Programming Problrims and Languages and Processor Design Considerations

5.1"The most important need is for a design bined hardware-software programming." (Rosen, philosophy that aims at the design of total informa- 1968, p. 1446). tion processing systems, and that will eliminate "Since, ultimately, the only computer system the mostly artificial distinction between hardware of any significance to the outside world is the systems and software systems. We need a con- system composedof hardwareplussoftware, tinuing development of the trend toward com- one cannot claim to have truly designed a computer 73

t't,46gi system without having designed both...'Inte- features are desirable so that parts of problemscan grated Hardware/Software Design' requires be handled independently." (Burkhardt, 1965,p. 5). not only a knowledge of hardware design and "Automatic segmenting of code isa critical and program design, but a thorough understanding difficult problem. To be efficient,a program should of the total relationships between hardware and riot loop from segment to segment and further, all software in a working system. Effective hardware/ code not normally executed should beseparate software design means consideration of themany from the main flow of theprogram. Since program potentialtradeoffsbetween capabilityresident segments will be retrieved on demand, this reduces in the machine configuration and capability resident the program accesses." (Perry, 1965,p. 247). in the supporting software." (Constantine, 1968, "Since running programs will haveaccess to only p. 50). a portion of the memory, frequent 'page turning' will 5.2For example, at the [ACMSUNY Conf.], be necessary as the programgoes through its major "Dr. McCann talked about the need for 'Language operating pieces. Segmenting of theprogram can Escalation' (Commun. ACM 9, 645(1966)), be very difficult if it must be done manually by the while Perlis (1965), Clippinger (1965), and Burk- programmer. Assemblers or compilers with auto- hardt(:1 965), among others, discuss the general matic segmentation or semi-automatic segmentation problems of hierarchies of language. are needed." (Bauer, 1965, p. 23). 5.3"Verylarge programs,roughlydefined, 5.5"We must now master, organize, systematl.., are those that (1) demand many times the available and document a whole new body of technical experi- primary storage for retention of code and im- ence that pertaining to programming systems." manent data, and(2)aresufficiently complex (Brooks, 1965, p. 88). structurallytorequiremore thanten coders "When a programmer is asked to changea record for implementation." (Steel, 1965, format, for example, in an unfamiliarprogram, or p. 231). when a systems analyst must estimate the time "For the hypothetical air traffic controlsystem required for such a change, dncultiesmay arise used as an example, approximately forty volumes which are out of proportion to the routine nature of of several hundredpages each would be necessary the problem... to specify the program subsystems fully, In addition "Inadequate documentation may be regardedas to detailing the many functions that must be one of the most serious problems confronting com- performed by the severalprograms, limits on puter users." (Fisher, 1966, p. 26). environmentalbehavior and transfer functions "One major concern is the between the environment and the communication program library. There network are many needs here. One is an indexing system to mustbedescribed...The amount permit users to retrieve aprogram from among the of interrelated, precise andnot easily obtainable many. Another is adequate documentation of library data required is staggering." (Steel, 1965,p. 233). programs. Documentation must specify what the "The point upon which the mind should focus in program will do and under what conditions. Still the foregoing description ofcurrent large-scale another need is a linking mechanism: That is,a programming procedures is thatmany people work means for passing data from one library program to a long time to prepare a computer to do quickly anotherwithoutmanualinterventionsothat what it would takemany people a long time to do. library programs can becomesegments of com- The number of people involved ina large program- posite, larger programs. These requirements make ming task is very great, and the time it takesto it difficult today to providea convenient library of prepare the program successfully is very long. For computer processes even in a single computer, not example, the programming of the entire Semi-Auto- to mention the problems raised by machine-to- matic Ground Environment system took roughlya machine incompatibility. Today itisdifficult to thousand man-years, and theaverage output for stand on the shoulders of previous programmers." the entire staff was betweenone and ten 'debugged' (David, 1966, p. 2). and final machine instructionsper day. Despite the "The importance ofa library to provide a reposi- fact that almost anyprogrammer can write a valid tory for programs was recognized early, but full subroutine of ten instructions in ten minutes, the exploitation has been impeded bypoor program figure for the Semi-Automatic Ground Environment documentation, lack of intereston the part of pro- system does not indicate that anyone was lazyor grammers, and language problems. Many program inept. It represents the extreme difficulty ofvery- libraries now consist ofprograms in languages either large-scale programming and the essential incoher- dead or destined foran early demise. Limitations ence, insofar as complex and highly interactive result from the dearth ofprogram 'readers' and the procedures are concerned, of large, hierarchical serious practical difficulties in translation between staffs of people." (Licklider, 1964,p. 121). machine languages." (Barton, 1963,p. 170). 5.6"One of the most significant elements in the 5.4 How can very large,very complex, problems orderly, rapid assimilation of multiaccesssystem be effectively segmented forprogrammer attack? technology is adequate and appropriate documenta- "For the practical application of programming tion. 'Appropriate' is the operative word here,since languages to large problems, efficient segmentation the historical norms for system andprogram docu- 74 mentation are probably inappropriate for the future. one machine to be run on another and they provide The time is rapidly approaching when 'professional' a bootstrap for changeovers from ore equipment programmers will be among the least numerous and system to another. Examples of such programs least significant system users," (Mills, 1967,p. 227). areControlData'scomputer-aidedtranslation 5.7"However, there is another means of rduc- system to transfer 7090 programs into their own ing programming costmaking use of program 3600 Compass language (Wilson and Moss, 1965) development already done by others. At present this and a system developedto reprogram Philco is difficult to do because of poor documentation and 2000 codes into IBM 7094 language (Olsen, 1965). maintenance of programs by their authors. The 5.10For example, "Any software system would primary reason for this sad state of affairs is the become increasingly useful if it could be adapted absence of any clear incentive forprogram authors to a variety of I/O configurations." (Salton, 1966, to provide good documentation and to tailor their p. 209). programs to the demands of prospective users "If the performance of input/output functions rather than theirown private whims." (Dennis, requires specialized coding in the master control 1968, p. 376). program of a system, then altering the set of periph- 5.8"The importance of documentation in the erals or changing its i/o functions requires modifica- management of large programming developments tions of the master control programs, leading to is gendrally accepted. A number ofgroups have the.. .problemofcopingwithevolution." foundaformal system of documentation the (Dennis and Glaser, 1965, p. 9). most effective management tool at their disposal. "If systems designcan be automated,i.e., Initsmost advanced implementation, such a programmed for a computer, ultimately thecon- system of documentation ison-line to a time figuration can be selected by systematically design- sharing system available to all participatingmem- ing the systems for a variety of configurations and bers of the system programming project. selecting the configuration which willrun these "Difficulties often are related to theprogram- applications at lowest cost." (Greenberger, 1965, mer'sresistanceto documentation which may p. 278). be due to several reasons: 5J1"Severalapparently mutually exclusive Lack of tangible evidence of benefitto hisfeatures of programming languages all have their own activity. advantages .How are we toresolve these The inaccessibility of his colleagues' docu-issues?" (Raphael, 1966, p. 71). mentation because of sheer quantity, lack 5.12"Walter F. Bauer, president of Informatics of organization and common format andout Incorporated,. .predicted that in ten years all of date status. computer systems Will be online systems and that Rejection of standards, imposed forreasons 90 percent of all workon computers will involve he does not appreciate. online interaction." (Commun. ACM 9, No. 3, 645 Belief (oftenconfirmed) that he can get (Aug. 1966).) along without, and in fact feel at his creative 5.13"Multiprogramming: That operation ofa `best' when free to improvise. (serial) processor which permits the execution ofa number of programs in such a way thatnone of the "Putting a documentation system on-lineappears programs need be completed before another is to have overcome this resistance in amanner started or continued." (Collila, 1966,p. 51.) acceptable to the programmer. "A subfield of multiprogramming is concerned The system itself can help him by rejecting with the problems of computersystem organization certain types of inconsistencies. which arise specifically because of the multiplicity He has instant access to the latest versionof input-output devices which interface with the of his colleagues' work. system. The problems in this area are referred toas Standardshave beentranslatedintofor- problems of multiaccessing." (Wegner, 1967,p. 135.) matting conventions with which he is familiar. "The requirements for console languages will He understands that the system must safe-pose a formidable problem for facility designers of guard itself and his programs from unauthor- the future." (Wagner and Granholm, 1965,p. 287). ized change. Thus, he more readily accepts "The whole system is multi-programmed, there the need for authorization to change and being a number of objectprograms in core at once. implement." (Kay, 1969, p. 431). Undoubtedly, we shallsee such systems in operation and undoubtedly they will work. In thepresent state 5.9"The greatestdifficulty in programming of knowledge, however, the construction ofa super- still concerns the language to be used and the visor for such a system isan immense task, and fact that any given program isrelatively non- when constructed it hassevere run-time overheads." interchangeable on another machine unlessit (Wilkes and Needham, 1968,p. 315). has been rather wastefully written in,say ALGOL 5.14Brooks says further, "thenew systems con- or FORTRAN." (Duncan, 1967, p. x). cepts of today and tomorrow are most '-eenly Machine language translationprograms are there- programming systems concepts: efficient time-shar- fore of interest, since they allowprograms written for ing,fail-softlymultiprocessing,, effectivemass 75

376-411 0 - 70- 6

kc 1ft information retrieval, algorithms for storage alloca- operator error checks, an interrupt system . . tion, nationwide real-time Teleprocessing systems." ablelogical modules [and] dynamic relocation (1965, p. 90). mechanism' as being essential for time-sharing." 5.15"Most individuals accustomed to scientific (Davis, 1966, p. 225). computation or commercial data processing fail to "From a programmer's point of view, one of the appreciate the magnitude of the programming effort most important features of the second generation required for real-time control system implementa- of computers is the way itis possible to exploit tion." (Steel, 1965, p. 231). their automatic interruption facilities to provide "Dr. Saul Rosen of Purdue University men- control programs and operating systems. A typical tioned several fallacies of current time-sharing computer will have stored in it, more or less per- systems of which the most important is the belief manently, a control program (which may be called that manufacturers who have great difficulty pro- the 'Director', 'Master', 'Supervisor', 'Executive', ducing relatively simple software systems will etc.) whose functions are usually to arrange the somehow be better able to produce the very com- loading and unloading of independent 'object' pro- plex systems required for time sharing." (Ccmmun. grams (the programs whichactually do the work) ACM 9, No. 8, 645 (Aug. 1966).) and keep a record of the sequence of jobs they 5.16"Typical functions of such executive sys- perform, to allocate input-output devices to these tems include: priority scheduling, interrupthan- programs, and to enable thecomputer's operators dling, error recovery, communications switching to exercise the necessary control over its operation. and the important and relevant area of cataloging, It may also provide facilities for performing various accessing and manipulating information and pro- kindsofinput-outputoperation.Thecontrol gram files." (Weissman, 1967, p.30). program may be able to arrangefor several object 5.17"Today's fastest machine cannot be loaded programs to be stored in the computer at once,and down and will be idle most oF the time unless it is to 'time-share' the use of the instruction-sequencing coupled to a large number of high speed channels unit of the computer between all these programs.... and peripheral units...In order to distribute "The relationship between a control program input-output requests in a balanced flow, it must and the object programs it controls in many ways also be controlled by a complex monitor that chooses resembles that between a deity and mere mortals wisely among the jobs in its job queue." (Clippinger, the analogy extends to the permanence,, privileges, 1965, p. 207). independence and infallability' of control programs. "In some systems, more than one single program Perhaps because of this, a misconception seems isprocessed with simultaneity...Inefficiency to have grown up about the extent of their activities. often results because the mix of individual pro- Although in a computer equipped with one instruc- grams, each written for sole occupancyof a com- tion-sequencingunitthecontrol program only puter, is unlikely to demand equal loading of each `comes to life' following an interruption of the parallel element." (Op ler, 1965,p. 306). object program, and effectively expires when the "System overhead includes scheduling and the latter is resumed, it seems to be half-believed continuousprocessing of console input. These that,all the time the object program is active, functions are almost uniformly distributed, degrad- the control program is leading some kind of inde- ing the processor's execution rate by almost a pendent existence;like an all-seeing presence, constant." (Scherr, 1965, p. 14). keeping a close watch on all the activities of the objectprogram.Thismythprobablysprings "The executive is usually multipurpose. It must from experience of the behaviour of the control be designed with a balance between the conflicting program when the object program is caughtobeying requirements of (1) continuous flow or batch process- an illegal instruction: but in factthis occurrence ing, and (2) control for a demand processor in case is detected by hardware, not by the control program time-sharing consoles should be attached. In addi- itself." (Wetherfield, 1966, p. 161). tion, it usually has facilities for on-line control in "Built-inaccounting and analysisof system particular for communications switching." (Wagner logs are used to provide a history of system per- and Granholm, 1965, p. 287). formance as well as establish a basis for charging "The utility programs provide three basic func- users." (Estrin et al., 1967, p. 645). tions: the movement of data within the system 5.18"A very significant development in soft- required by time sharing or pooled procedure, the ware and one which must be given serious attention controlling of the printout of information on a pooled by the facility system designer, is the relatively basis, and the controlling of accesses to auxiliary new concept of Data Base Management." (Wagner memory." (Bauer, 1965, p. 22). and Granholm, 1965, p. 287). "From the system designers' point of view, in "The layout and structuring of files to facilitate time-sharing systems the most important thing is the efficient use of a common data base for a the supervisory program. Gallenson & Weissman wide range of purposes requires careful analyses pursue this subject in considerable detail andhigh- of the applications, the devices which store the light other features such as 'memory protection, files,and thefileorganization and processing. error checking circuitry for hardware, software and The criteria for efficiency are, as usual, maximum 76 throughput and minimum requirement for storage format, but all records within a file must be identical space, Ease of programming, program size, and in format. New files may be created or old files running time are also important considerations," changed to meet new requirements. Data can be (Bonn, 1966, p. 1866). added to files, or changes can be made to correct "Themostwidespread informationretrieval errors in existing files." (Baker and Triest, 1966, systems are those for data base file management, p. 5-1). which process records organized into fields, each "The Formatted File System (FFS) developed for containing a type of data in the record," (Hayes, the Defense Intelligence Agency is a general- 1968, p. 23), purpose data management system for the IBM 1410 "Data Management Problems. Itis interesting which is coupled to the 1410/7010 Operating Sys- to review a few of the problems raised by G. H, tem. It is oriented to a set of users (technicians) who Dobbs at the summary session of the first sym- can maintain an intimate knowledge of the structure posium on data management systems mentioned of their files and the query language to access them. above. One of these was the diverse terminology It employs both tapes and disc to define, maintain, and points of view, which make it difficult to extract and query a set of independent files. A table of any basic principles. Another was lack of concern contents and cross index can be defined and main- to input quality control. Still another was lack of tained on tape or disc. An FFS file must have a appreciation for the real-life data base problems unique key field group in each record. A single level as the user sees them, At the second symposium, of embedded files (periodic sets) is permitted in the two years later, Galantine described tin relative record. Except for the last field of a record, all fields lack of progress as `apalling.' Dobbs, at this sec- are Fixed in length. The query language permits ond session, identified several specific technical general logical conditions and relations and provides areas needing further development among these, several geographical and statistical operators. FFS is theabilityto allow an unsophisticated user to one of the few general-purpose data management describe data structures, capability to change data systems which are operational." (Minker and Sable, and file organization, ability to share files among 1967, p. 148). simultaneoususers with adequatefilesecurity "The users of non-numeric systems had require- `lockout' procedures, and the need for more flexible ments for very long alphanumeric records. Some .report formatting." (Climenson, 1966, p. 128). of the records were formatted as were unit records 5.19One example of a developmental system but the fields were not all of predetermined length. claiming to incorporate these features is the Catalog To cope with this, the formatted file concept was Input/Output System at RAND Corporation. More developed. It had the ability to handle records of specifically: "Computer applications in linguistics, variable length by referring to a data definition library science, and social science are creating a which described the permissible record contents, need for very large, intricately structured, and in context, and internal structure. The data definition some cases tentatively organized files of data. The could be carried within each record but was more catalog a generalized format for data structures normally separated into a data definition table to eliminate redundant entries. The formattedfile is designed to meet that need ...The computer programs will: could handle variable length records but could not interpret completely free form text. Special tech- a) Facilitate partitioning, rearranging, and con- niques were developed to handle free text which, in verting data from any source in preparation general, relied on the usual delimiters in the text, for writing the catalog. such as periods and commas, to identify the end of b) Format and convert data for printing on one each structuralunit.Free text then could be of a variety of printers. interpreted by scanning it as though the computer c) Sort the data elements within a catalog and were reading it from left to right." (Aron, 1968, p. 7). merge data from two or more separate catalogs. 5.21"Univac's B-0 or Flow-Matic, which was d) Restructure a file by rearranging the order running in 1956, was probably the first true Data- of classes of datacatalog transformations. Processing compiler.It introduced the idea of e) Address nodes in the structure, retrieve data file descriptions, consisting of detailed record and from the structure, and add to or delete from item descriptions, separate from the description of the structurefile maintenance." (Kay et al., program procedures. It also introduced the idea of 1966, pp. 1-2). using the English language asa programming language." (Rosen, 1964, p. 8). 5.20"In recent years there has been a rapid 5.22"General Electric has announced GECOS growth in the use of so-called 'formatted file sys- III (General Comprehensive Operating Supervisor tems'. These systems are general-purpose data III), an advanced operating system for large-scale storage, maintenance and retrieval systems de- computers. GECOS III integrates requirements for signed to provide the user with a maximum amount on-line batch, remote batch, and time-sharing into of flexibility. They feature the use of a single set ofone system using a common data base. The 'heart' programs to handle a variety of demands on a group of the GECOS III is a centralized file system of of large files. Each hle may possess a different hierarchical, tree-structured design which provides 77 multiprocessor access to a common data base, full scribing and generating a report. Another work fileprotection, and access control." (Commun. in this area is by Roberts." (Minker and Sable, ACM i i, 71 (Jan. 1968).) 1967, p. 137). 5.23"The Integrated Data Store (IDS), devel- "The file organization of TDMS isan inverted oped by Bachman of the General Electric Company, treestructurewithself-definingentries.This is a data processing programming system that relies organization has made it possible for TDMS to on linkage of all types for its retrieval and mainte- meet itsgoal of providing rapid responses to nance strategies. Through extensions to the COBOL unpredictable queries ina time-shared environ- language and compiler, IDS permits the programmer ment, Although this organization requires more to use mass random-access storage as an extension on-line, random-access storage than most other of memory," (Minker and Sable, 1967, p, 126). file organizations, the benefits obtained far out- "The IDSfilestructureallowsa linked:list weigh this storage cost." (Bleier and Vorhaus, structure in which the last item on every list 1968, p. F97). is linked back to the parent item that started the 5.25"IBM has developed a Generalized Infor- list. Thus, it is possible to return to the parent mation System based onexperience gained with item withouta recursivelistof return points. military file applications. Because of IBM's intention In IDS, each record is an element in a linked to provide this system as part of its applications list, A file of records may be subordinated to a library for the System/360 series, this system un- master record by linking itto the first member doubtedly will be examined quite closely by a variety of the subordinatefile and chaining from that of potential users. The system has two basic point, through each record in the subordinate modules: one for defining, maintaining, and retriev- file, through the last one, and back to the master ing files of 'formatted' data, and one for text process- record. There is no inherent limit in IDS to the ing and concordance-type retrieval." (Climenson, number of records that may exist in a chainor 1966, p. 126). to the number of detailed chains that may be "The text-processing module of GIS includes linked to a given record with a single master three basic files:(1) a dictionary ordered on key record. There isalso no inherent limit to the word, each record containing: pointers to synonyms depth of nesting that is permitted; i.e., a record and equivalents, key word frequency data, and a in a chain that is subordinate to a given record pointer to (2) the inverted file, which can contain a may, in turn, have subordinate record chains." variable number of document numbers indexed by (Minker and Sable, 1967, pp. 126-127). the given key word. Finally, (3) the master file can "The G.E. Integrated Data Store is an example contain bibliographic data and all words stored for of a linked file organization. Master and detail that document. Given the document number, the items are organized in a series of linked chains bibliographic data, and key words from the docu- to form records. Each chain at least one master ment, the system can automatically generate the item and one or more detail items. Each item above files." (Climenson, 1966, p. 127). contains linking or chaining information which 5.26"GRED, the Generalized Random Extract contains the addresses of the next item and the Device developed at Thiokol Chemical Corporation previous item in the chain. An item may belong and described by Heiner & Leishman, is written toseveralchains, andlinking informationto in COBOL. Files within the system follow the all chains is included in the record." (Bonn, 1966, COBOL restrictions of fixed-record size and fixed- p. 1867). field length size for each field; the files are also restricted to tape. File definitions are provided at 5.24"Franks describes the SDC Time-Shared run time by the user, who specifies the file and Data Management System (TDMS), whose design record description. A file definition library option draws upon ADAM and the earlier LUCID. TDMS is provided and can be maintained by input request. employs an interesting data structure involving The system, developed for the IBM 7010 computer, only a single appearance of each item of data has the ability to sort and output data, and is with appropriately organized pointers to represent useful for small files." (Minker and Sable, 1967, order, multiple instances, etc. This is a much- p. 147). discussed idea that has needed exploration in 5.27"At a second symposium held in September. a large system. TDMS, like too many similar 1965, a benchmark problem was used to organize systems, lacks means by which the system can the discussion of specific systems. The problem `learn'frequently traversedpaths through the involved a management data basethat is, an organ- data, a mechanism that would permit subsequent ization table and personnel files. Five systemswere identical or similar searches to be handled more given the same file data and asked to create the efficiently." (Mills, 1967, p. 240). file(s) and perform several kinds of operationson it. "Williams & Bartram have developed a report The five systems were: COLINGO of MITRE Cor- generator as part of the TDMS. The object of poration, the Mark III File Management System of this programistogive a nonprogrammer the Informatics, Inc., the on-line data management ability, while he is on-line with the system, to system by Bolt, Beranek, and Newman, Inc., the access a large file of data for the purpose of de- BEST System of National Cash Register, and the 78 `IntegratedDataStore'of GeneralElectric." referencingfacii;ltyrequiredforsucha (Climenson, 1966, p. 125). systemof multi-level,multi-subjectfiles. 5.28"This work has led to the definition of This work must take into account also the FILER(FileInformationLanguageExecutive planning and development of several kinds Routine), the formalization of a calculus of opera- of computer programs which are required tions with specific relevance to problems of file for file maintenance... organization." (Hughes Dynamics, 1964, p. 1.3). "4. To determinethekindoforganization 5.29"Apart from languages at the implementa- ofinformationwhichwillmostreadily tion or procedural level, a number of user-oriented permit questioning of the file by different systems employ list-structured data as abasic groups of questioners who have varied (and mechanism, Bernstein & Slojkowski describe a varying) requirements for the kinds of infor- system called Program Management System(PMS), mation contained in the file, and who have, whic* stores data in a two-level file structure. Asso- furthermore,requirementsfordifferent ciated with each file is a file name, a list of pointers degrees of specificityinthe information to its subfiles, and a list of pointers to otherfiles." they are seeking. (Anderson et al., 1966, (Minker and Sable, 1967, p. 126). pp. 2-4). 5.30"The long-range objectives are: 5.31"Fossum & Kaskeyexaminedifferent "1. To define some of the charActeristics of large kindsoffileorganization and investigatethe files, and to develop the structure for a large potential of a 'list ordered file' for economizing file of heterogeneous scientific and technical on the work-load on a computerwhen carrying informationincludingchemicalstructure out a search. The data base used is the terms representations, with particular attention to assigned to a number of DDC documents, using the necessity for: the DDC thesaurus, The relevance of this study a. Manipulating informationwhich is in for this chapter is the analysis of word associations, some cases formatted and in others and the objective of the analysis is, in effect, to completely amorphous. permit acertain amount of precoordination of b. Making provision for inclusion of cer- terms, in order to reduce the amount of processing tain kinds of information when it exists, necessary in the retrieval operation.The means and for later filling of gaps when the to achieving this objective is the determination information is not available at the time of the mutual exclusiveness of terms. Unlike the of file initiation, rigorousintellectualmarshallingoftermsin c. Creating a multi-level fileof information facet analysis, the procedure here is to determine so that provision may be madefor the mutual exclusiveness on the basis of whether or inclusionof general information, as not terms appear together in the indexing of well as the addition of various levels of individual documents. The attempto develop specificity as required by the user, either an economic list-organizedfile, the specific purpose simply because the additional levels of of the work reported here, is abortive, but the specificity exist and might be useful at a technique of handling terms in this way remains later date or because there is a user an intriguing possibility as amachine aid to the requirement for that degree of specificity. generation of schedules for faceted classification d. Examining the inter-structuring of files schemes." (Sharp, 1967, p. 102). that are part of the larger files but which 5.32"The fundamental importanceof data may be geographically separated. structures may be illustrated by consideringthe "2. To provide list-processing capability in the problem of designing a single language that would file structure in order to: be the preferred language either for a purely arith- a. Maintain flexibility in the file. metic job or for a job in symbol manipulation. b. Provideforanefficientmeansof Attempts to produce such a language have been updating. disappointing. The difficulty is that the data struc- c. Permitadditionstofiles,bothin tures required for efficient implementation in the classes existing already in the system two cases are entirely different. Perhaps we should and in the entry of new classes of recognize this difficulty as a fundamental one, and information. abandon the quest for an omnibus language which d. Free the system from the constraints will be all things to all men." (Wilkes, 1968, p. 5). of fixed-length and formatted files. 5.33"Historically primary preoccupation with e. Permit aggregations ofdata from files threeclassesof data structures (real-complex that are geographically separated. scalars and arrays, alphanumeric strings and files, 3.To investigate the techniques of file manipula- and symbolic lists) have led to three major language tionin order to provide systems of sub- developments; exemplifiedbut not exhaustively files,specialfiles,desirableredundancy defined by ALGOL, COBOL, and LISP, respec- in exchange for multiple access. to informa- tively. A major concern of procedural language tion, and the necessary keying or cross- designers is the reconcilization of these diverse 79

tsunA.1. 4 I-4.1.02,..1 data types and their transformations withone languages has as yet been establishedas 'standard' language." (Perlis, 1965,p. 189). even in an informal sense." (Raphael, 1966, p. 67). "Studies in artificial intelligence call for powerful "The concept of list processing,or chaining, has list-processing techniques, and involve suchopera- been used as a technique for the manipulation of tions as the placing of itemson lists, the searching logical data strings formany years and has been of lists for items according to specified keys, and formalized as a language for handling data incom- so on. Languages in which such operations can be puter storage. List processing has also been utilized easily specified are indicated. The ,ioneer language with limited success for the control of data in direct- in this regard was IPL developed Iv Newell, Simon access storage devices. In general, when list struc- and Shaw." (Wilkes, 1964,p, F 1-3). tures are used for external data control, onlya sub- 5.34"Earlylistprocessing languages were set of the possible data structures is implemented, invented in order to carry out specific projects ofa and the logical and physical relationshipsare non-numerical nature. Thesequence of languages approached as a single entity. Thus, themany ven- called IPL ..Started out in order to provide tures into this area are highly individualized, result- satisfactory methods of organizing information for ing in duplication and incompatibility." (Henry, work in theorem proving and problem solving,a 1969, p. 2). field in which the amount of storage needed isvery 5.38"List processing is a convenient mode of variable and unpredictable, and in which thestruc- description for many of themore interesting and ture of the listscarries important information. provocative areas of computer application, but the LISP... wasdeveloped partly for theuse of a availablelanguagessuchas IPLV, COMIT, project called the 'Advice Taker' whichwas in- LISP, SIMSCRIPTrequire a level of programmer tended to operate ina complex way on English state- sophistication that makes them almost prohibitive ments about situations. A language called FLPL... for normal classroom use." (Conwayet al., 1965, which was embedded in FORTRANwas produced p. 215). to write programs for proving theorems in geome- 5.39"While the list processing languages.. try. COMIT... wasdevised for languagere- make it possible to handle reasonably complicated search." (Foster, 1967,p. 3). data structures, theprogrammer is nevertheless 5.35"Furthermore, since alistisitself an faced with a number of difficult problemsas soon ordered set of items whichmay themselves be lists, as he attempts to use such languages in practice... there is no limit to the complexity of thestructures it becomes necessary to fit the given datastructures that can be built up except the totalmemory space into what often turns out to bean unnatural for- available. Also, there isno restriction to the number mat. ..At the very least,thistype of data of lists on which an itemcan appear." (Hormann, organization wastes a great deal of storagespace ... 1960, p. 14). Furthermore, the inefficientspace allocation also "Liststructureshave, astheir fundamental results in extremely slow execution times for the property, the substitutionrule that allowsany object programs." (Salton, 1966,p. 205). symbol in a list to be substituted bya list structure. "The user of list-processing languages isfre- The natural control procedure for such rulesis quently plagued by the lack ofmemory space. recu[r]sion: the use of rules whose definitioncon- Sophisticated means of 'garbage collection' have tains calls on itself." (Perlis, 1965,p. 189). become important to circumventmemory space 5.36"List structures allow dynamicstorage limitations;however, they constitutemostly a allocation for units of data larger thancan be stored palliative that seldom satisfies the user'sappetite in a single computer word. Normally blocksof con- for extra space." (Cohen, 1967,p. 82). secutive storage cellsare allocated for data sets "In their relation to machines the listprocessing prior to program execution basedon estimates of languages are burdensome, firstbecause the data the maximum size of thesesets. If the sizes of data structuresinvolved may be representedonly sets are variable or unknown, wastefulamounts of through explicit links, and secondbecause the storage must be assigned t.o guard againstprogram free growth and cancellation ofstructuresre- failure caused by overflow atsome block of assigned quires a troublesome administration ofthe storage." storage. Faced with this problem, Newell, Shaw, (Naur, 1965, p. 197). and Simon organized informationin memory into an "Powerfulsoftwareschemes(e.g.,thelist associative list structure format." (Fuller, 1963,p. 5). processing languages) have beendeveloped to "Since the order relationamong items is deter- deal with the problem of treatingscattered data mined only by links, this relationcan be changed as a contiguous string, but theypay a very heavy simply by changing the links withoutmoving the price in memory overhead (insome schemes over items physically, thus allowing simple andquick three-fourths of the availablememory is required to changesoforganizationof memorycontent. handle the addressing mechanisms)and in the Processes suchas insertion and deletion of items in processing time required to perform theaddress a list become very simple." (Hormann, 1960,p. 13). arithmetic. 5.37 For example, "list processing is relatively "An alternative solution is proposedhere which new; none of the forms for the components of list involves the addition ofa small Associative Memory 80 (AM) to the addressing machinery of the computer tures, and arithmetic procedures may be corre- (or peripheral dirF-3 Pr..,cess storage device). As spondinglydifficultto perform." (Salton, 1966, will be shown, this hardware modification will p. 208). permit scattered data to appear contiguous, with "The well-known list structure languages such as only a token overhead cost in memory andprocess- LISP were not designed for graphics and are not ing time." (Fisch ler and Reiter, 1969, p. 381). very efficient or easy to use for such multidimen- 5.40"The process of item coletion involves sional problems. They are well suited for processing the elimination of an existing item in thefile. strings of text but break down when two-way asso- This task is complicated by the fact that after ciations between list elements are the rule rather deletion, linkage information must be modified than the exception." (Roberts, 1965, p. 212). if the system isto operate properly. This can 5,44"The list processing section of the compiler be accomplished in two ways. In the first case should make it possible to handle variable length the item can be physically deleted and each list data structures, in such a way that each data item associated with the item can be searched and may be associated with a variable number of list theassociatedlinkdeleted.Alternatively,the pointers, as specified by a special parameter." specified item can be flagged in the control section. (Salton, 1%6, p. 209). Then during retrieval flagged itemsare ignored." 5.45"It might...be useful to think about a (Prywes, 1965, p. 20). general graph and tree manipulator, capable of "There are several penalties paid for obtaining performing most of the common transformationson the flexibility of a memory organized intoa list abstract graphs and trees." (Salton, 1966, p. 209). structure. One of these is the requirement for "The value of a generalized file processing system additional memory bits to store appropriate linking is directly related to the variety of file structures it addresses. A far more serious penalty is the time can accommodate. With most existing systems, the required to retrieve symbols from lists and for user is limited to structures having the form of operations required to add symbols to lists, reorder hierarchies or trees. For many problems, such lists, delete symbols from lists,erase lists, transfer structures are either not adequate or not efficient; lists to and from bulk storage, and manipulate instead, theie problems require structures of the push-down registers. These tasks are taken out form of graphs, of which the tree is a specialcase. of the hands of the programmer by various list- Examples of problems of this type are network and processing languages, but they still must beper- scheduling problems, computer graphic problems, formed by the machine." (Fuller, 1963, p. 12). and information retrieval problems." (McGee, 1968, 5.41"Mostlist-processinglanguageshave p. F68). suffered from their inability to deal directly with "Most complex data structurescan be ade- complex data structure and/or from their inability quately represented as directed graphs. A directed to perform the complete range of programming graph is defined informally as a set of labeled points language operations upon the data list structures." or nodes, and a set of directed line segments or (Lawson, 1967, p. 358). arcs connecting pairs of nodes. An arc running from 5.42". ..In many cases, especially among the node x to node y denotes that the entities repre- list-oriented languages, they simply have not geared sented by x and y are related in some way, e.g., themselves to the large amounts of data and datax is superior to y, or x precedes y. If x and y bear processing required in this special field." (Simmons, multiple relationships to each other, they may be 1966, p. 214). connected by multiple arcs, each arc being labeled "The reason list processing has yet to acquire with the relation being represented." (McGee, 1968, universal popularity in non-numeric data processing p. F68). is that, while these mechanics are easy in the list 5.46"Since Rover hasmore features that are processing language, they can be very time-consum- common to IPLs than any other language, it may ing when performed by a computer, so much so be said to be a member of the IPL family. Some that the economics of use of present-daylist processing languages in most information retrieval additional features are present in Rover which applications is questionable. Because of the saving attempt to provide a more powerful tool for syn- in programming time that is possible, however, we thesizing complex processes.... may anticipate future hardware and software devel- "The original design of IPLmemory structure opments that will enable these methods to be made has only DLs [down links] which, by themselves practical." (Meadow, 1967, pp. 200-201.) can form lists and achieve a great flexibility. We "This suggests that a conventional linked list are introducing ULs [up links] here in order to give structure is inadequate for representing map (or greater flexibility and to overcome some of the difficulties encountered by having only DLs.. ." picture) structure. Alternative forms might be an "Since the order relationamong items is deter- associative memory or a more general linked ele- mined only by links, this relationcan be changed ment structure." (Pfaltz et al., 1968, p. 369). simply by changing the links without moving the 5.43"It must be said also that numeric informa- items physically, thus allowing simple and quick tion is often awkward to store in terms of list struc- changesoforganizationof memorycontent. 81 Processes such as insertion and deletion of items in and constraints in such a way as to facilitate the a list become very simple.... processing of these associations." (Roberts, 1965, "Another disadvantage is the loss of ability to p. 211). compute addresses. Since the order relation is "A block of elements collects many ties together introduced to a set of items only by extending and thus allows the multi-dimensional associa- `adjacency' defined by links, the 'table look -up' tions required for graphical data structures .. . feature is lost. A block is formed from a sequence of registers "A part of the seemingly lost feature is regained, of any length and contains a blockhead identifier and furthermore, an added feature is gained by at the top, a group of ring elements and any number introducing 'side links' which are extension of ULs of data registers...Blocks are used to represent and DLs.... items or entities and the rings form associations "The first item on a list... usesits UL to store between blocks." (Roberts, 1965, pp. 212-213). the name of the list [the location of the cell that "There are two operators for moving around contains the name of the first item on the list]. classes, one to go through all members of a class The last item of a list contains the name of the first (arrows leaving this item), and one to find all the item as its DL as well as the 'end bit' description, classes an item belongs to (arrows pointing at "This internal convention gives a list the effect this item)." (Roberts, 1965, p. 214). of being a ring; having reached to the end item, the "Before a new free block is added to the free processor has an access to the first item aswell as lists, the block just after it in memory is examined the name of the list without backtracking." (Hor- and if it is free also, then the two blocks are merged mann, 1960, pp. 2, 13, 15-16, 21). into one larger free block. This merging technique makes the allocation of variable length blocks 5.47"The Association-Storing Processor (ASP) almost as efficient as the allocation of fixed length consists of a language designed to simplify the blocks." (Roberts, 1965, p. 213). programmingofnon-arithmetic proble Y to- 5.50"In the current DEACON work, data is gether with a number of radically new r < ~chine organized into ring structures. These structures are organizations designed to implement this language. similar in many respects to the plex structures These machine organizations are capable of high- defined by Ross and used by Sutherland in Sketch- speedparallelprocessing, and take advantage pad, and are an extension of the notion of list of the low cost of memory and logic offered by structure." (Thompson, 1966, pp. 354-355). large scale integration (LSI). 5.51"SLIP was the first embedding of a list- "The ASP concept has been developed spe- processing capability within a higher level language cifically for applications having one or more of the and has a formative ring structure. The idea of following characteristics: rings was crystallized by Sutherland and Roberts (1) The data bases are complex in organization and used with data systems designed primarily for and may vary dynamically in both organiza- graphics and computer-aided design. Roberts has tion and content. also developed a language to refer to rings (Class (2) The associated processes involve complex Oriented Ring Associative Language: CORAL). combinations of simple retrieval operations. In such languages, the associations are built into (3) The problemdefinitionsthemselves may the structure by allowing blocks of information to change, often dramatically, during the life of be threaded by rings which carry the associations the system." (Savitt et al., 1967, p. 87). between the blocks of data. This isillustrated ..,where JOHN's"parent of ring. goes through 5.48"Sketchpad'stopological fileusesa `NUB C' and 'NUB A', which in turn reference special ring structure that permits storage structure `Edith' and 'Arnold' respectively. Thus John is rearrangement with a minimum of searching ... the parent of both Edith and Arnold. A similar The use of redundant interconnecting blocks in its structure has been implemented with PL/1 by Dodd. liststructure givesitthe appearance of being The duality of certain relationships, such as: 'de- ordered more like a tree. This allows fast forward fined by' and 'defines' or 'to the left of' and 'to the and backward searches for subgroups. The structure right of, etc., led to the need for a connector block, used insures that at most two steps must be taken here illustrated by the three NUBS. in essence, the to find either the header block or the previous NUB represents a two-way switch for transferring element." (Coggan, 1967, p. 6). out of one ring and into another. The subroutines 5.49"If some connection is deleted, it is not or macros pass along the ring until they arrive at a sufficient to just remove the line from the display NUB. They 'switch' it, and pass into the other ring, file. That line represents an association between passing along the second ring (and others as found) two elements and may constrain the movement picking up information until they return to the of the elements, the distance between them and original NUB and re-enter the first ring. This allows their deletion, as well as their external properties answers to questions such as 'Who is the mother of (e.g., electrical, program flow, etc.). Thus, each Arnold?' as well as 'Who is the sou of Mary?' One line or other picture element must be attached of the major disadvantages of these structures to an undetermined number of graphical elements occurs on adding a new, not previously anticipated, 82

St 1r association. The operation eitheris impossible, requiring a complete re-compilation, the immediate disposal ofa human user. Evidence or else clumsy, of today's great interestin on-line programming patching on additional blocks . .and requiring systems is that more andmore of them are being sophisticated garbage collectors. Arecent survey used." (Sutherland, 1965,p. 9). by Gray describes these andsimilar structures." (Ash and Sibley, 1968, "The on-line nature oftime-sharing permits direct p. 145). man-machine communication 5.52"An interesting variation in languages thatare on the basic list beginning to approach naturallanguage, ata pace processing techniquewas described by Cheydleur approaching normal human in 1963. In this technique,a datum is stored just conversation, and in once in memory, but provision is made for some applications, at graded difficultylevels appro- many priate to the skill andexperience of the user." pointers both to andaway from a given element, (Sackman, 1968,p. 1). Data redundancy of normallist processing is elim- "To improve all inated at theexpense of addressing redundancy. our on-line systems, we need more :Indbetter languages ofcommunication The address bookkeepingbecomes quite complex, between the but it is interestingto note that Cheydleur's man and the machine whichare concept `natural' in thesense that they are easy touse and is applicable to cells ofvarying sizes. Hesuggests fit the task. Why can't I methods for partitioningstrings of symbols write mathematicalequa- so that tions which look like mathematicalequations and there is a reasonable distributionof pointers through- have the machine out the store. That is, if it accept, compile and perform can be predicted that the them? Why can't I describenetwork problems co-occurrence of individual lettersor words will be to of high frequency, these the computer bymeans of the picture showing the co-occurrences would be network? Why can't I,in filter design, place poles stored in thesame cell. Since the storagespace taken by pointers and zeroson the complex plane? Theanswer in competes with data storage each case is: Ican in principle, but not in practice. requirements, methods for balancingcell size and As yet, the techniques pointer distributionare quite important." (Climen- which let me do these things son, 1966, p. 114). are not widely used." (Sutherland, 1965,pp. 11-12). 5.53, "Ring structuresare adequate for storing 5.57"Overhead computationcan be thought of a wide range of richly interrelateddata that is as degrading the effective operatingrate of the pertinent to such functionsas intelligence analysis, processor ac, seen by the user." (Scherr, 1965,p. 28). management planning, and decisionmaking. Typi- 5.58Examples include, butare not limited to, cal of these functionsare resource allocation prob- the following: "With each lems, in which the pertinent computation there is data is an inventory of associated a set of informationto which it requires the resources, theircharacteristics, and their inter- a high density (in time) of effective reference. relations. This type of datais specifiable in ring The membership of thisworking set of informa- structures." (Craig et al., 1966,p. 366). tion varies dynamically during the 5.54 "APLwas conceived at the General Motors course of the Research Laboratories computation." (Dennis and Van Horn,1965, p. 11). to satisfy the need forcon- "Various problemareas make different demands venient data association anddata handling tech- niquesin for data structures and forsyntax. The argument a high-level language. Standing for is not to restrict specializationto transliteration. ASSOCIATIVE PROGRAMMINGLANGUAGE, it It is to avoid is designed to be embedded unnecessary diversity and to achieve in PL/1 as an aid to the necessary diversity by specializing in the various user dealing with data structures in whichassocia- directions at tions are expressed." (Dodd, as high a point on the scale as possible 1966, p. 677). and thus by handlingthe bulk of the language- 5.55"The principal tool used bythis system is implementationprocess in a uniform way with the Associative StructurePackage (ASP7) implemented by W. A. Newman for common facilities." (Licklider, 1965,p. 181). the PDP7. It is used "Most writers ofpapers on on-line applications to build data structures whoseassociative proper- of time-shared ties are expressed using systems are universal in theiragree- rings, where a ring isa ment on the importance of interactivelanguages. series of addresses in wordsso that the first points They also to the next, and so seem to concur that such languages will on until the last points back to differ substantially from theprogramming languages the first. The first pointeris specially marked and now in existence, such as FORTRAN, ALGOL, is called a ringstart. Theringstart may point to and itself; i.e., the ring IPLV. The difference is primarilydue to thenew may be null." (Pankhurst, 1968, kinds of operations madepossible by the remote p. 410). console through which 5.56"On-line interaction introduces communication between into the man and computer takes place." (Davis,1966, language picture the possibilityof 'conversation'. p. 229). This possibility, together withthe need to bringon- "In programmingas well as in hardware design line languages abreast ofconventional programming languages, and system organization,time sharing calls fornew opens an inviting field to research and departures. Perhaps themost significant is the development." (Licklider, 1965,p. 124). one referred to by the term 're-entrant programs'. We note also the following:"On-line Program- When many computer ming Systems put the programs are currently raw power of a computer at active, it is likely that several ofthem are doing 83 essentially the same thing at the same time. When- either 'born in the U.S.' or 'naturalized'. If these ever that is the case, efficiency in use of memoryterms have been previously defined, the computer space would be gained if the several programscan find an answer to the question; additionally, shared a common subprogram." (Licklider, 1965, the next time it is asked about eligible voters, it p. 26). will know what is meant." (Carr and Prywes, 1965, 5.59"In artificial intelligence problems, this pp. 88-89). process [code, run & see, code] must often be 5.63"Each user, and each user's program, must prolonged beyond the debugging phase.Itis be restricted so that he and it can never access importantfortheprogrammertoexperiment (read, write, or execute) unauthorized portions of withthe working program, making alterations the high-speed store, or of the auxiliary store. This and seeing the effects of the changes. Only in this is necessary (1) for privacy reasons, (2) to prevent a way can he (evaluate it or extend it to cover more defective program from damaging the supervisor general cases." (Teitelman, 1966, p. 2). or another user's program, and (3) to make the "This appears to be the best way to use a truly operation of a defective program independent of the interactive man-machine facilityi.e., not as a state of the rest in store." (Samuel, 1965, p. 10). device for rapidly debugging a code representing 5.64"The TRAC (Text Reckoning And Com- a fully thought out solution to a problem,but piling) language system is a user language for control rather as an aid for the exploration of problemof the computer and storage parts of a reactive solving strategies." (Weizenbaum, 1966, p. 42). typewriter system." (Mooers, 1966, p. 215). "If computers are to render frequent and inten- "A solution to this problem is to use a machine- sive service to many people engaged in creative independentcomputerlanguage,designedto endeavors(i.e., working on new problems, notoperate with a reactive typewriter, to operate the merely resolving old ones), an effective compromise local computer. With this method, the computer between programming and ad hoc programming is acts in place of the human controller to gain access required." (Licklider, 1965, p. 181). to remote computer systems. This approach is 5.60"Programs very likely to contain errors possible only with an extremely versatile language, must be run but must not be permitted to interfere such as the TRAC language....It is relatively with the execution of other concurrent computa- easy to describe in TRAC the set of actions which tions. Moreover, it is an extremely difficult task to must be taken in order to make the remote computer determine when a program is completely free ofperform and bring forth the desired files." (Fox errors.Thus, in a large operational computer et al., 1966, p. 161). system in which evolution of function is required, 5.65"The basic property of symbolic languages it is unlikely that the large amount of programming is that they can make use in a text of a set of local involved is at any time completely free from errors, symbols, whose meaning and form must be declared and the possibility of system collapse through soft- within the text (as in ALGOL) or is to be deduced ware failure is perpetually present. It is becoming by the context (as simple variables in FORTRAN)." clear that protection mechanisms are essential to (Caracciolo di Forino, 1965, p. 227). However, "it any multiprogrammed computer system to reduce is...regrettable from thestandpoint of the the chance of such failure producing catastrophic emerging real-time systems that languages like shutdown of a system." (Dennis, 1965, p. 590). COBOL are so heavily oriented toward processing 5.61"The functional language makes no refer- of sequential tape file data." (Head, 1963, p. 40). ence to the specific subject matter of the prob- 5.66Someotherrecentexamplesinclude lem...The program must be organized to separate LECOM, L6, LISP II, CORAL, and TREET, char- its general problem-solving procedures from the acterized briefly as follows: application of these to a specific task." (Newell and "The compiler language, called LECOM, is a Simon, 1959, p. 22). version of COMIT, and is especially designed for "There has been a shift away from a concern small(8K)computers. The microcategorization with difficulty andt9w-ard a concern with general- program was written in LECOM, and assigns an ity. This means--both a concern that the problem appropriate syntactic category_ to each word of an solver accept a general' anguage for the problem input sentence." (Reed and Hillman, 1966, p. 1). statement, and that the internal representation be "Bell Telephone Laboratories' Low-Level Linked very general." (Newell, 1965, p. 17). List Language (L6, pronounced 'L-six') contains 5.62For example, "Multilang is a problem- many of the facilities which underlie such list oriented language that translates the user's state- processors as IPL, LISP, COMIT and SNOBOL, ment of the problem into requests for relevant but it permits the user to get much closer to machine programs and data in the system's memory. The code in order to write faster-running programs, to language was designed specifically to assist in prob- use storage more efficiently and to build a wider lem-solvhig and, in so doing, to 'accumulate knowl- variety of linked data structures." (Knowlton, 1966, edge'. For example, it may not recognize the term p. 616). `eligible voter', but it can be told that an eligible "L6...is presently being used for a variety voter is a thing that is 'human', 'age over 21' and of purposes, including information retrieval, simu- 84

V/ 'alliketVdd I lation of digital circuits, and automatic drawing of pendence for transition to other computers, and flowcharts and other graphical output." (Knowlton, otherwiseforcompatibility with hardware... 1966, p. 617). [and] better documentation (compatibility among "The most important features of Lsix which programsanddifferentprogrammers)."(Burk- distinguish it from other list processors such as hardt, 1965, p. 4). IPL, LISP, COMIT and SNOBOL are the availa- "The user needs to employ data structures and bility of several sizes of storage blocks and a flexible processes that he defined in the past, or that were means of specifying within them fields containing defined by colleagues, and he needs to refresh his data or pointers to other blocks. Data structures understanding of those objects. The language must are built by appropriating blocks of various sizes, therefore have associated with it a metalanguage defining fields (simultaneously in all blocks) and and a retrieval system. If there is more than one filling these fields with data and pointers to other working language, the metalanguage should be blocks. Available blocks are of lengths 2" machine common to all the languages of the system." (Lick- words where n is an integer in the range 0-7. The lider, 1965, p. 185). user may define up to 36 fields in blocks, which "The over-all language will be a system because have as names single letters or digits. Thus the D all the sublanguages will fall within the scope of one field may be defined as bits 5 through 17 of the metalanguage. Knowing one sublanguage will make first word of any block. Any field which is long it easier to learn another. Some sublanguages will enough to store an address may contain a pointer be subsets of others." (Licklider, 1965, p. 126). to another block. The contents of a field are inter- 5.68"The most immediate need is for a general preted according to the context in which they are compiling systemcapableof implementing a used." (Housden, 1969, p. 15). varietyof higher-levellanguages, including in "LISP 2 is a new programming language designed particular, string manipulations, list processing facil- for use in problems that require manipulation of ities, and complete arithmetic capabilities." (Salton, highly complex data structures as well as lengthy 1966, p. 208). arithmeticoperations.Presentlyimplemented 5.69Licklider, 1965, p. 119. on the AN/FSQ-32V computer atthe System "It will be absolutely necessary, if an effective Development Corporation...A particularly procognitive system is ever to be achieved, to have important part of the program library is a group excellent languages with which to control processing of programs for bootstrapping LISP 2 onto a and application of the body of knowledge. There new machine.(Bootstrappingisthestandard must be at least one (and preferably there should be method for creating a LISP 2 system on a new only one) general, procedure-oriented language for machine). The bootstrapping capability is suffici- use by specialists. There must be a large number of ently powerful so that the new machine requires no convenient, compatible field-oriented languages for resident programs other than the standard monitor the substantive users." (Licklider, 1965, p. 67). system and a binary loader." (Abrahams et al., 5.70"There is, in fact, an applied scientific 1966, pp. 661-662). lag in the study of computer programmers and "Thisliststructureprocessing system and computer programming a widening and critical language being developed atLincoln iscalled lag that threatens the industry and the profession CORAL (Class Oriented Ring Association Lan- with the great waste that inevitably accompanies guage). The language consists of a set of operators the absence of systematic and established methods for building, modifying, and manipulating a list and findings and their substitution by anecdotal structure as well as a set of algebraic and con- 4inion,vestedinterests,andprovincialism." ditional forms." (Roberts, 1965, p. 212). (Sackman et al., 1968, p. 3). "TREET is a general-purpose listprocessing "Work on programming languages will continue system written for the IBM 7030 computer at to provide a basis for studie9 on languages in gen- the MITRE Corporation. All programs in TREET eral, on the concept of grammar, on the relation are coded as functions. A functionnormally has between actions, objects and words, on the essence a unique value (which may be an arbitrarliycomplex ofimperativeanddeclarativesentences,etc. list structure), a unique name, and operates with Unfortunately we do not know yet how to achieve a zero or more arguments." (Bennett etal., 1965, definition of programming languages that covers pp. 452-453). both their syntactic and pragmatic aspects. To this 5.67"The growing importance of the family goal a first step may be the thorough study of special concept accentuates the need for levels of soft- languages, such as programming languages for ware. These levels of software will be geared to machine tools, and simulation languages." (Carac- configurationsizeinsteadoffamilymember. ciolo di Forino, 1965, p. In other words, the determining factor will be the 5.71"As Levien & Maron point out, and Bobrow amount of memory and the number of peripheral analyzes in detail; natural language is much too units associated..." (Clippinger, 1965, p. 210). syntactically complex and semantically ambiguous "Theadvantagesof high-levelprogramming to be efficient for man-machine communication. An languages...[include]higher machine inde- alternative is to develop formalized languages with 85 a simplified syntax and vocabulary. Examination of 5.75"A systems problem that has received several query languages, for example, COLINGOconsiderable attention is how to determine which and GIS, reveals a general (and natural) depend-data should be in computermemory and which ence on, and adaptation of, the rules of formalshould be in the various members of themass logic. However, even with English words for logical storage hierarchy." (Bonn, 1966, p. 1865). operations, relations, and objectnames, formal "The key requirement in multiprogrammingsys- query languages have been a less-than-ideal solution tems is that information structures be represented to the man-machine communication problem. Ex- in a hardware-independent form until the moment cept for the simplest queries, punctuatinga nested of execution, rather than being convertedto a Boolean logical statementcan be tricky and can hardware-dependent form at load time. This require- lead to errors. Furthermore, syntacticproblems ment leads directly to the concept of hardware- aside, a common difficulty arises when theuser independent virtual addressspaces, and to the does not know the legalnames of the data for which concept of virtual processors which are linked to he is searching or the structural relationshipsamong physicalcomputerresourcesthroughaddress the data items in the data base, whichmay make mapping tables." (Wegner, 1967, p. 135). one formulation of his query very difficult andex- 5.76"With respect to the central processing pensive to answer whereasa slightly altered one unit, the major compromise of future needs with may be simple to answer." (Minker and Sable, present economy is the limitation on addressing 1967, p. 136). capacity." (Brooks, 1965, p. 90). "The possibility of user-guided natural-language "Other major problems of large capacitymem- programming offers a promise of bridging theman- ories are related to the tremendous amount of elec- machine communicationgap that is today's greatest tronic circuitry required for addressing andsens- obstacle to wider enjoyment of the services of the ing." (Kohn, 1965, p. 132). computer." (Halpern, 1966,p. 649). 5.77For example, "the problem of assigning "Such a language would be largely built by the locations in name space for procedures thatmay be users themselves, the processor being designed to referenced by several system functiols andmay facilitate the admission ofnew functions and nota- perhaps share references to other procedures, is tion of any time. The user of sucha system would not widely recognized and leads to severe complica- begin by studying not a manual ofa programming tions when implementation is attempted in the language, but a comparatively fewpages outlining contextofconventionalmemory addressing." what the computer must be told about the location (Dennis and Glaser, 1965, p. 6). and format of data, the options it offers inoutput media and format, the functions already available 5.78"A particularly troublesome phenomenon, in the system, and theway in which further func- thrashing, may seriously interfere with the perform- tions and notation may be introduced. He would ance of paged memory systems, reducing comput- then describe the procedure he desired interms ing giants (Multics, IBM System 360, and others natural to himself." (Halpern, 1967,p. 143). not necessarily excepted) to computing dwarfs. 5.72"Further investigation is required in search- The term thrashing denotes excessive overhead and ing and maintaining relationships represented by severe performance degradation or collapse caused graph structures, as in by too much paging. Thrashing inevitablyturns a 'fact retrieval' systems. shortage of memory space into a surplus of Problems in which parts of the graph exist inone processor store while other parts are in another store must time." (Denning, 1968, p. 915). be investigated, particularly whenone breaks a 5.79"...Global weather prediction. Herea link in the graphs. The coding of data and of rela- three-dimensional grid covering the entire world tions also needs much work." (Minker and Sable, must be stepped along through relatively short 1967, p. 151). periods of simulated time to producea forecast in a 5.73"This program package has been used in reasonable amount of time. This type of problem the analysis of several multivariate data bases, with its demand for increased speed in processing includingsociologicalquestionnaires,projective large arrays of data illustrates the applicability ofa test responses, and a sociopolitical study of Colom- computer designed specifically for array processing." bia. It is anticipated that theprogram will also prove (Senzig and Smith, 1965, p. 117). useful in pattern recognition,concept learning, "Most engineering data is best represented in the medical diagnosis, andso on." (Press and Rogers, computer in array form. To achieve optimum capa- 1967, p. 39). bility and remove the restrictions presentlyasso- 5.74"The execution ofprograms at different ciated with normal FORTRAN DIMENSIONed installations whose total auxiliarystorage capacities array storage, arrays should be dynamically allo- are made up of different amounts of random access cated. Dynamic allocation of data ahieves the storage media with different access characteristics following: can be facilitated by the organization of the auxiliary "1. Arrays are allocatedspace at execution storage devices into a multilevel storage hierarchy time rather than at compilation time. They and the application of level changing." (Morenoff areonlyallocated the amount ofspace and McLean, 1967,p. 1). needed for the problem being solved. The 86

Itta-.114221a1,41 size of the array (i.e., the amount ofspace access memory, and the use by many processors of used) may be changed atany time during peripheralandinput/outputequipment.This program execution. If an array is not used implies that high speed switching devicesnot now during the execution of a particular problem, incorporated in conventional computers be devel- then no space will be allocated. oped and integrated with systems." (Bauer, 1965, "2. Arrays are automatically shifted between p. 23). See also note 2.52. primary and secondary storage to optimize 5.85"The actual execution of data movement the use of primary memory. commands should be asychronous with the main "Dynamic memory allocationisa necessary processing operation. It should be an excellentuse requirement for an engineering computer system of parallel processing capability." (Opler, 1965, capable of solving different problems with different p. 276). datasizerequirements. A dynamic command 5.86"Work currently in progress [at Western structured language requiresa dynamic internal Data Processing Center, UCLA] includes: investi- data structure. The result of dynamicmemory gations of intra-job parallel processing which will allocation is that the size ofa problem that can attempt toproduce quantitative evaluations of be solved is virtually unlimited since secondary component utilization; the increase in complexity storage becomes a logical extension of primary of the task of programming; and the feasibility of storage." (Roos, 1965, p. 426). compilers which perform the analysisnecessary 5.80 "Any languagewhichlacksprovision to convert sequential programs into parallel path for performing necessary operations, suchas bit programs."(DigitalComputer Newsletter16, editing for telemetered data, forces theuser to No. 4, 21 (1964)). write segments in assembly language. This destroys 5.87"The motivation for encouraging theuse the machine independence of the program and of parallelism in a computation is notso much to complicatesthecheckout."(Clippinger,1965, make a particular computationrun more efficiently p. 207). as it is to relax constraints on the order in which 5.81"Thus one must consider not only whether parts of a computation are carried out. A multi- the logical possibilities of anew device are ignored program scheduling algorithm should then be able when one is restricted to a binary logic, but also to take advantage of this extra freedom to allocate whether one is sufficiently using the signals when system resources with greater efficiency." (Dennis only one of the parameters characterizing that and Van Horn, 1965, pp. 19-20). signal is used," (Ring et al., 1965,p. 33). 5.88"The parallel processing capability ofan 5.82"For a variety of reasons, not the least associative processor is well suited to the tasks of of which is maturing of integrated circuits with abstractingpatternpropertiesandofpattern their low cost and high density, centralprocessors classificationbylinearthresholdtechniques." are becoming more complex in their organization." (Fuller and Bird, 1965, p. 112). (Clippinger, 1965, p. 209). 5.89"The idea of DO TOGETHERwas first 5.83 "No large system isa static entity mentioned (1959) by Mme. Jeanne Poyen in dis- it must be capable of expansion of capacity and cussing the AP 3 compiler for the BULL Gamma 60 alteration of function to meet new and unforeseen computer." (Opler, 1965,p. 307). requirements." (Dennis and Glaser, 1965,p. 5). 5.90"To date, there have been relatively few "Changing objectives, increased demands for attempts made to program problems for parallel use, added functions, improved algorithms and processing. It is not known how efficient, forex- new technologiesallcallforflexible evolution ample, one can makea compiler to handle the of the system, both as a configuration of equipment parallel processing of mathematical problems. Fur- and as a collection of programs." (Dennis and thermore, it is not known how Van Horn, 1965, p. 4). one breaks down problems, such as mathematical differentialequa- "By pooling, the number of componentspro- vided need not be large enough to accommodate tions, such that parts can be processed independently and thenrecombined. These tasksarequite peakrequirementsoccurringconcurrentlyin each computer, but may instead accommodate formidable, but they must be undertakento establish whether the future development lies in thearea a peak in one occurring at the same time as an of parallel processing or not." (Fernbach, 1965, average requirementintheother."(Amdahl, p. 82). 1965, pp. 38-39). 5.91For example, inmachine-aided simula- 5.84"The use of modular configurations of tionsof nonsensesyllablelearning processes, components and the distributed executive princi- Daly etal. comment: "Presuming that, for the ple. ..insures there are multiple components parallellogicmachine, the nonsense syllables of each system resource." (Dennis and Glaser, 1965, were presented on an optical retina in a fixed p. 14). point fixed position set-up, there would bea re- "Computers must be designed which allow the quirement for recognizing (26)3or about 104 dif- incremental addition of modular components, the ferent patterns. If three sequential classification use by many processors of high speed random decisions were performedon the three letters 87 of the nonsense word only 3(26)or 78 different simultaneously (e.g., magneto-hydrodynamic patterns would be involved. problems or pattern recognition). "In the above simple example converting from (3) The computer's structure and belaviorcan, purely parallel logic to partially sequentialprocess- with simple generalizations, be formulated ina ing reduced the machine complexity by two order(s) way that provides a formal basis for theoretical of magnitude. The trend is typical and may involve study of automata with changing structure much larger numbers in a more complicated prob- (cf.the relation between Turing machines lem. Using both parallel and sequential logicas and computable numbers)." (Holland, 1959, design tools the designer is able to trade-off time p. 108). versus size and so has an extra degree of freedom 6 6. in developing his system." (Daly et al., 1962, pp. 5.93 .The development of the Illiac III 23-24). computer, which incorporates a 'pattern articulation unit' (PAU) specifically designed for performing 5.92"...The SOLOMON concept proposed by Slotnick at Westinghouse. Here it is planned local operations in parallel,on pictures or similar that as many as a thousand independent simple arrays." (Pfaltz et al., 1968, p. 354). processors be made to operate in parallel under "One of the modules of the proposed ILLIAC III an instruction from a network sequencer." (Fern- will be designed as a list processor for interpreting bach, 1965, p. 82). the list structure representation of bubble chamber "Both the Solomon and Holland machines belong photographs." (Wigington, 1963, p. 707). to a growing class of so-called 'iterative machines'. 5.94"I use this term ['firmware] to designate These machines are structured with many identical, microprograms resident in the computer's control and often interacting, elements. memory, which specializes the logical design for a "TheSolomonmachineresultedfromthe special purpose, e.g., the emulation of another study of a number of problems whose solution computer. Iproject a tremendous expansion of procedures call for similar operations overmany firmwareobviously at the expense of hardware pieces of data. The Solomon system contains, but also at the expense of software.. .. essentially, a memory unit, an instruction unit, "Once the production of microprogrammedcom- and an array of execution units. Each individual puters was commenced, a further area of hardware- execution unit works on a small part of a large software interaction was opened via microprogram- problem. All of the executioli unitsare identical, ming. For example, more than sothatallcan operate simultaneously under one set of micro- control of the single instruction unit. programs can be supplied with one computer. A "Holland, on the other hand, has proposed second set might provide for execution of the order a fully distributed network of processors. Each set of a different computer perhaps one of the processor has its own local control, local storage, second generation. Additional microprogramsets local processing ability, and local ability to control might take over certain functions of software pathfinding to other processors in the network. systems as simulators, compilers and control pro- Since all processors are capable of independent grams. Provided that the microsteps remain a small operation, the topology leads to the concept of fraction of a main memoryaccess cycle, micro- `programs floating in a sea of hardware'." (Hudson, programming is certain to influence future software 1968, p. 42). design." (Opler, 1966, p, 1759). "TheSOLOMON(SimultaneousOperation "Incompatibilitybetweenlogicand memory Linked Ordinal MOdulator Network), a parallel speeds...has also led to the introduction of network computer,isa new system involving microprogramming, in which instruction execution theinterconnectionsand programming, under is controlled by a read-only memory. The fast the supervision of a central control unit, of many access time of this memory allows full use of the identical processing elements (as few or as many speed capabilities oflered by the fast logic." (Pyke, as a given problem requires), in an arrangement 1967, p.161). thatcan simulatedirectly the problem being "A microprogrammed control sectionutilizes solved." (Slotnick et al., 1962, p. 97). a macroinstruction to address the first word of aseriesof microinstructions contained "Three features of the computerare: inan internal,comparativelyfast,controlmemory. (1) The structure of the computer is a 2-dimen- These microinstructions are then decoded much sional modular (or iterative) networkso that, as normal instructions are inwired-in control if it were constructed, efficient use could be machines..." (Briley, 1965, p. 93). made of the high element density and 'temp- "Themicroprogrammedcontrollerconcept late' techniques now being considered in has been used to implement the IBM 2841 Storage research on microminiature elements. Control Unit, by means of which randomaccess (2) Sub-programs can be spatially organized and storage devices may be connected to a System/360 can act simultaneously, thus facilitating the central processor. Because of its microprogram simulation or direct control of 'highly-parallel' implementation,the2841 canaccommodate systems with many points or parts interacting an unusually wide variety of devices, including 88 CO two kinds of disk storage drive, a data cell drive, differences in performing tasks involving man- and a drum," (McGee and Petersen, 1965, p. 78). computer communications for the general public." "In microprogram control, the functions of the (Sackman, at al., 1968, pp. 9-10). controllerfor source experimental data auto- 5.96"The dynamic nature of multiprogram mation]are vestedina microprogram which on-line computation should have a strong influence is stored in a control memory. The microprogram on memory organization." (Lock, 1965, p. 471). is made up of microinstructions which are fetched "The tradeoffs in speed, cost, logic complexity, in sequence from memory and executed. The and technology are inherent to the design of microinstructions control a very general type of systems and are not separable in spite of the good hardware configuration, so that merely by changing intentionsof the semiconductor manufacturers the microprogram, the functions available in the or the abstract logicians." (Howe, 1965, p. 506). controller can be made to range between wide 5.97"I cannot emphasize toostrongly the limits." (McGee and Petersen, 1965, p. 78). interdependence of hardware and software (the 5.95"The computer science community has statementsofprocedures,implementationof not recognized (let alone faced up to) the problem which in a given equipment configuration constitutes of anticipating and dealing with very large individual the processing capability)." (Schultz, 1967, p. 20).

6. Advanced Hardware Developments

6.1"It will be necessary to scan the photo- while a laser beam can operate in air ant' be the chromic plane very quickly and accurately, with an source of light directly.... extremely fine pinpoint of light. Lockheed Elec- "The major significance of the laser in a display tronics has been exploring a method of rapidly system is that all of the energy is usable since the deflecting a laser beam, nonmechanically, in two apparent source of this light is a diffraction-limited dimensions. The technique is based on refraction point-dipole radiator. Conventional light sources of the beam by acoustic energy." (Reich and such as tungsten filament (. a mercury arc are quite Dorion, 1965, p. 573). wasteful since light is emitted into a 360-degree 6.2"Laser holography is finding some practical solid angle from a relatively large area. When these applications. Technical Operations, lnc., of Burling- light sources are used to illaminate the limited aper- ton, Mass., says it has delivered what is believed to ture of a practical optical sy stem, only a small frac- be the first operational holography equipment to tion of the emitted light is used." (Baker and Rugari, Otis Air Force Base in Massachusetts, where it will 1966, p. 37). be used to photograph fog in three dimensions." 6.7"Since the polarization of the light can be (Electronics 38, No. 20, 25 (1965).) electro-opticallyswitchedinnanoseconds,the 6.3"A work horse of unsuspected power was inherent speed of the electro-optic effect does not harnessed in 1960 when the first operating laser was limit the rate of data projection. However, in prac- demonstrated at Hughes Research Laboratories." tice the rate is limited by dissipation in the deflection ("The Lavish Laser", 1966, p. 15). elements and by the stored-energy requirements of "The first device to be successfully operated was theassociatedcircuitry. Therefore the voltage a pulsed ruby laser." (Baker and Rugari, 1966, p. 37). across the half-wave plate and the loss tangent of the Further reference is to Maiman, 1960. dielectric are important parameters." (Soref and 6.4"Gas lasers are the most monochromatic, McMahon, 1965, p. 59). fluorescent crystal lasers are the most powerful, 6.8"Coherent light from lasers will provide a while semiconductor lasers are the smallest, the revolutionary increase in the volume of communica- most efficient and can be directly modulated." tionthat can be bent over a single pathway." (Gordon, 1965, p. 61). (McMains, 1966, p. 28). 6.5This area of technological development has "In communications the laser can far surpass con- already received such a responsive interest, in ventional facilities. Operating on frequencies many general, that Lowry-Cocroft Abstracts, Evanston, times higher than radio, it can carry many times as Illinois, provides a punched card abstracts service much information. In fact, one laser beam could in the field of laser developments. carry thousands of TV signals at once. Experiments 6.6"The development of the laser as a practical, now under way with lasers enclosed in large pipes continuous, coherent light source has created a new indicate their wide employment for mass com- display technology, that of the laser-beam display. munications for the future." ("The Lavish Laser", This type display can be considered to be analogous 1966, p. 16). to well-known electron-beam type displays, e.g., the "As man goes farther away from the earth in cathode ray tube and the liquid-light valve. The space exploration, laser communication will become primary difference is that the electron beam is more important, because the problems of power constrained to a vacuum environment and requires supply and background noise besetting conventional, a special screen for the emission or control of light microwaves at distances beyond the moon will be 89 minimized. In an example of speed comparison, studies that involve moving unpredictable phe- eight hours were required to transmit the pictures nomena of either uncertain or changing location. from Mars, but a laser beam could carry even televi- Physicalapplications includeterminal and in- sion images across the same distance in a few flightballistics,aerosol-sizedistributions, cloud minutes." ("The Lavish Laser", 1966, p. 16). physics, studies of sprays, and combustion and "The laser with its extremely narrow beam due rocket-exhaust studies, among others." (Stephens, to its short wavelength, notwithstanding its high 1967, p. 26), quantum and background noiseoffers the pos- "Recently at Boulder, Colo., Michael McClintock sibility of surpassing RF techniques in its ability to of the NBS Institute for Basic Standards used an satisfy deep-space requirements." (Brookner et al., argon laser as a source to obtain and analyze the 1967, p. 75). Raman and Rayleigh spectra in several transparent In terms of immediate practicality, however, liquids...His mathematical evaluation of the experiments in the use of laser techniques for data experimental data related scattered light spectra to transmission have been limited to very short dis- viscosity, to molecular rotation and vibration, and to tances. For example, "television signals have been certain molecular concentrations in mixtures of two transmitted on laser beams for distances of the order unassociatedliquids.Analysisofthe Raman of a mile in clear weather..." (Gordon, 1965, p. spectrum also provided, new data on molecular 60), and "the Lincoln Laboratory developed an coupling... optical communications system based on an array "In general, the beam from an argon ion laser was of multiple semiconductor lasers that propagates first passed through a dispersing prism to eliminate pulses through a 1.8-mile path in most weather all but the 4880-angstrom radiation. The light was conditions." (Swanson, 1967, p. 38). then examined from various angles by a spectrom- Goettel therefore concludes that "years of con- eter. Photomultiplier tubes served to increase the tinued research remain beilme an economical laser- intensity of the spectral lines so that they could be transmission system becomes a reality." (Goettel, recorded." ("Laser Applied to Molecular Kinetics 1966, p. 193). Studies", 1968, p. 242). 6.9"Scientists from the Honeywell Research 6.11"A very special hologram, called a spatial Center in Minneapolis say that optical techniques filter, has the capability of comparing two patterns provide a means for increasing information storage and producing a signal which is a function of the cor- density above the levels obtainable with current relation or similarity of the patterns. Experimentally, technology. A memory element under development it has been found that complicated, natural objects will permit over two million bits of information to be with irregular patterns can be recognized with stored on a surface the size of a dime. The informa- greater confidence than can man-made objects tion can be read at the rate of 100 million bits per whichtendtobegeometricallysymmetrical. second using a low power -1 milliwatt laser." Fingerprints, because of their randomness, appear (Bus. Automation 13, No. 12, 69 (Dec. 1967).) to be ideal objects for the spatial filtering method 6.10"The laser will transform Raman spectros- of recognition... copy from a time-consuming tool of limited useful- ness to an important analytical technique; for "The spatial filtering method of fingerprint recog- example, the hour-long exposures of Raman spectra nition has several advantages over other methods on photographic plates are eliminated. Raman spec- of recognition. troscopy with gas-laser beams should have wide- "Recognition is instantaneous, limited only by spread application in analytical chemistry and solid- the mechanical pattern input mechanism. state physics." (Bloembergen, 1967, p. 85). "Partial prints can be recognized. As long as the "The microscopic electrified fluid streams studied information whichisavailable does correlate, occur at high speed, and are virtually impossible to recognition will take place even though one of the record with a conventional optical microscope or two patterns being compared is incomplete. This imaging system. In order to overcome the working- propertyisespecially advantageous when you distance and depth-of-field limitations of the clas- are attempting to correlate partial latent prints sicalmicroscope,atwo-step imaging process with complete recorded prints." (Horvath et al., (holographic photomicroscopy) of high resolution 1967, pp. 485, 488). was developed and applied to the study of the 6.12"A laser image processing scanner (LIPS), electrostatic charging process. In this technique, able for the first time to quantize high resolution one first records the optical interference pattern of photographs for computer manipulation, has been the 'scene', and then uses this record to reconstruct developed by CBS Laboratories...and accepted the original scene. The reconstructed scene can be by the Air Force...LIPS can simultaneously leisurelyexaminedwithconventionaloptical digitizea developed high-resolution photograph systems of limited object volume ... from a negative and produce a much more detailed "The practical consequences of pulsed laser negative from the computer image adjacent to holographic photomicroscopy go beyond the re- the original on the same drum. This makes it much quirements of the present application. Reasonable easier for a photo interpreter to recognize important projection would indicate application to scientific details... 90 "The Air Force systemusesa commercial are not now available." ("The Lavish Laser", helium-neon gas laser to produce black and white 1966, p, 15). images...However,colorphotoscouldbe "The original idea of holography and specifically, producedbysubstitutinganargon,ionlaser spatial filtering dates back to 1886 when Ernst system .. Abbe suggested their existence. However, itre- "Elements of LIPS include a laser lightsource mained for Dennis Gabor to show in 1951 thata focused on a five microns spot sizeon, the negative hologram, which has little recognizable information being digitized inturn feeding into a linearly could be 'reconstructed' to a normal recognizable scanning microdensitometer and a computer buffer image. Various other workers showed his analysis storage. to be correct. Spatial filtering was investigated at "On reconstruction of the higher quality negative about the same time by Marechal and others on the same drum, the same laser is employed in primarily as a means of improving photographic combination with an optical modulator in duplicate images. These pioneers demonstrated that the scanning. ."(Electronic News 14, No. 714, concepts of holography and spatial filtering would 38 (June 23, 1969).) work, but theywere handicapped by the lack of "When the Air Force permitted CBS Labs to a strong source of coherent light. The advent of talk about its high-resolution laser-scanning system the laser in 1960 as a source of essentiallya single (part of Compass Link) used to transmit reconnais- wavelength of light ,excitednew interest in the field sance photos from Vietnam to the Pentagon in of holography. Scientists at General Electric demon- minutes [Electronics, April 14,p. 56], CBS officials stratedthefeasibilityofusingatwo-beam were optimistic about the possibility of broader holographic spatial filteras a means for recogniz- applications. A step in that direction has been ing patterns. A. Vander Lugt, at the University of taken with the modification of the laserscanner Michigan, also investigated methods by whichtwo- so that it can convert high-resolution photos for beam spatial filters could be produced." (Horvath handling by a computer. et al., 1967, p. 485). "Called LIPS-laser image processingscanner "Laser beams will be used to print the catalogs the system digitizes the image, then feeds the and newspapers of the future usinga new technique signal through a buffer to an IBM 360/40 computer. developed by RadioCorporationof America. The computer processes the picture to emphasize Announcement of the development of the technique fine details or improve the contrast. The recon- that can eliminate the need to print in signatures structed image is then read out of the computer was made last month by the company. The method onto photographic film. Thus, LIPS enables the uses the intense light produced by the laser to fuse photo interpreter to manipulate hispicture to powdered ink spreadover the paper to reproduce bring out any desired detail with a high degreethe original. Excess ink is removed byvacuum. of resolution. "The image comes froma photograph of the material to be printed half- tones, line drawingor "Routine work. In operation, the interpreternewspaper pageon a transparency. The image tells the computer what areas he wants empha-is transferred with the aid ofa laser beam to a sized. For example, he could call fora routine hologram or lensless photograph whichserves as that would bring out high-frequency detail. If thea permanent plate. A separate hologram is used finished picture were unsatisfactory he couldgo for each page. to a routine that not only would emphasize high- "Dr. Kenneth H. Fishbeck,a technical advisor frequency detail, but also wouldsuppress or cleanat the David Sarnoff Research Center, Princeton, up large areas of black. N.J., and holder of the patent, said publishers will "LIPS uses a sequentialscan to attain a resolu- be able to eliminate signatures since thenew tion of 100 lines per millimeter. Itcan digitize, or process reproduces pages in sequence, from title record from digital data,a 1.8-centimeter-square to index. He claims publishers could almost print area in 15 minutes; that's at least twice as fast as on demand." (SPIE Glass 5,5 No. 1, 12 (June 1969)). conventional scanners suchas those used on the 6.14"Dillon et al. have proposed and operated Ranger moon probes. a limited-population memory using a ferramagnetic "CBS saystheadvantagesofLIPS highgarnet and driven by a laser beam." (Kump and resolution and geometric fidelity, high-speed read- Chang, 1966, p. 255;see also Dillon et al., 1964). write rates, and operation in standard room lighting 6.15Reimann reportsthat:"The neuristor can be used by map makers, meteorologists, or laser computer, conceived at RCA, isan 'all- news organizations." (Electronics 42, No. 13, 46 optical' computer in which all information and (June 23, 1969)). control signals are in the form of optical energy. .. 6.13"Laserphotographs,calledholograms, "A theoretical study of the neuristorconcept are true three-dimensional representations, and in form of Fiberglas lasers concluded that the the process of holography not only provides a'means fundamentalrequirementsofaneuristorline for lensless microscopy but may make possible could, at least in principle, be met with lasers... microscopic systems at wavelengths where lenses "The main result of the laser neuristor feasibility 91

376-411 0 - 70 - 7 study was the conclusion that lasers are capable film circuits results primarily from the spectral of satisfying all the requirements for digital devices. purity and degree of collimation of the laser light. It was shown that, in addition to the neuristor-type These characteristics allow the beam to be focused logic, lasers in form of resonators and amplifiers to a very fine and intense spot. The high heat can have input-output characteristics that resemble flux which occurs when the light is absorbed by those of conventional logic circuits such as gates the target material, and the sharp definition and or flipflops." (Reimann, 1965, pp. 250-251). localized nature of the working region allow heating, "Fiber-opticelements,with appropriatecon- melting, or vaporizing minute amounts of material, centrations of active emissive ions and passive with minimum effect to adjacent material or com- absorptive ions, are the basic components of tbiz ponents." (Cohen et al., 1968, p. 386). system. The computer is powered by being in a 6.17See Hobbs, 1966, p. 42. continuous light environment that provides a con- 6.18"A new laser data storage/retrieval system stant pump power for maintaining an inverted thatprovidesa 1000-time increaseinpacking population of the emissive ions. Among the poten- density over conventional mag tape, an error rate tially attractive features of such a system a)re the of1 X 108orbetter,permanent (nonerasable) freedom from power-supply connections for indi- storage, a transfer rate of 4 megabits/sec., and vidual circuits, the possibility of transmission of instantaneousread-while-writeverificationhas signals without actual connections between certain been developed by PrecisionInstrument Co., locations, and a promise of high-speed operation." Palo Alto, Calif. (Reimann and Kosonocky, 1965, p. 182). "A working demonstrator of the Unicon' system 6.16"The feasibility of machining resistive and uses a 1-watt argon gas laser, which makes a hole capacitive components directly on thin film metal- in the metallic coating of a mylar-base tape wrapped lized substrates with a laser has been demonstrated. around a drum. The current system, using 5-micron Tantalum films can be shaped into resistor geo- holes, offers a packing.denSity of 13 million bits/sq. metries and trimmed to tolerance by removing in. metal. These films also can be oxidized to value "Readoutisaccomplished by reducing the using the laser beam as the heat source. Resistors laser power; beam reflection or non-reflection in- can be made with tolerances in value of less than dicates nonholes or holes. The tape being used on 0.1 per cent.... the current system offers storage equivalent to "Patterngeneration by laser machining has 10 2400-ft. reels of 800 bpi tape. The system can beer demonstrated on various thin films as well serve on- and off -line, and is capable of recording as on electroplated films. Vaporized lines asfine analog, FM or video data, all of which require as 0.25 mil are readily attainable in thinfilms, high speed." (Datamation 14, No. 4, 17 (Apr. 1968)). as are 0.4 mil lines in plated films. Much narrower 6.19"Byearly1968,PrecisionInstrument lines may be obtained under particularly well- Co. had developed a massive-scale laser recorder/ controlled conditions. Uniform lines as fine as 1 reader storage system, but the first order for the micron have been scribed in thin films on suffi- device was not received until this year. Ed Gray, cientlyflatsubstrates. These films have beenthe chief engineer on the UNICON (Unidensity removed with minimum effect to the substrate Coherent Light Recorder/Reproducer) Laser Mass surface." (Cohen et al., 1968, p. 403). Memory System, said that convincing thefirst "Semiconductor laser digital devices offer an potential customers that they should acquire a improvement in information processing rates of $500K to $1 million memory system was not easy, one to two orders of magnitude over thatexpected especially when you had to 'tell someone that you fromhigh-speedintegratedtransistorcircuits. were not going to store data withmagnetics like Data processing rates of 10 to 100 gigabits perGod intended.' Now that the first order has been second may be possible using semiconductor lasers. placed, by Pan American Petroleum Corp. of However, the technology for fabricating low-power Tulsa, Oklahoma, Mr. Gray feels that the systems laser circuits is still undeveloped and low-tempera- will move a little faster in the marketplace. ture operation may be required." (Kosonocky and "The $740K system placed with Pan American Comely, 1968, p. 1). is to be installed with all requisite software about "Laser digital devices may be used for general- March of 1970. Four other potential customers, purpose logic circuits in very much the same way including some government agencies and a private that transistors are now used, except that all of credit-reporting firm, are also expected to place the processing is done with optical rather than orders." (Datamation 15, No. 3, 116 (Mar. 1969).) electric signals." (Reimann and Kosonocky, 1965, 6.20"TheNationalArchivesandRecords p. 183). Service has begun a cost-effectiveness study of "Semiconductor current-injection lasers are most archival storage systems in an effort to shrink its attractive for digital devices because of their small mag tape library, which contains one million plus size, high pumping efficiency, and high speed of reels. The study, due for completion next month, operation." (Reimann and Kosonocky, 1965, p. 194). is using the capabilities of Precision Instruments' "The utility of a laser as a tool for fabricating thin Unicon device as a model. The Unicon employs a 92 laser-etched aluminum strip with a 30-year shelf "Such a system could be used with a matrix life." (Datamation 14, No. 10, 171 (Oct. 1968)). containing alphabetical or other symbols. The 6.21"Honeywellscientistsareinvestigating laser would be used as a print-out device, pro- a method that uses a laser for mass storage and jectingthevarioussymbolsonto a recording retrievalof informationincomputer memory. medium. Although emphasizing that development isstill "The Air Force Systems Command at Wright- in the research stage and may be several yearsPatterson AFB is interested in IBM's work on a away from practical application, the researchersvariable frequency laser which might be used in believe the discovery is a possible key to inex- conjunctionwithacolor-sensitivecomputer. pensive mass storage of data for the enormous This type of setup is said to have a potential capacity computer networks envisioned for the 1970's." of a hundred million bits per square inch of photo- (Commun. ACM 11, 66 (Jan. 1968)). graphic material." (Serchuk, 1967, p. 34). 6.22"The system...uses a modulated laser 6.25"Instead of recording a bit as a hole in beam to inscribe data onto photosensitive discs... a card, it is recorded on the file as a grating pattern Each disc contains 3, 100 tracks with a capacity of a certain spacing...A number of different of 67,207 bits per track, including error corrections grating patterns with different spacings can be bits. The storage unit holds 2, 600 discs, storedsuperposed and when light passes through, each on edge, in four [or eight] trays ... twoauxiliary grating bends the light its characteristic amount, disc banks can be added to achieve the maximum with the result that the pattern decodes itself... memory capacity . ..of 150 billion characters. The new system allows for larger areas on the film The reader reaches any piece of information onto be used and lessens dust sensitivity and the the3,100 tracks(per disc) within 15milli- possibility of dirt and scratch hazards." (Commun. seconds..." (Business Automation 12, No. 6, ACM 9, No. 6, 467 (June 1966).) 84 (1965)). 6.26"RecentlyLonguet-Higginsmodeleda A CW helium-neon laser is used to "achievetemporal analogue of the property of holograms real-time writing of information on the system'sthat allows a complete image to be constructed photosensitive memory discs." (Connolly, 1965,from only a portion of the hologram. In the present p. 4). paper a more general analogue is discussed and 6.23"A method for producing erasable holo- twotwo-steptransformationsthatimitatethe grams may enable an optical memory to store 100recording-reconstruction sequence in holography million bits in a film one inch square. are presented. The first transformation models "The memory could be read out, erased and the recall of an entire sequence from a fragment reusedrepeatedly,accordingtoDr.William while the second is more like human memory in Webster, vice president in charge of RCA Labora- that it provides recall of only the part of the se- tories. quence that follows the keying fragment." (Gabor, "Information can be written into the magnetic 1969, abstract, p. 156). film in 10 billionths of a second, and erased in 6.27"A new recording mechanism...con- 20 millionths of a second. Laser light split intosists of the switching of magnetization under the two beams, one going directly to the film and theinfluence of a stress resulting from a heat gradient other going to the information bit pattern, interferesintroduced by a very narrow light or electron beam. constructively to produce heat and consequentlyThe mechanism is assumed to be magnetostriction a realignment of atoms. with a rotation of the anisotrophy. The model pre- "Where the two beams interfere destructively, sented and the criteria for recording are supported, nothing happens." (Data Proc. Mag. p. 21 (Sept.at least in part, by experimental observations." 1969)). (Kump and Chang, 1966, p. 259). 6.24In the IBM-laser system developed for 6.28"In attempts to provide computers with Army Electronics Command andinstalledat previously unavailable amounts of archival (read- Fort Monmouth, it is noted that: "Through employ-only) storage, various techniques involving optical ment of a deflection technique, the shaft of light and film technology have been employed to utilize can be focused on 131,072 distinct points within the high information capacity of film (approximately a spa smaller than a match head....To 106 bits/in.2) and the high resolution and precision provide a readout in printed form, the ilser beam oflasers and electron beams. The trillion-bit can scan through a mask inscribed with the alpha-IBM 1350 storagedevice, an offshoot ofa the bet and other symbols andthrough the action `Cypress' system,... uses35 mm X 70 mm silver of light-bending (deflection) crystalsturn out the halide film 'chips.' final product on photo-sensitive paper." (Commun. "A total of 4.5 million bits are prerecorded on ACM 9, 467 (1966)). each chip by an electron beam. For readout, a "At International Business Machines Corp.... plastic cell containing 32 film chips is transported one method, devised for the Army Electronics to a selector, which picks the proper chip from Command, Fort Monmouth, N.J., makes use ofamong the 32; average access time to any of the a high-speed switching arrangement with electron- 1012 bits is 6 seconds. After a chip is positioned, ically controlled crystals. information is read using a flying-spot CRT scanner. 93 Two IBM 1350 units are scheduled for mid-1967 model, that one UNICON device with 35 mm film delivery to the Atomic Energy Commission at could store a trillion bits on 528 feet of film, with Livermore and at Berke ly for use with bubble an average access time to a record of 13 seconds." chamber data. Other techniques of reading and (Van Dam and Michener, 1967, p, 205). (See also writing withelectron beams are explained by note 6.14), Herbert." (Van Dam and Michener, 1967, p. 205). 6.31"Considerable experimentation in modula- 6.29"Results of basic theoretical studies con- tion and transmission is needed before optical ducted at the NCR research laboratories have communication by laser can be said to be really indicated that CW lasers of relatively low power useful except in very specialized cases." (Bloom, should be capable of permitting very high resolu- 1966, p. 1274). tion real-time thermal recording on a variety of 6.32"Atfirstsighta laser communication materials in the form of thin films on suitable system with its extremely wide information carrying substrates.Subsequent laboratory studies have capacity would appear to be a natural choice for shown that such thermal recordingisindeed an interplanetary communication system. How- possible. This recording technique has been termed ever, among other things, the acquisition and heat-mode recording." (Carlson and Ives, 1968, tracking problems are considered to be so severe p. 1). that such a system is not thought to be realistic at "The recording medium is coated on a 5- by the present time. This may be indicative of an in- 7-inch glass plate, a quarter of an inch thick. The formation technology utilization gap." \Asendorf, plate carrier mechanism is capable of stepping in 1968, p. 224). thehorizontal and verticaldirectionsto form 6.33"Ingeneral,earthboundlaser-ranging matrices of up to 5,000 images at an overall reduc- systems are limited by local atmospheric conditions. tion of 150 to 1." (Carlson and Ives, 1968, p. 5). A typical value of range routinely measured is "The results of the studies described in this 20 km or less." (Vollmer, 1967, p. 68). paper have established laser heat-mode recording "Earthbound applications of coherent optical asa veryhighresolutionreal-timerecording radiation for communications appear to be severely process capable of using a wide 'variety of thin limited for two reasons. The first, and most signifi- film recording media. The best results were ob- cant, is the effect of atmospheric turbulence on the tained with images which are compatible with coherence of the radiation. The second is the effect microscope-type optics. The signals are in elec- of smallvibrations on the coherent detection tronic form prior to recording and can receive efficiency and signal-to-noise ratio. This can be extensive processing before the recording process minimized by careful design, but the first factor is occurs. In fact, the recordings can b e completely beyond the designer's control. Although coherent generated from electronicinput. For example, opticaldetectionhas been demonstrated over Figure 6 shows a section of a heat-mode microimage some useful paths, the vulnerability of the link to with electronically generated characters, produced atmospheric variations makes practical applica- by the Engineering Department in our Division. tion somewhat doubtful." (Cooper, 1966, p. 88). The overall image is compatible with the 150-to-1 6.34"Is the enormous increase in bandwidth PCMI system (less than 3 mm field), and consists offered by light as a carrier frequency in communica- of 73 lines of characters, 128 characters per line. tions needed? For transmission in space the acquisi- Although this image was recorded in 1.6 seconds, tion and aiming of the light beams pose formidable faster recordings are anticipated. A description problems. In the atmosphere, rain, smog, fog, of this work will be published in the near future." haze, snow, etc., make light a poor competitor of (Carlson and Ives, 1968, p. 7). microwaves. Can a system of enclosed tubes with 6.30"Another scheme for storing digital in- controlled atmosphere and light repeater stations formation optically is the UNICON system, under be built on a technologically sound and econom- development at Precision Instrument Company. ically feasible basis?" (Bloembergen, 1967, p. 86). This system uses a laser to write 0.7-micron- 6.35"InformationAcquisition,Sensing, and diameter holes in the pigment of a film. Information Input", Sect. 3.1.1. Some additional references is organized in records of at most a million bits; are as follows: each record is in a 4-micron track extending about a "Since the first laser was demonstrated in 1960, meter along the film. Individual tracks are slanted considerable interest has developed in its possi- slightly so that they extend diagonally across the bilities for use in communication systems. The film. (The amount of slant and the width of the basic sources of this interest are the coherent film determine the length of the records.) Each nature of the radiation obtained as compared with recordisidentified by information stored next all previously known extended sources of optical to the beginning of that record, in an additionalradiation, and the laser's short wavelength. This track at the edge of the film. Readout of a particular latter characteristic provides the potential ability record involves scanning the identifier track for to achieve bandwidths, or information capacities, the proper code and then scanning the track with a that are orders of magnitude greater than anything laser weaker than that used for writing. It is pre- obtained heretofore. A more realistic advantage, dicted, on the basis of an experimental working in terms of presently available information sources, 94

wyasaluitasmArin.citt 74.7.4",14,7, 4

results from the combination of high coherence and similarreceiving and recordingstationthere short wavelength. It is the ability to generate a reconstructs the photographs to their original form highlycollimated beam (limited by diffraction for immediate inspection." (Spie Glass 5, No. 2, phenomena), which leads to the ability to achieve 9 (Aug. 1969).) communications over great distances. Of equal "According to Air Force officials, picturespro- importance is the fact that with a coherent signal, duced by the CBS Laboratories Image Scanning coherent detectionof the information can be and Recording System contain the highest resolu- obtained with greatly improved immunity to natural tion ever reported in the transmission of aerial incoherent noise sources such as the sun." (Cooper, reconnaissance photographs. High-altitude photo- 1966, p. 83). graphs processed by the system show such detailed "The firstenthusiastic suggestions that laser information as identification numbers of ships in technology potentially provides many orders of port, planes on runways and troop movements... magnitude more communication capability than "In operation, the lightweight system uses a pre- RF technology, and that it might, therefore, offer cisely controlled laser beam to scan rapidly across the only solution to the problem of general wideband photographic film. The laser converts each picture communications with deep-space probes, needs frame to an electronic video signal. The signal is to be more carefully assessed." (Dimeff etal., then fed to a transmitting device for satellite relay, 1967, p. 104). saidJohnManniello, CBS LaboratoriesVice "For deep-space, wide-band communication... President for 'Government Operations, who con- another factor may be...importantnamely, the ceived the system application. Once the signals size of the transmitting aperture. A very large contact the satellite, they are flashed to a receiving aperture, as would undoubtedly be required by a station in Washington within seconds, he added. microwave channel, is likely to prove an obstruction "The receiving station which has related photo- to the sensors of the aircraft and will, therefore, scanning, recording and developing equipment reduce the time available for collecting information reconstructs the video signal to the original film or transmitting it. In this respect, the laser has an image andproduceshigh-qualityphotographic important advantage over microwave." (Brookner prints. et al., 1967, p. 75). "Because of the laser-scanning technique in- "Several optical links which use GaAs injection volved, no photographic resolution is lost between lasers as transmitters have been constructed. One recordingandtransmissionfromtheoriginal of them has been demonstrated to be capablefilm taken in Vietnam." (Spie Glass 5, No. 2, 9 of transmitting 24 voice channels over 13 km." (Aug. 1969).) (Nathan, 1966, p. 1287). 6.38"Superficially, it appears attractive to have 6.36"If we classify our communication require- fast switching, high storage density, direct visual ments on the basis of range, we find that lasers can display. Such developments would depend heavily be helpful at the range extremities that is, for on the availability of cheap, small, high-quality distances less than about 15 km and for those semiconductor lasers. If these were available, the greater than 80 million km." (Vollmer, 1967, p. 66). entire organization of computers using them would 6.37"Anelectronicsystemthattransmits probably be different." (Bloembergen, 1967, p. 86). military reconnaissance pictures from Saigon to 6.39"The power and efficiency available from Washington in minutes viasatellite may soon lasersat the desired wave lengths (particularly enable news media to dispatch extremely high- ultraviolet) must be improved, and adequate laser quality photographs and type around the world deflection techniques must be developed before for instant reproduction. laser displays will be feasible for widespread use." "Potential benefits are also foreseen for medicine, (Hobbs, 1966, p. 1882). earth resources surveys and industry. "Since lasers don't require vacuums, there is "The high-performance system was developed asignificantconveniencerelativetoelectron for the U.S. Air Force Electronics Systems Divisions beams. But there is a severe penalty compared to by CBS Laboratories, adivisionof Columbiaelectron beams due to problems indeflecting, Broadcasting System, Inc.It combines electro- modulating, and focusing." (Gross, 1967, pp. 7-8). optical and photographic techniques to relay high- "Lasers offer great promise for future imple- resolution aerial photographs of ground activitymentation of display systemsparticularly large- in Vietnam to the President and Pentagon officials. screen displays. The ability of a laser to deliver Pictures seen by the President are many times highly concentrated light energy in a coherent beam sharper than the best pictures shown on home of very small spot size is well known. Several dif- television sets. ferent approaches to laser displays are being in- "Within minutes after photographs have been vestigated. Since they all require some means for taken in Vietnam, they are readout by the system's deflecting and modulating the laser beam, con- electronicscanningdeviceandconvertedto siderable development efforts are being expended video signals. The signals are then fed to a com- ondeflectiontechniques.Digitaldeflectionof munication link, which relays them over the U.S. lasers by crystals has been satisfactorily demon- DefenseSatelliteNetworktoWashington. A strated for 256 positions in each direction, but at 95 least 1024 positions in each direction are needed and under a de-activation condition and/or at a for a practical large-screen display system." (Hobbs, later time." ("Investigation of Inorganic Photo- 1966, p. 1881). tropic Materials...", 1962, p. 1). "The laser is an efficient light source, and its "The property of certain dyes and other chemical output can be focussed to small sizes and high compounds to exhibit a reversible change in their power densities. There is confidence that laboratory absorption spectrum upon irradiation with specific means for modulating lasers and deflecting their wavelengths of light has been termed phototropism, beams will be found practical." (Bonn, 1966, p. or photochromism. The emphasis in this definition 1869). is on reversibility, because, upon removal of the "More rapid progress would be made in utilizing activating radiation the systems must revert to laser recording if better means of deflecting laser their original states to be considered pho;.ochromic." beams at the desirable speeds and resolutions (Reich and Dorion, 1965, p. 567). existed or were clearly foreseeable." (Smith, 1966, "By definition, photochromic compounds exhibit p. 1297). reversible spectral absorption effects color "An experimental device that can switch the changes, resulting from exposure to radiant energy position of a light beam more than a thousandin the visible, or near visible, portions of the times faster than the blink of an eye could become spectrum. For example, one class of photochromic an important part of computer memories of the materials consists of light-sensitive organic dyes. future. The device, a digital light deflector, was NCR photochromic coatings consist of a molecular developedatthe IBM SystemsDevelopment dispersion of these dyes in a suitable coating Div. laboratory in San Jose, Calif. material.Photochromic coatings are similar to "The experimental deflector changes the location photographic emulsions in appearance and with of a beam in 35 millionths of a second by a unique respect to certain other properties. Coatings can method of moving a glass plate in and out of contact be made to retain two-dimensional patternsor with a prism. images which are optically transferred to their "High-speed deflectors of this type are poten- surface." (Hanlon et al., 1965, p. 7). tially useful in future optical memories to randomly "Photochromic film, a reusable UV sensitive position a laser beam for data recording and reading. recording media hasprogressedtothepoint Such beam addressable memories are expectedwhere prototype equipment is being designed." to be many times faster than present magnetic (Kesselman, 1967, p. 167). storage methods because of the relative speed 6.43"Photochromic coatings exhibit excellent of relocating a light beam in comparison to moving resolution capabilities. In addition, both positive- a bulky recording head." (Computers & Automation to-negative and direct-positive transfers are pos- 18, No. 5, 68 (May 1969).) sible...The coatings are completely grain-free, 6.40"Another attractive approach is the use of have low gamma (excellent gray scale characteris- a laser beam to write directly on a large luminescent tics), and exhibit inherently high resolution... screen. This is somewhat equivalent to an 'outdoor' Further, because the coatings are reversible, the cathode-ray tube in which the laser beam replaces information stored can be optically erased and the electron beam and the luminescent screen rewritten repeatedly." (Hanlon et al., 1965, p. 7). replaces the phosphor face plate of the tube. "Photochromism may be defined as a change in It offers advantages over a CRT in that a vacuum color of a material with radiation (usually near is not required and a large-screen image can be ultraviolet)and the subsequent returntothe generated directly. One feasibility system has been original color after storage in the dark. Reversible developed using a 50 milliwatt neon-helium gas photochromism is a special case of this phenomenon laser, a KDP crystal modulator, a piezoelectric in which a material can be reversibly switched by crystal driven horizontal deflecting mirror, and a radiation between two colored states. Photochromic galvanometer driven vertical deflecting mirror to compounds may be valuable for protection from provide a television rastor scan image projected radiation; reversibly photochromic materials are onto a 40 inch screen. Brightness of 50 foot-lam- potentially valuable for data storage and display berts, contrast ratio of 100 to 1 (dark environment), applications."(Abstract,talkonphotochromic resolution of 1,000 to 2,000 lines, and update time materials for data storage and display, by U. L. of 33 milliseconds are anticipated for direct view Hart and R. V. Andes, UNIVAC Defense Systems laser systems." (Hobbs, 1966, p. 1882). Division, at an ONR Data Processing Seminar, 6.41"Electron-beam devices, including those May 4, 1966, See also Hart, 1966). which use photographic emulsions and thermo- 6.44"Most of the systems so far reported are plastic films, operate in a vacuum, which is a only partially, or with difficulty, reversible, or are nuisance." (Bonn, 1966, p. 1869). subject to fatiguea change in behavior either 6.42"By definition, phototropy is the photo- with use or with time in storage." (Smith, 1966, chemical phenomenon of changing some physical p. 40). property (color) on exposure to electromagnetic "Organic photochromic materials fatigue with radiation (light) and returning to its original (color. use." (Bonn, 1966, p. 1869). less) state after removal of the activating source 6.45"Photochromic films permit the storage 96 ,".tORIPAIrrrrr,

of images containing a wide contrast of gray reader...which was designed specifically for scale because they are inherently low gamma and possible use aboard a manned space vehicle. The grain-free." (Tauber and Myers, 1962, p.409). reader would have a self-contained, fixed reference 6.46"It is recorded that Alexander the Great file of up to 50,000 pages of information, such as discovered a substance, whose composition has navigationalcharts,planetary and space data, been lost in the obscurity of antiquity, that would andcheckout,maintenance,andemergency darken when sunlight shone upon it. He dipped a procedures." (Hanlon et al., 1965, p. 13). narrow strip torn from the edge of his tunic into "Current design emphasis by 'NCR has been a solution of the material and wore this strip toward the development of low cost PCMI readers wrapped about his left wrist. Many of his soldiers forcommercialapplications."(Hanlonetal., did the same. By observing the changes of color 1965, p. 19). during the day, they could tell the approximate 6.48"The information glut threatening to swamp hour.This became known as Alexander's rag the engineer and the scientist is being eased by a time-band. (I am sorry that I cannot identify, and British organization called Technical Information hence cannot give proper credit to the author of on Microfilm. The medium of the 'message' is the this delightful footnote to history.)" (Smith, 1966, National Cash Register Company's PCMI process. p. 39). This makes possible the storage of over 3000 printed 6.47"1. Photochromic films provide very high pages of information on a single 4-by-6-inch trans- resolution with no grain. parency. The system used by TIM enables the "2. Photochromic films permit the storage engineer to locate the data he wants in a matter of images containing a wide contrast of seconds. He simply selects the proper trans- of gray scale because they are in- parency and immediately locates the appropriate herently low gamma and grain-free. page images with an NCR reader which displays "3.Photochromic films provide immediate the selected pages on an illuminated viewing screen. visibility of the image upon exposure. TIM points out that one of the most valuable sources No development process is required. of information to engineers and scientists is manu- "4. Photochromic films provideboth facturers' literature. The problem has been that erasing and rewritingfunctions. thisis produced in an extraordinary variety of This permits the powerful processes forms. These are difficult to catalogue compre- hensively and they also create an enormous bulk. editing,updating,inspection,and The NCR-developed PCMI technology involves error correction to be incorporated a photochromic coating which produces an image into systems. that is virtually grain-free. The process permits "5.The PCMI process incorporates the a microscopic-size reduction which is not practical ability to effect a bulk-transfer read-out with conventional microfilm processes. NCR is of micro-images at the 200:1 reduction producing the transparencies for TIM in its Dayton, level by contact printing. Ohio processing center from 35-millimeter micro- "6. Use of high-resolutionsilver halide film supplied by the British firm. All data is updated films provides both permanency for every six months." (bema News Bull., Dec. 9, 1968, the storage of micro-images and eco- p. 8). nomical dissemination of duplicates. 6.49"Informationstoredon photochromic "7.The very high density of 200:1 micro- coatings is semipermanent...This is a result images offers the possibility of using of the reversible nature of the photochromic coating. same form of manual retrieval tech- The life of the photochromic micro-image is de- niquesfor many applications. This pendent upon the ambient temperature, of the coat- eliminates the normal requirement in ing. At room temperature, image life is measured systems of this size for expensive and in hours, but as the temperature is lowered, life complex random access hardware." can be extended very rapidly to months, and even (Tauber and Myers, 1962, p. 266). years." (Hanlon et al., 1965, p. 8). In the photochromic micro-image (PCMI) micro- 6.50"Thetemperature-dependentdecayof form process developed by the National Cash image life obviously prohibits the use of photo- Register Company therehave been achieved chromic micro-images in their original form for "linear reductions from 100-to-1 to greater than archival storage. To overcome this problem, means 200-to-1, representing area reductions from 10,000 - have been developed for contact-printing the photo- to-1 to greater than 40,000-to-1, [which] have been chromic micro-images to a high-resolution successfully demonstrated by using a variety of photographic emulsion, thereby producing image formats, such as printed materials, photo- permanent micro-images." (Hanlon et al.,1965, graphs, drawings, and even fingerprints." (Hanlon pp. 8-9). et al., 1965, p. 1). "The entire contents of the photochromic micro- "NCR has developed a number of research image plate are then transferred (as micro-images) prototype readersfor viewing PCMI transpar- in one step, by contact-printing onto a high-resolu- ences ...[including] a miniaturized microimage tion silver halide plate...Micro-image dissemi- 97 nation (duplicate) films are prepared ina similar by 6 inchescan now be used to store a pattern manner, using the silver masters to contact-print of bits instead of images of pictorialor alphabetical onto high-resolution silver halide film." (Hanlon information.Photochromichigh-resolutionfilms et al., 1965, p. 9). coupled withproper light sources and optical 6.51Further, "a more realistic assessment... systems can provide the storage of millions of bits so that spillover, halation, and registration restric- to the square inch. A micro-holographic indexing tions would not be impossiblysevere, still results system used with such storage devicesmay revolu- in a contiguous bit density of 10Vcm2." (Reich and tionise data storage and retrieval." ("R andD for Dorion, 1965, p. 572). Tomorrow's Computers," 1969,p. 53). 6.52"Transparentsilicateglasscontaining 6.56"The breadth of the sensitivity charac- silver halide particles darkens when exposedto teristics of the photochromic films in conjunction visible light, and is restoredto its original trans- with the width of the spectral characteristics ofthe parency when the light source is removed. These available phosphors presenta potential systems glasses have been suggested for_self-erasingmemory designer witha choice of a number of component displays, readout displays for air traffic controls, parts...Future improvementsin CRT-photo- and optical transmission systems.... chromic film display systemsare dependent upon "Photochromic glass appears to be uniqueamong the capabilities of each of thecomponents. The other similar materials because of itsnon-fatiguing basic parameters which enter into the cathode characteristics. No significant changes in photo- ray tube efficiency are the fiber optic plate and chromic behaviorhave resulted fromcycling phosphor. An increase in the fiber opticefficiency samples with an artificial 3600A black lightsource is doubtful except through theuse of higher numeri- up to 30,000 cycles. There were also no apparent calaperturefibers.Increasingthenumerical solarization effects causing changes in darkening aperture has the disadvantage of requiringa higher or fading ratesafter accelerated UVexposure degree of controlon the film-CRT gap. An improve- equivalent to 20,000 hours of noon-day sunshine." ment in basic phosphor efficiency is difficult (Justice and Leibold, 1965, to p. 28). foresee although several militaryagenciesare "Another potentially important applicationof now or will be sponsoring programs to achieve photochromic materials is in the display ofinfor- this goal. mation. Data can be recorded in photochromic "An advance of the state-of-the-art ofphosphor glass in two ways: by darkening the glass withshort- technology should be possible bya factor of 4, wavelength light in the desiredpattern; or by uni- but probably not beyond. By carefuloptimization formly darkening the glass and bleachingit, in the of the phosphor deposition withrespect to particular desired pattern, with longer wavelengthlight." applications some improvement ispossible. At (Smith, 1966, p. 45). the same timean increase in the efficiency of the 6.53"To produce a display,patterns of varying photochromic film bya factor of 2 is theoretically optical densityare written on photochromic film possible. Of more importanceto the system de- with a deflected ultraviolet light beam. Thefilm signer is the understanding and optimizationof so exposed forms the 'object' in a projection system. writing and rewriting ratesas they affect phosphor Visible light is projected through the filmonto a efficiency and life and in the matching of the CRT screen. This display has mechanical simplicity, with the photochromic film." (Dorionet al., 1966, controllable persistence, anda brightness com- p. 58). parable to conventional film-projectiondisplays." 6.57"Wavefront reconstructionwas invented (Soref and McMahon, 1965,p. 62). 6.54 by Gabor and expounded by him ina series of "For dynamic applications suchas target classic papers [1948-1951] ." (Armstrong,1965, tracking, this techniquenot only permits a real- p. 171). time target track, but also providestarget track history in the form of "The wavefront reconstruction method ofimage a trace with 'intensity' formation was first announced by Gabor in1948." decreasing with time. The time periodcovered by 6.58Stroke gives a derivation of theterm: the visible target track historyis a function of"Hence, the name 'hologram' from theGreek the photochromic material. At thepresent time, roots for 'whole' and `writing'." (Stroke, 1965,p. 54). the speed of photochromic materialslimits the And also defines it: "A hologram istherefore an character generation rateto less than 100 characters interference pattern between persecond. a reference wave Successful development of faster and the waves scattered by the objectbeing re- photochromic materials will providean attractive corded." (Stroke, 1965,p. 53). electro-optical dynamic large-screen displaywith 6.59See also the following: no mechanically moving parts." (Hobbs, 1966, "Arbitrary objects...areilluminatedby p. 1879). parallel laser light. In the generalcase, the light 6.55"One of the technological trends which reflected by these objects will be diffuseand the will give us mass memoriesat a viable price is reflected wavefronts will proceedto interfere in photochromic microimagery. Photochromictech- the photosensitive medium wherethe interference niquesby which asmany as 2,000,000 words pattern can be recorded. After the photosensitive can be stored on a film transparency only 4 inches medium has been exposed and processeditis 98 called the hologram, whichmay be defined as the spatial frequency content recorded interference of or bandwidth of the two or more coherent device isavailable for the single image.Com- wavefronts. When the hologram is illuminatedby putationally, kinoform constructionis faster than one of the original wavefronts used to form it, the hologram construction because remaining reference beam wavefrontsarereconstructed. and image separation calculationsare unnecessary, Observation of these reconstructed wavefrontsis "A kinoform operates onlyon the phase of an nearly equivalent to observing theobjects from incident wave, being basedon the assumption that which they were originally derived." (Collier,1966, only the phase information p. 67). in a scatteredwave- front is required for the constructionof an image "An optical hologram isa two-dimensional photo- of the scattering object. The graphicplate, amplitude of the which preserves information about wavefront in the kinoform plane is assumedcon- the wavefront of coherent light whichis diffracted stant. The kinoform may therefore be from an object and is incidentupon the plate. A thought of properly illuminated as a complex lens which transforms the known hologramyieldsathree- wavefront incident on it into the wavefrontneeded dimensional wavefront identicalto that from the to form the desired image. Although it original object, and thus the observed was first image is conceived as an optical focusingelement, the an exact reconstruction of the object. The observed kinoform can be used image has all of the usual optical as a focusing element for properties as- any physical waveform, e.g., ultrasoundor micro- sociatedwithrealthree-dimensionalobjects; waves." (Lesem et al., 1969, e.g., p. 150). parallax and perspective." (Lesemetal., 6.63 "A new hologram madeat Bell Telephone 1967, p. 41). Laboratories now allows the viewerto see a 3D 6.60"HolograjAy is the science ofproducing image rotate through images by wavefront reconstruction. In a full 360 degrees as he moves general his head from sideto side... no lenses are involved. The reconstructed image To make a flat hologram with a 360-degree view, verticalstrips may be either magnified or demagnified compared of the photographic plateare exposed sequentially to the object. Three-dimensional objectscan be from left to rightacross the plate. A narrow slit reconstructed as three-dimensionalimages." (Armstrong, 1965, in a mask in front of the plate allowsonly one p. 171). strip to be exposed ata time, each strip becoming 6.61"Are HologramsAlreadyOutdated? a complete hologram of one view of the object." Holography is one of themost exciting developments (Data Proc. Mag. 10, No. 4, 16 (Apr. 1968)). of today's technology. Holograms make use of a 6.64"Holographyprovidesan high-energy laser beam tostore or display three- alternative dimensional images for such applications description of pictures, which mightbe more as read- amenable to bandwidth compression. Toinvestigate only storage; packing densities anddevice speeds are this possibility, it is desirableto measure various extremely impressive. However,at today's statistics of the hologram, andto try various opera- pace of innovation, holographymay be outmoded tions on it to see what their effects would before it approaches being practical.One of the be on the reconstructed pictures....Holography and latest competitors for 3-D display,storage, and wave other coherent optical processing...techniques conversion applications is the kinoform,a new wave- have made possible relatively simple front reconstruction device which also ways of obtain- projects a ing the Fourier transforms oftwo-dimensional 3-D image, but requires one-fourth ofthe computer functions and operatingon them in the frequency time to generate and creates images roughlythree times as bright. domain." (QuarterlyProgress Report No.81, ResearchLaboratoryforElectronics,M.I.T., "A computer program is usedto produce a coded 199 (1966)). description of light being scattered froma particular object. The resultant computationsare used to 6.65"Gabor and others have proposedthe use produce a 32-grey-level plot which isphotoreduced of the wavefront reconstructionmethod to produce and bleached. Then, when subjectedto even a a highly magnified image, using eithera change in very small light source, such as the girl's earring wavelength between recording ofthe hologram and in the photo above, the 3-D imageis formed. A its reconstruction,or by using diverging light for kinoform image can be produced ofany object which one or both steps of the process. The two-beam can be computer-described. Examples might in- processis readily amenableto such magnifica- clude proposedbuildings,autodesigns,relief tion..." (Leith et al., 1965,p. 155). maps, or two-dimensional alphanumeric data." 6.66"Color reconstructions should beattainable (Datamation 15, No 5, 131 (May 1969)). from black and white holograms ifsuitable temporal 6.62"The kinoform isa new, computer-gener- coherence conditionsare ensured." (Stroke, 1965, ated, wavefront reconstruction devicewhich, like p. 60). the hologram, provides the display of a three- "Holography is another field forwhich the laser dimensionalimage. Incontrast, however, the has opened illuminated kinoform yields many possibglies, Perhaps itwill a single diffraction find useful applications inpattern recognition and order and, ideally, all theincident light is used in storage of three-dimensional to reconstruct this one image. Similarly, information as a all the Fourier transform.... 99

1Mmft.eLou1a..1Lehndki19.sollg..aiNAIILI "Three-dimensionaldisplaysofairfieldap- graphs. In the NCR system, a hologram containing proaches in the cockpit of a jet liner with the correct signatures and numbers randomly located is placed viewing angle from the position of the aircraft in the optical path of a laser. would be a more interesting application [of laser- "If matching occurs when a signature card is holographicrecordings]."(Bloembergen,1967, inserted into a receiving device, the system locates p. 86), the picture [which] is projected for comparison." "We read about images having three-dimensional (Se.4huk, 1967, p. 34). properties, magnification obtained by reconstruct- 6.67"Thewavefrontreconstruction method ing with a wavelength greater than that used in offers the possibility of extending the highly de- forming the hologram, diffuse holograms which, veloped imagery methods of visible-light optics even when broken, produce whole images, multi- to regions of the electromagnetic spectrum where color imagesobtaining from emulsions which high-quality imagery has not yet been achieved..." normally produce only black and whites." (Collier, (Leith et al., 1965, p. 157). 1966, p. 67). 6.68"A Megabit digital memory using an array "Therecordingofsurfacedeformationsin of holograms has been investigated by Bell Labo- engineering components demonstrated here shows ratory scientists. The memory is semipermanent, how these techniques may be applied at low cost with information being stored in the form of an and in a short time. For teaching purposes it has array of holograms, each hologram containing a been shown that interference holography of the pageof information. A pageisread...by distortions of a rajor blade can be demonstrated deflecting a laser beam to the desired element of adequately to a large group of people in only a the array, so as to obtain reconstruction of the few minutes." (Bennett and Gates, 1969, p. 1235). image stored in the element the digital informa- "With practical applications for holograms still tionon a read-out plane which is common to all in the few-and-far-between stage, the Office of elements of the array. Photosensitive semicon- Naval Research and IBM believe they have a holo- ductors arrayed on the read-out plane then sense graphicapplicationthatisboth practical and the stored information... unique: in a head-up, all-weather landing system. "In the Bell Labs experimental system, the light "The systemnow atthelaboratory model sourceisa continuous-wave helium-neon laser stageemploys a hologram of an aircraft carrier. operating in the lowest order transverse mode. The hologram is picked up by an infrared vidicon Two-dimensional deflectionisaccomplished by and projected on a crt cockpit display... cascaded water-cell deflectors, using Bragg dif- "The achievement is one of application in which fractionfromultrasonicwaves inwater, and atwo-dimensionalrepresentationwiththeso- capable of deflecting the beam to any of 300 called six degrees of freedom encountered in a addresses in less than 15 psec. carrierlanding, and full ranging capability,is "The present system comprises 6 k bits per page, produced without employing a computer. The and a 16 X 16 matrix of pages, for a total capacity demonstration model simulates an approach window of 1.5 M bits access time is 20 pc sec. Total optical two miles wide and a half-mile high and offers a insertion loss is 75 db, resulting in 70 k photons 3.5-degree glide slope. The six degrees (glide-slope impinging on eachbitdetector. ..and Bell deviation, localized deviation, depression angle, Labs scientistsproject that, by straightforward bearing angle, roll, and slant angle) are achieved extensions of the present system, 25 M bits with an mechanically,electronically, and optically.For access time of 7 p,sec is a feasible system. This example, roll is achieved as the vidicon itself is system would have 65 X 65 matrix of 6 k bit pages, a rolled; glide-slope deviation is simulated by manip- faster deflection system, and a, reduced insertion ulating the hologram. In the model, the generated loss of 65 db, resulting in 0.5 M photons per bit image allows a view which includes magnification at the detectors. of the holographic image of the carrier up to "Ultimately, it is predicted that a memory can 16-to-1 and permits views including one below be built having greater than 100 million bits of the deck of the carrier." (Electronics 42, No. storage, with an access time in the one microsecond 13, 46 (June 23, 1969).) range." (Modern Data Systems 1, No 2, 66 (Apr. "c .meral Electric has also examined the hologram 1968).) foi :,otential in charactek recognition. One method "Bell Labs has already constructed a 'bread- suggested by GE is to create a spatial filter using board' hologram memory system...that may a hologram. This filter can be used to detect, or eventually be able to display any one of 100 million recognize, specific shapes from among a random units of information upon one millionth of a second's assortment. notice. "This general scheme is the basis for a personnel "It is based on using a number of closely spaced identification system being developed by National holograms on a single photographic plate. Bell Cash Register Co., Dayton, Ohio. Labs had in mind switching operations as one funda- "According to NCR, two of the most important mental application. .. aspects of identification are signatures and photo- "This memory system works by directing a laser 100 beam to one 'page' (location of a hologram) inan stored not on a microscopicspot on the film but as array. Initial goals are to make each hologram part of an optical interference about a millimeter in diameter pattern which is and to space them contained in the entire hologram,"(Modern Data rather closely ina pattern of 100 rows by 100rows, Each hologram will Systems 1, No. 2, 66 (Apr. 1968),) store, encoded in the form of "A bad spot ina photographic image will not an interference pattern, another 100 by 100matrix, spoil all bits of informationcompletely; the Fourier This will be coded in dotsor blanks to represent transform of such information, a plate will still give a good Thereconstructedhologramwill image," (Bloembergen, 1967,p. 86). be aligned precisely withan array of phototran- sistors "Since information fromany one bit of the object (also under developmentat Bell Labs), is spread out over the whole hologram,it is stored which will `report' to the electronicdevice which of the dots there in a redundant form, andscratches or tears are present and which are absent, This of the hologram make onlya minor deterioration rollcallisthe message," (Photo Methodsfor in the overall reconstructed image, Industry 12, No. 3, 61-62 (Mar. In particular, 1969)), no single bit is greatly marred by such damageto 6.69"CarsonLaboratories,Bristol,Conn., for example is working the hologram." (Smith, 1966,p. 1298). on the development of "Leith reports that diffusedillumination holo- potassium bromide and similar crystalsas holograms have an immunity to dust and scratches graphic materials. and that particles have littleeffect in producing "The , laser is usedto bleach the crystal in erroneoussignals as in previous photographic accordancewith theholographicinterference memories." (Chapman and. Fisher,1967, p. 372). patterns. Such a memory device is saidto have a "Since light from the point capacity of 1 million bits source is spread over per square half-inch of the entire hologram's surface(thus ensuring inter- material." (Serchuk, 1967,p. 34). ference patternsover the entire film surface), 6.70"An experimental opticalmemory system that could lead. to any part of the hologram will reproduce thesame computer storage devicesa image as any otherpart of the hologram. Itcan thousand times faster than today'sdisk and drum be seen that the only effect of dustand scratches storage units was reported. .by three Inter- is to reduce the activearea of the hologram." national Business MachinesCorporation engineers. "In the experimental (Vilkomerson et al., 1968,p. 1199). system; blocks of infor- "Generally, the light projectedinto an image by mation are accessed bya laser beam in just ten- a hologram is not associated with millionths of a second. More any specific than 100 million bits point of the hologram, thus, if the hologrambecomes of computer information couldbe stored ona nine marred by dust square inch holographic plate or scratches there is little degrada- ... tion of any one point in theimage. Dust and film "The experimentalmemory system uses a laser imperfections can bea severe problem in non- beam to project blocks of informationcontained on holographic storage, because the hologram onto errors arise from the a light-sensitive detector. The degradationof specificbits."(Gamblin,1968, detector then converts theprojected hologram into pp. 1-2). electronic signals whichcan be processed by a computer. 6.72Further, "the results of thisstudy have indicated that holographic techniquesare particu- "In a feasibility model,assembled at IBM's larly suited to satisfy the Poughkeepsie, N.Y., SystemsDevelopment Divi- functional requirements of read-onlymemory ...Holography offers solu- sion Laboratory, size, direction,and focus of the tions to two key problems laser beamare determined by a series of lenses. associated with the The beam is positioned requirement for a single removablemedia storing on the hologram by a up to 160,000 bits. First, the unique redundance crystal digital light deflector.By controlling the polarization of the light from the inherent in holograms constructedwith diffused laser the deflector illumination eliminates the loss is used to selectany block of information storedon of data due to such a single plate. environmentaleffectsasdustandscratches. Second, the potential freedomfrom registration "The hologram splits thelaser beam into two effects which separate rays: one non-functional and Cit» be achieved by proper selection the othera of constructiontechniquesallowsthe manual first-order diffractionpattern which carries the insertion and removal of mediawith high bit holographic information. Thisfirst-order diffraction packing densities and does pattern is then focused not add a requirement on a light-sensitive detector for complicated mechanicalpositioning or complex array, which converts the optical informationto electrical interconnectionin the read unit." (Chap- electronic signals. The signals,representing data, man and Fisher, 1967, p. 379). are then sent to the computer's central processing 6.73 unit at high speeds." (bemaNews Bull. 5, Nov. "One can constructcomputer techniques 18, 1968). which would takean acoustic hologram (thewave- ont from a scattered sound wave)and transform 6.71"An advantage ofstoring information in it into an optical hologram, the form ofa hologram rather than asa single real thereby allowingus image is that the loss of data due to construct the three-dimensionalimage of the to dust and scatterer of the sound waves." (Lesemetal., film defects is minimized,since a single bit is 1967, p. 41). 101 6.74"In a paper presented at the International least a factor of four reduction in thecomputer Symposium on Modern Optics, researchers at the time needed to calculate the wavefrontpattern IBM Scientific Centre at Houston, Texas, de- necessary for equivalent image quality.Corre- scribed how they have programmeda computer spondingly there is a reduction in plotting time an IBM System/360 Model 50 to calculate the for the kinoform. interference patterns that would be created if "A further economy is achieved in thatno cal- light waves were actually reflected froma real culations involving a reference beam arenecessary. object. Neither the real objectnor actual light waves Finally, in the cases of one- and two-dimensional are required to produce holograms with the com- objects only real-number additionsare required, puter technique. While the initial IBM computer once the basic transform is calculated, to determine hologram experiments have been restrictedto the wavefront phase for plotting, The corresponding two-dimensional objects for research simplicity, quantity to be plotted for digital holograms is the the authors said further work is expectedto make wavefront intensity which requires multiplication possible digital holograms whichcan be recon. of complex numbers." (Lesem et al., 1969,p. 155). constructed into 3.D pictures. An engineer could 6.76"A[n] .important reason for synthe- then get a 3.D view of a bridgeor car body design sizing holograms is to create optical wavefronts without actually building the physical objector from objects that do not physically exist. A needto even drawing it by hand." (The Compoto Bull, form such a wavefront from a numerically described 1 1,No. 2, 159 (Sept. 1967)). object occurs whenever the results ofa three- "Morerecently,firmshaveexperimented dimensional investigation, for example, the analysis withcomputer-generated holograms for unique of an x-ray difiractogram must be displayed in data display, NASA's Electronics Research Center three dimensions." (Brown and Lohmann, 1969, in Boston, Mass., is said to be investigating making p. 160). real -time holograms for such applicationsas air- "Scientists, stock brokers, architects, statisti- port display to approaching aircraft. cians and many others who use computersmay "A team at IBM's Houston Scientific Research soon have a practical, fast, and inexpensive way Center has programmeda System/360 Model 50 of converting memory data into three-dimensional to create hologram by calculating the necessary pictures and graphs. interference patterns. "With a process devised at Bell Telephone "Thus it may soon be possible to use the computer Laboratories, it takes only a few seconds ofcom- to create a mathematical model of a device and puter time to turn equations, formulas, statistical then translate equations intoa three-dimensional data and other information intoa form suitable hologram of the mathematical model." (Serchuk, for the making of holograms. Viewed under ordinary 1967, p. 34). light,the holograms produce three-dimensional 6.75"Holograms of three dimensional images pictures that can display a full 360-degree view of have been constructed witha computer and re- the object shown. constructed optically. Digital holograms have been "Holography, which has been called gensless generated by simulating, witha computer, the wave photography,'recordsasubjectthroughthe fronts emanating from optical elements, takinginto interference of two laser beamson a photographic account their geometrical relationship. We have plate. One beam is aimed directly at the plate, studied in particular the effects of varioustypes of and the other reaches the plate after beingtrans- diffuseillumination. Economical calculations of mitted through, or reflected by, the subject being high resolution images have been accomplished `photographed.' using the fast finite Fourier transform algorithmto "In the BTL method, the original subject exists evaluatetheintegralsinKirchoffdiffraction only as a group of numbersor coordinate points theory. We have obtained high resolutionthree in three dimensions, far example, in thecom- dimensional images with all the holographicprop- puter's memory. The hologram is made intwo steps. erties such as parallax, perspective and redun- First, the computer is programmed to construct dancy." (Hirsh et al., 1968, abstract,p. H 104). a series of two-dimensional pictures, or projections, "Kinoforms serve for all of the applications of each showing the 3-D data froma precisely defined computer-generated holograms,e.g., three-dimen- unique angle. A microfilm plotter, connectedto sional display, wave conversion, read-onlystorage, the computer, producesa microfilm frame for each etc. However, kinoforms give a more practical, picture. computationally faster display construction that "In the second step, a holographictransparency yields more economicaluse of the reconstructing is made. The frames of the microfilmare used as energy and that yields only the desired image. subjects to make very small holograms (1to 3 mm "The principal computational advantage of kino- across), which are positioned sequentiallyon a forms as compared with digital holograms isem- holographic medium." bodied in the fact that all of the spatial frequency "Thus, a composite hologram is madeup of a content of the device is used in the formation of the series of small holograms, each of which is formed real image; none is required for the separation of with a two-dimensional image. But thecomposite the real and conjugate images. There is thenat image appears three-dimensional, and showsa 102 360-degree view of the object. With thistype of marginal improvementsover the past few years, hologram also invented at BTL the viewercan but no startling improvements have beenmade see the object rotating through a full cycle by simply in comparison to reductions in the basicdevice moving his head from sideto side in front of the geometry. hologram." (Computer Design 8, No. 6, 28(June "The consumption of realestate may be reduced 1969).) by interconnecting thelogiccircuits with the 6.77"By the use of photographic recording narrow lines allowed by the masking technology, techniques a very high information densitycan be thus reducing to a minimum thearea requirements achieved to which rapid randomaccess can be made for external lead pads. At thispoint, the semi- by appropriate electronic and opticaltechniques. lf, therefore, there conductor manufacturer relaxes andsays in effect are any classes of information to the computer designer: Reduceyour logic to a which must be read frequently, butare not changed for at least a week, then such few standard configurations, and Iwill reduce a storage technique costs by a large factor. Hence,we have a search for would be appropriate. This is evidentlythe case magic standardlogicfunctions." (Howe, 1965, for all system programmes includingcompilers and p. 507). monitor programs. ." (Scarrott, 1965,p.141). "Interconnections are alreadyour problem for "Photographic mediaare quite inexpensive, are designing and buildingsystems, and applying capable of extremely high bit densities,and exhibit Large Scale Integration (LSI)to digital systems an inherent write-once, read-only storage capability. will inevitably force the realization thatintercon- The optical read-out techniques, whichare used, nections will be more important in determining are nondestructive." (Chapman and Fisher, 1967, p. 372). performance than all other hardware factors.This is because the problems of physical sizeand bulk, 6.78"To get an order of magnitudeidea of the memory capacity, DC Shift over long cables, reflectionsand stub we will consider a memory lengths, crosstalk and RFI, and skineffect degrada- plane of 2 in.square ...There will be approximately 645 subarrays tion are making computer systemsinterconnection [individuallyaccessible]. limited." (Shah and Konnerth, 1968,p. 1). Consider that only one-half thememory plane is 6.81 composed of active film. The "One example of thesemore exploratory memory would then attempts is the optically addressedmemory with contain almost 13 million bits." (Reich andDorion, 1965, p. 579). microsecond nondestructive read cycle andmuch longer write cycles. Chang, Dillon andGianola 6.79"The inherentpower of optical processing propose such a changeable memory employing can be exploited without suffering the speed limita- gadoliniumirongarnetsasstoring elements." tions usually associated with static spatialfilters. (Kohn, 1965, p. 133). The method consists of usingan electron beam- 6.82"Maintaining lowpower supply and dis- addressed eloctro-optic light valve(EOLV) as the spatial filter. Thus the filter need tribution impedances in thepresence of nanosecond no longer be a noise pulses is an increasingly difficult problem .. fixed transparency, butcan instead be a dynamic As more circuitsare placed on a chip, decoupling device whose orientation is controlledelectronically of power supply noise will be requiredon or in rather than mechanically. Thisopens the method close proximity to the chip." (Henle and Hill, of optical processing to the domainof real time and 1966, p. 1858). presents exciting possibilities for itsuse in a variety 6.83 of applications." (Wieder "Integratedcircuitryhasbeen widely etal., 1969, p.169). held to be the most significant advancein com- 6.80"In optical transmission lines, thewave- puter technology since the development of the length of the signals will be shorterthan any of transistorinthe the circuit dimensions; therefore, mid-fifties...Semiconductor one could elimi- integratedcircuitsaremicrominiaturecircuits nate, for example, all the reactive effects in the with the active and passivemicrocomponents on interconnections." (Reimann, 1965,p. 247). "High-speed electronic or in active substrate terminals. In thin-film inte- computer circuitryis grated circuitry, terminals, interconnections,re- becoming 'interconnection limited. Thereactance sistors and capacitorsare formed by depositing a associated with the mounting andinterconnections thin film of various materialson an insulating of the devices, rather thanthe response of the active substrate. Microsize active components, is becoming the main factor limiting components are then the speed of operation of the inserted separately to complete the circuit. Micro- circuits. modules are tiny ceramic wafers made fromsemi- "A possible approachto computer development conductive and insulative materials. Thesethen that might circumvent interconnectionlimitations function either as transistors, resistors,capacitors, is the use of optical digital devices ratherthan or other basic components." ("The Impact...", electronic devices as active components."(Reimann, 1965, p. 9). 1965, p. 247). 6.84"Integrated circuit technology willbring "One factor of growing significance,as circuit revolutionarychangesinthesize,cost, and size is reduced, is the increasingamount of surface reliability of logicalcomponents. Lesser improve- area consumed by areas devoted to interconnections mentswillbe realizedincircuitspeed.... and pads for interconnections. There have been "Advances in integrated circuit logiccomponents 103 and memories,.,will provide significant reduc- central processor LSI might' be used to carry out tions in cost since the implementation of flexible more micro-operations per instruction;address character recognition equipment involves complex moreoperandsperinstruction;controlmore logical functions." (Hobbs, 1966, p. 37). levels of look-ahead; and provide both repetition "Of course, the cost of electronics associated and more variety in the types of functions to be with peripherals will be drastically reduced by executed. In system control, LSI might provide LSI. But the promise of LSI is greater than that. greater system availability through error detection, Functions that are now handled by mechanical error correction, instruction retry, reconfiguration parts will be performed by electronics, More logic to bypass faulty units, and fault diagnosis; more will be built into terminals, and I/O devices such sophisticated interrupt facilities; more levels of as graphic displays, in which the major cost is memory protection; and concurrent access to inde- circuits, will come into more general use." (Hudson, pendent memory units within more complex pro- 1968, p. 47). gram constraints. In system memory, LSI might "This speed power performance requires only provide additional fast local memory for operands modest advances from today's arrays. The board and addresses; improved address transformation modulesizeisconvenientforsmall memory capability; content-addressablememory; and applications and indicates the method whereby specialfast program statustables.In system LSI memories will establish the production volume input,output, LSI might provide more channels; and the impetus for main frame memory applica- improved interlacing of concurrent input-output tions. The LSI memory being produced for the Air operationswithautomaticmemoryprotection Force by Texas Instruments Incorporated falls into features; and more sophisticated pre- and post- this category." (Dunn, 1967, p. 598). processing of data and instructions to relieve the "The Air Force contract [with Texas Instruments] central processor of these tasks." (Smith and Notz, has as a specific goal the achievement of at least a 1967, p. 88). tenfold increase in reliability through LSI tech- "LSI has an inherent functional advantage over nology as compared with present-day integrated magneticsinassociativeapplications,namely circuits." (Petritz, 1967, p. 85). that fast bit-parallel searches can be achieved. "Impetus for continued development in micro- The main drawback of magnetic associative memo- electronics has stemmed from changing motivations. ries, even in those applications which require Major emphasis was originally placed on size re- relatively simple match-logic per word, isthat duction. Later, reliability was a primary objective. imperfect cancellation of analog sense signals and Today, .development of new materials and processes other noise effects give rise to a low signal-to-noise point toward effort to reduce cost as well as to ratio and thereby limit the technology to essentially further increase reliability and to decrease weight, bit-serial operation. Thus, the more nearly binary cube, and power." (Walter et al., 1968, p. 49). signals available from semiconductor associative "This [LSI] technology promises major impact devices seem to provide a unique advantage over in many areas of electronics. A few of these are: magnetics which is not strongly evident in compar- 1. Lower cost data processing systems. isons of the two technologies over other categories 2. Higher reliability processing systems. of memory." (Conway andSpandorfer,1968, 3. More powerful processing systems. p. 842). 4. Incorporationofsoftwareintohardware, "Content Addressable Memories: As the semi- with subsequent simplification of hardware." conductor manufacturer learns to produce more (Petritz, 1966, p. 84). and more components on a single silicon chip, reasonablysizedcontent-addressablememories "Computer systems built with integrated cir-may become feasible. Memories of thistype, cuits have higher reliability than discrete-com- available on a large scale, should permit significant ponent machines. This improved reliability is due to changes in the machine language of the computer, two factors: (1) the silicon chip has a higher relia- and possibly provide simplification in the design bility than the sum of the discrete components it of such software as operating systems and com- replaces, and(2)thehighdensity packaging significantly reduces the number of pluggable con- pilers." (Graham, 1969, p. 104). tracts in the system." (Hen le and Hill, 1966, p. 6.87"Revolutionary advances, if they come, 1854). must come by the exploitation of the high degree 6.85"One of the most interesting and significant of parallelism that the use of integrated circuits paradoxes of the new technology is the apparent will make possible." (Wilkes, 1968, p. 7). reconciliation of a desire to achieve high speed "One area in which r feel that we must pin our and low cost. The parameters which yield high hopes on a high degree of parallelism is that of speed, i.e., low parasitics, small device geometry, pattern recognition in two dimensions. Present-day alsoyieldlowestultimateproduction costin computers are woefully inefficient in this area." silicon integrated circuits." (Howe, 1965, p. 506). (Wilkes, 1968, p. 7). 6.86"Some examples of functional expansion 6.88"The recent advance from discrete tran- we would naturally consider are as follows. In the sistor circuits to integrated circuits is, about to be 104 overshadowed by an even greater jump to LSI memory elements, at least for some portion of circuitry. This new jump will result in 100-gate and primary storage. Even today some systems have then 1000-gate circuit modules whichare little scratchpadmemoriesconstructedof machine larger in size or higher in cost than the present logic elements, so that the fastprocessor logic four-gate integrated circuit modules." (Savitt et al., is not held back by the slowermemory capability." 1967, p. 87). (Pyke, 1967, p. 161). 6.89"Discrete components have givenway to 6.94"LSI memories show considerable potential integrated circuits based on conventional etched in the range of several hundred nanoseconds down circuit boards. This fabrication technique isin to several tens of nanoseconds. In contrast with turn giving way to large scale integration (LSI), logic, LSI memory isideally suited to exploit in which sheets of logic elements are producedas theadvantages andliabilitiesof largechips: a unit." (Pyke, 1967, p. 161). partitioningisstraightforwardandflexible,a "The initialresults suggest the possibility of high circuit densitycan be obtained with a manage- fabricating, in one step, a complete integrated able number of input-output pads, and the major circuit with all the passive elements. Sucha process economic barriers of part numbers and volume would start with a metallized substrate and would which confront LSI logicare considerably lower. use a programmable laser and work stage. Com- Small-scalememorycellchipshavealready plete laser fabrication of hybrid circuits will require superseded film memories in the fast scratchpad a process in which a metal film is removed selec- arena; the depth of penetration into the main- tively, exposing a different film. For example such frame is the major unresolved question." (Conway a process ma; be necessary in order to remove the and Spandorfer, 1968, p. 837). nonductor and expose the resistor film. 6.95"As technological advancesare made, "In the present tantalum-chrome-gold technology the planar technology permitsus to pack more such a selective removal of the gold presents sub- and more bits on a single substrate thus reducing stantial difficulties because the reflectivity of the theinterconnectionproblemandsimplifying gold is much greater than that of the tantalum the total memory packaging job. This integration nitride. It is quite probable, however, thatsome will reflect in the long runon product cost and combination of films will be found for which the product reliability." (Simkins, 1967,p. 594). upper film can be removed from the resistor without 6.96"The new technology hasa number of damaging it." (Cohen et al., 1968, p. 402). problems whose solution can be facilitated by 6.90"Integratedcircuits(moreimportantly, arranging the circuitry on thesearrays in a 'cellular' largescaleintegration(LSI)whichinvolves form thatis,inatwo-dimensionaliterative numerous integrated circuits tied together on the pattern with mainly local intercell connections same chip) offer the best promise from the stand- that offers such advantagesas extra-high packing point of size, reliability and cost [for scratch-pad density, ease of fabrication,simplifiedtesting memories]." (Gross, 1967, p. 5). and diagnosis, ease of circuit and logical design, 6.91"Of all the potential applications of large- the possibility of bypassing faulty cells, andpar- scaleintegration, new memory techniquesare ticularly an unusually high flexibility in function the most startling. Ferritecore memories have and performance." (Kautz et al., 1968,p. 443). just about reached their limit in terms ofaccess 6.97"LSI offers improvements incost and speeds required for internal scratchpads. Magnetic- reliability over discrete circuits and older integrated thin films, while fast enough,are too costly. Studies circuits. Improvement in reliability is dueto the show that because of LSI, semiconductor memories reduction of both the size and the number of are less costly than any other approach for speeds necessary connections. Reductions incost are from 25 to 200 nanoseconds and for capacitiesup due largely to lower-cost batch-fabrication tech- to 20,000 bitsjust the range required by scratch- niques. One problem in fabrication is the increased pads." (Hudson, 1968, p. 41). repercussion of a single production defect, necessi- tating the discarding ofan entire integrated com- 6.92"In addition to being used for circuitry, ponent LSI if defective instead of merelya single techniques applytotheconstructionof transistor or diode. This problem isattacked memories, since some of the newmemory elements by a technique calleddiscretionarywiring;a mentioned abovecan be fabricated using the computer tests for defective cells ina redundantly micro-constructiontechniques.Thepossibility constructed integratedarray and selects, for the also comes to mind of fabricating both thecom- good cells, an interconnectionpattern that yields parison circuitry and the memory cells ofa content- the proper function." (Van Dam and Michener, addressable memory intoa single unit. Thus the 1967, p. 210). developmentoflarge-scale integrationholds "Sincepackagingandinterconnectionsare considerablepromiseforimproving computer major factors in the cost of hardware." (Van Dam and Michener, 1967,p. 210). an integrated circuit, the cost potentials... canbe achieved only by 6.93"In view of theeconomy that should batch fabricating largearrays of interconnected accompany widespread use of LSI, it may become circuits in a single package. This raisesmany less expensive to use LSI logic elementsas main difficult and conflicting questions, suchas packag-

..541.1Stikisat4..k...1,,,akvaqAiglisliataUs.2}4 ingdesign,maintenance philosophy,flexibility advantage of these new elements is that a memory and functional logic segmentation." (Hobbs,1966, array made withthem can be batch-fabricated lower-cost pro- p. 39). in one step, leading to simpler, "The rapid and widespread use ofintegrated duction (the production of core memories requires circuitlogicdevicesby computer designers, making the magnetic cores and then stringing coupledwithfurtherimprovementsinsemi- the cores together to make a memory).Another conductor technology hasraisedthequestion advantage appears to be in memory speed.The of the impact of Large Scale Integration(LSI) new memoryelements have taken over the fields on computerequipment. Itis generally agreed of high-speed registers and temporary'scratch- that thisisa very complexproblem. The use pad'memories.Integratedcircuitry...will of Large Scale Arrays for logic requiresolutions probably replace the planar magneticthin-film to the problems such asforming interconnections, currently used for high-speed registers;however, debugging logic networks, specifyingand testing the planar film will be extended tointermediate- multistate arrays, and attempting tostandardize sized stores (105-106 bits)." (Van Dam andMich- arrays so thatreasonable production runs and ener, 1967, p. 207). low per unit design costs can beobtained." (Pet- 6.100"Today most common types[of core schauer, 1967, pp. 598-599). memories] have about a million bits and cycle "The advent of large-scale integrationand its times of about one microsecond,with bigger resultant economy has made it clear that acomplete and faster types available. Capacity andspeed re-evaluation of what makes a good computer have been constantly increasing and cost con- organizationis imperative. Methods ofmachine stantly decreasing." (Rajchman, 1965, p. 123). organization that provide highly repetitivelogical "The ferrite core memory with 106 bitsand subsystemsareneeded. As notedpreviously, 1/.4 sec cycle time is the presentstandard for main certain portions of present computers(such as memories on the computer market.Larger mem- successive stages in the adder of aparallel machine) ories up to 20.106 bits at 10/h seccycle time have are repetitive;but others (such as the control already been announced." (Kohn, 1965, p. 131). unit) tend to have unique nonrepetitivelogical "Ferritecores dominatedthe main memory configurations." (Hudson, 1968, p. 42). technologythroughoutthesecondgeneration. 6.98"Graceful performance degradation through Mostalthough notall, of the third generation use of majority votinglogic. machines thus far announced have corememories." "LSI will allow a single logical element tobe (Nisenoff, 1966, p. 1825). replaced by several logical elements in a manner 6.101"The NCR 315 RMC (Rod Memory such that the elements can be used todetermine Computer) has about the fastest main memory the state or condition of a situation.The state or cycle time of any commercial computer yetde- condition of the situation indicated by amajority livered800 nanoseconds. The uniquethin-film of the elements can be accepted asvalid hence, memory isfabricated from hairlike berryllium- majority voting logic." (Walter et al.,1968, p. 52.) copper wires platedwitha magnetic coating. "Because of low cost modules, penniesand less In the Rich's system the 315 RMC processesdata per logic function,maintenance will be simplified from over 100 NCR optical print cash registers ..." and maintenance cost will be reducedby using (Data Proc. Mag. 7, No. 11, 12 (1965)). throw-away modules. By 'wiring' the sparesin, fabricated on the same LSI wafer thatthey are A later version of NCR's rod-memory computer, toself-repaira the 315-502, adds multiogramming capability sparing,itbecomes practical time. (Datamation 12, computer.Thisisaccomplished by including at an 800 nanosecond cycle diagnostic logic (coupled with programs) toeffect No. 11, 95 (Nov. 1966)). theself-repair.Such aself-healingcomputer "NCR's new thin-film, short-rod memory repre- system, using electronic surgery,need only be sents one of the mostsignificant technical in- manually maintained when its spare partsbank novations in the Century series ...The rods becomes exhausted. Some advantagesof self- are made bydepositing a thin metallic film and repair are: then a protective coating on 5-mil copper wire. Increased system reliability This process yields a plated wire 0.006 inch in Continuous operation (system always avail- diameter, which is then cut into 0.110-inch lengths able) to form the 'bit rods'.The basic memory plane Long term (years) remote system operation is formed by inserting the bit rods intosolenoid without manual repair coils wound on a plastic frame. Then the entire Considerable reduction in maintenance plane is sealed between two sheets of plastic. costs." (Joseph, 1968, p. 152). Automated processes are used to plate the wire, solenoid coils, insert the 6.99See, for example, Rajchman (1965) and cut the rods, wind the Van Dam and Michener as follows:"However, rods into the solenoids, and seal theplanes. The new memoryelements, such as plated wires, result is a high-performance memory at an un- planar thin films, monolithic ferrites, and inte- usually low bit cost." (Hillegass, 1968, p. 47). grated circuits, are becoming competitive. One 6.102For example, "laminate-diode memories 106 of millions of bits operating in, less thanone micro- and the readout signals small." (Shively, 1965, second seem possible in thenear future." (Rajch- p. 637). man, 1965, p. 125). "The thinfilm transistor isbarely emerging "The cryotrons, the memory structure, and all from the laboratory and itmay require several connections are constructed by a single integrated years before it becomes a serious contender for technique. Thin films of tin, lead and silicon integrated-all-transistor-random-access-memories of monoxideareevaporatedthroughappropriate large capacities." (Rajchman, 1965,p. 126). masks to obtain the desired pattern of lines. "The development of higher-speed conventional The masks are made by photoforming techniques memory devices, of cores and thinfilms, has and permit simple fabricationof any desired slowed, and progress with such devices in breaking intricate patterns.. .. the hundred nanosecond barrierwillprobably "The superconductive-continuous sheet-cryotron- take some time." (Pyke, 1967,p. 161). addressed approachtolarge capacity memory "Thin films appeared to bemore hopeful and are offers all the qualities, in its principle ofopera- certainly an area where extensive research isbeing tion and itsconstruction, to support ambitions carried out. The main problems stillappear to be of integration on a grand scaleas yet not attempted those of production, especially the problem of by any other technology. No experimentalor achieving reproduceability fromone film to an- theoretical result negates the promise...There other." (Fleet, 1965,p. 29). is, however, a serious difficulty: The variation of "The development of Cryogenic memories has the thresholds of switching between elements in reached the point where planes storing several the memory plane." (Rajchman, 1965,p. 127). thousand bits can be fabricated with high yield. 6.103"A planar magnetic thin filmmemory However, there are still many problems, suchas has been designed and built by Texas Instruments interconnections, cost, data rate, etc., to be solved using all integrated circuits for electronics achiev- before considering a mass store large enoughto ing a cycle time faster than 500ns, and an access justifythe overhead of cryostat and dewar." time of 250 ns. The memory is organizedas 1024 (Bonn, 1966, p. 1869). words by 72 bits in order to balance the costs of "Key problems in the fabrication of largemono- the word drivecircuits against the sense-digit lithic memories are reliability (what to do ifa bit circuits. The inherent advantage of this particular fails) and the volatility of the monolithic cell (if organizationisthat the computer can achieve the power goes off, the information is lost)." (Henle speed advantage not only because ofa fast repeti- and Hill, 1966, p. 1859). tion rate, but also because four 18 bit wordsare 6.106 Kohn points out, for example, that "at accessed simultaneously. (Comparablecore memory present, such optically addressed memories seem designsareordinarilyorganized4096words to be capable of storing 105...106 bits/sq in. of 18 bits each.) The outlook is for higher speed to have about 0.1 pi, sec read access time, and one (faster than 150 ns) memories in similar organiza- cell can be written in 100 pi, sec. Very high voltages tions to be developed in planar magnetic films. for the light switches are required. Thisappears The cost of these memories will be competitive to be quite unfavorable from a technical point of with 2-1/2 D core memories of thesame capacity view; however, an intensive materials research but the organization and speedcan be considered may overcome the weakness of electrooptic effects to offer at least a 4:1 improvement in multiple and lead to more realistic devices." (Kohn, 1965, word accessing and a 3 :1 improvement in speed. As p. 133). a result of this, more computers will be designed to 6.107"In the subsystems ofa large computer, take advantage of the long word either by extend- one serious problem is ground plane noisespurious ing the word length of the computer itselfor by signals generated by large currents flowing in cir- ordering instructions and data in sucha manner cuits which have a common ground... that sequential addressing will be requireda large Another percentage of the noise nuisance arises when signals haveto be time." (Simpson, 1968, pp. coupled from two subsytems whichare operating 1222-1223). at two widely different voltages. Lumped together, 6.104"If the high-speed memory is to operate such difficulties are knownas the 'interface prob- at a cycle time in the 100-nanosecond region, the lem'." (Merryman, 1965,p. 52). class of storage elements that can be used issome- 6.108 what limited. Storage elements capable of switching "Superconductive cryogenic techniques, which were advocated for quick, on-linestorage, speedscompatiblewith100-nanosecondcycle will probably not become operational because of times include (a) thin magnetic films of several the high costs of refrigeration." (Van Dam and types, (b) some forms of laminated ferrites, (c) Michener, 1967, p. 207). tunnel diodes, and (d) semiconductor flip-floptype devices." (Shively, 1965,p. 637). 6.109"Theprojected'break-even'capacity, includingrefrigerationcost,for a cryoelectric 6.105For example, "most forms of thin films memory is approximately 107 bits." (Sass et al., and laminated exhibit adequately fast switching 1967, p. 92). times, but the drive current requirementsare large "The cryoelectronicmemory is made up of 107

376 -411 0 - 70 - 8 striplines, which, though interconnected from tions. Each program microinstruction from the main plane to plane, display low characteristic impedance memory addresses a sequence of microinstructions and high propagation velocity, and require modest in the read-only memory. Each microinstruction in peripheralelectronics.Therefore,propagation the sequence describes the state of the entire velocity is the only real limit to memory cycle time. machine duringitsnextcycle. The read-only Typical cycle time for the 108-bit AB system... memory divides easily into segments, since its is approximately 1 As." (Sasset al., 1967, p. 97). only external connectionsare the words address 6.110"The use of small specialpurpose mem- inputs and contri)1 signal outputs. LSI read-only ories has become prevalent inrecent years. The memories are being offered by several manufactur- Honeywell H-800 employeda small core memory ers." (Hudson, 1968, p. 42). to permit multiprocessing as far backas 1960." "If a read only-memory (ROM)module were (Nisenoff, 1966, p. 1826). used to store subroutines, the relative-efficiency 6.111"Much interest has recently been shown of the code would be much less important. ROM in the computer art in incorporatinga low-cost, modules cost less than one-fifthas much as com- mechanically changeable, read-only store in the parable amounts of maincore storage. Use of control section of a central processing unit. Flexi- ROM to 'can' bread and butter subroutines in low- bility of organization and compatibility with other cost hardware provides an effective solution toa systems can be built into a computer that has a particular problem. The mainor read/write memory, readily changeable read-only store. The printed thus liberated, can be used to provide feasible card capacitor Read-only store isone of three flexibility for the main program and to incorporate technologies selected for the ROS function in inevitable, unforeseen, jobs that arise during the System/360." (Haskell, 1966,p. 142). development and operating lifeof a computer "The Read Only Store (ROS)memory is a pre- system.... wired set of micro-instructions generally setup "One of the most significantaspects of fourth for each specific application. Thismeans that the generation computers will be theuse of read-only specifications of the computer may be tailoredto memories. Tradeoffsof hardware forsoftware suit the specific application of theuser. Thus in and/or speed, and/or reliability, will significantly an application where square root or some other affect computer organization. Advantagesto be special function must be performed rapidlyor gained through theuse of ROM include (1) increased repeatedly, such a sequence of operationsmay speed, output signal level and reliability, (2) de- be hard wired into the ROS." (Dreyer, 1968,p. 40). creased read-cycle time, operatingpower, size, "Micro-steps, the basic instructions ofa stored weight, and cost, and (3) nonvolatility." (Walter logic computer, permit theprogrammer to control et al., 1968, pp. 51, 54). the operation of all registers ata more basic level "The 'read-only' function includes thestorage than ispossible in the conventional computer. of indirect accessing schemes, the implementation Sequences of micro-steps (each of which requires of logic functions, the storage of microprogrammed only 400 nsec to perform)are stored in the ROS instructions, and related applications." (Chapman as 'micro-routines' which are executed much as and Fisher, 1967,p. 371). a conventional subroutine. However, unlike the "The attractions ofa good read-only storage unalterable commands of the conventionalcom- include not only extremely reliable nondestructive puter, stored micro-routines may be designed by readout, but also lower cost." (Picket al., 1964, the programmer to form the most efficientcom- p. 27). bination of basic computer logical operations for "Special hardware functions implemented in the a given application. The speed increase availableread-only memory of the 70/46 supplement the by use of a ROS does notcome from faster com- [address] translationmemory. They are used as puting circuits, but from operating instructions additional privileged instructions. Among thecapa- built into the hardware formore efficient ordering bilities they provideare the ability to load or un- of gates,flipflops, registers etal. Thus at the load all or selected parts of translationmemory outset of each application, tradeoff studies must and to scan translationmemory in such a way that be made to determine to what extent softwaremay only the entries of thosepages that have been be replaced by hardware throughuse of the ROS." accessed are stored into main memory." (Oppen- (Dreyer, 1968, p. 41). heimer and Weizer, 1968, p. 313). "The implementation of read-only memoriesas 6.112"A 1VIYRA memory element isa MYRi the control element in a computer has significance Aperture ferrite disk which, when accessed,pro- for maintainability and emulation. Instruction de- duces sequential trains of logic-level pulsesupon coders and controls presenta difficult problem to 64 or more otherwise isolated wires...A pico- the designer. These elements containno repetitive programmed system, then, consists essentially of patterns like those in data paths and arithmetic an arithmeticsectionand a modified MYRA units. In addition, they havemany external con- memory. A macroinstruction merely addresses nections. A read-onlymemory c n be used to pro- an element in the MYRA memory; when the element vide these same control signals. It would contain is accessed, it produces the gating signals which a long listhundreds or thousandsof microinstruc- cause the arithmetic unit to perform the desired 108

11 functions. In addition, itprovides gating pulses memory is that it has no explicit addresses. Any which fetch the operand (if needed),increment the reference to information stored inan associative control counter, and fetch thenext instruction." memory is made by specifying the contents ofa (Briley, 1965, p. 94). part of a cell. All cells in the memory which meet "Picoprogramming is realized by theuse of the the specification are referred to by the statement." MYRi Aperture (MYRA) element,a multiaperture (Feldman, 1965, p. 1). ferrite device which is the basic building block "We have described herean iterative cell which of the instruction mode. Each instruction module can be used as a content addressable memory is a complete entity and is fabricatedon a con- from which an entrymay be retrieved by knowing ventional printed wiring card thatcan be inserted part of its contents." (Gaines and Lee, 1965,p. 74). in a conventional PC connector. Incorporation ofa "Memory systems which retrieve information new instruction in the computer or alteration of an on the basis of a given characteristic rather than existing one is accomplished by the additionor byphysicallocation...arecalled'content- substitution of the appropriate instruction module addressed','catalog',or'associative'.In these card." (Valassis, 1967,p. 611). types of memory systems, an interrogation word 6.113 "With this GaAs diodearray system a ...is presented to the memory anda parallel very fast, medium capacity, read-only memory with search of all words within thememory is conducted. changeable contents becomes realizable.Since Those stored words which havea prescribed rela- many of the existing third generation computers tionship (e.g., equal to, nearestto, greater than, contain microinstructions in read-onlystores of etc.) to the interrogation wordare tagged. Sub- about the same capacityas that indicated for the sequently, the multiple tagged wordsor responses diode-accessed holographicmemory, it would seem areretrievedby some interrogationroutine." that the existing read-only memories could be (Miller, 1964, p. 614). replaced by this type of holographicmemory; "Associative Memories. An associativememory such a system could bean order of magnitude is a collection of storage cells thatare accessed faster and allow for increased flexibility of CPU simultaneously on the basis of content rather than configuration by easy change of the microinstruc- location. The ability to associate with circuit logic tion repertoire." (Vilkomerson et al., 1968,p. 1198). those cells with similar contents achievesa hard- 6.14In general, these termsare interchangable. ware implementation of an associatively linked Some typical definitionsare as follows: "Basically or indexed file. Sufficient quantitative results have an associative memory involves sufficient logical not yet been developed to establish conclusively capability to permit allmemory locations to be the merits of the hardware implementationas searched essentially simultaneouslyi.e., within against software associative systems. Acompre- some specified memory cycle time...Searches hensive study of hardware associativememory may be made on the basis of equality, greater-than- systems given by Hanlon should be read by those or-equal-to,less-than-or-equal-to, between limits, interested in this growing technology." (Minker and and in some cases more complex criteria." (Hobbs, Sable, 1967, p. 129.1 1966, p. 41). "An associativememory which permits the 6.115"It is extremely unlikely that large fast specification of any arbitrary bitpattern as the associative stores will become practicable in the basis for the extraction of the record within which near future ...We cannot expect associative this pattern appears is called stores to contribute to a solution of our problems a fully associative other than in very small sizesto carry out special memory." (McDermid and Petersen, 1961,p. 59). tasks, e.g., the page register addresses in Atlas." "The content-addressablememory (CAM) was (Scarrott, 1965, pp. 137-138). initially proposed by Sladeas a cryogenic device... "The concept of the content-addressablememory For this type of memory, word cellsare accessed has been a popular by the character of stored data ratherthan by one for study in recent years, physical location of the word cell. The character but relatively few real systems have usedcontent- of data is evaluated in parallel throughout addressable memories successfully. This hasbeen the partly for economicreasonsthe cost of early memory. A common addressing characteristic is designsof equality of stored data andsome externally pre- content-addressablememorieshas sented key. Memories of thistype have also been been very high and partly because itis a difficult called associative sincea part of the cell contents problem to embeda content-addressable memory may be used to call other cells in an 'associative' into a processing systemto increase system effec- chain." (Fuller, 1963, tiveness for a large class of problems." (Stone, p. 2). 1968, p. 949). "An associativememory is a storage device that "Considerations ofcost make it impossible for the permits retrieval of a record stored within by associative memory 'to contain referring to the contentsor description rather than many registers, and the relative address within the memory." (Prywes, the number that has been adoptedin current de- 1965, p. 3). signs is eight. Unless the associativememory has very recently been cleared, it will benecessary to "The distinguishing feature ofan associative suppress an item of information in order to make 109

1 IA room for a new one; the obvious thing is tosuppress total of 1.0 microsecond, the item that has been there for or an average of 250 the longest period of nanoseconds per word." (Gluck,1965, p. 662). time, but other algorithms slightlycheaper to imple- "The rationale behind theinclusion of local ment have also been proposed. It is claimedon the scratchpad memories in basis of simulations that eight the B8500 computer associative registers moduleencompasses...the need for buffering enable the full procedure of threememory cycles to of four-fetches of instructions and be shortcircuited data in advance on 90% of occasions." (Wilkes, of their use, i.e., lookahead. Alsoimportant are 1967, p. 4). its uses as storage for intermediate "In the past, associative results, as an or content addressable economicalimplementationforregistersand memories of any significant size have beenimpracti- cal for widespread counters, and for the extension of the push-down use. Relatively small associative stack." (Gluck, 1965,p. 663). memories have been built withvarious technologies, such as multiaperture 6.118 "A specific application fora CAM is ferrite cores, cryotrons, and encountered when assemblingor compiling pro- various thin-film techniques. Thelogical flexibility grams where it is common to refer to variables, of microelectronicsnow makes at least seratchpad- locations and other items interms of a symbol. size associative memories practical."(Hudson, 1968, The value or information associatedwith each p. 42). symbol must be stored somewhere 6.116 in memory "By a slavememory I mean one which and a table must be madeto relate each symbol automatically accumulatesto itself words that come to its value. As an example, the symbol ABLKR from a slower main memory and keeps them avail- may be assigned the value 5076. Thecomputer able for subsequentuse without it being necessary may take this information and store the value for the penalty of mainmemory access to be in- 5076 at location 1000 for example. curred again. Since the slave Then the first memory can only be a entry in the symbol table will relate thesymbol fraction of the size of the mainmemory, words ABLKR to the location 1000where the value of cannot be preserved in it indefinitely, and there ABLKR is stored. Asmore symbols are defined, must be wired into the systeman algorithm by which this symbol table willgrow in length." (Rux, 1967, they are progressively overwritten.In favorable p. 10). circumstances, however, a good proportion of the 6.119"Tied in with scratchpad No. 2is a small words will survive long enoughto be used on sub- 28-word associativememory (19 bits per word) sequent occasions and a distinct gain ofspeed whose use enhances the utilizationof the scratchpad results. The actual gain dependson the statistics of the particular situation. memory by providing content addressingas well "Slave memories have as the conventional binary coded word addressing recently come into promi- capability." (Gluck, 1965,p. 663). nence as a way of reducing instructionaccess time in an otherwise conventionalcomputer. A small, 6.120"Each cell of thememory receives signals very-high-speed memory of,say, 32 words, accumu- from a set of pattern linesand command lines in lates instructionsas they are taken out of the main parallel, and the commandsare executed simul- memory. Since instructions oftenoccur in small taneously in each cell. One ofthe commands orders loops a quite appreciable speedingup can be each cell to match itscontents against the pattern obtained.... lines. Each cell in whicha match occurs sets its "A number of base registers couldbe provided matchflip-flopand alsogenerates an output and the fastcore memory divided into sections, each signal..." (Gaines and Lee, 1965,p. 72). serving as a slave toa separate program block in These investigators describesome of the differ- the mainmemory. Such a provision would, in prin- ences between their proposed system and others, ciple, enable shortprograms belonging to a number in part as follows: "Thememory we describe here of users to remain in the fastmemory while some is a logical and practical outgrowth ofthe content other user was active, beingdisplaced only when addressable distributed logicmemory of Lee and the space they occupiedwas required for some other Paull.However,thereareseveralsignificant purpose." (Wilkes, 1965,pp. 270-271). differences: the inclusion ofa 'match' flip-flop and 6.117"The B8500 scratchpadsare implemented a 'control' flip-flop in each cell of thememory, the by magnetic thin film techniquesdeveloped and addition of a 'mark' lineto activate many cells organized into linear-select memory arrays...To simultaneously, and the control ofthe propagation realize the high speedaccess requirement of 45 of the marking signal. Asa consequence of these, nanoseconds, the reading functionis nondestructive, the memory hassome novel capabilities, among eliminating the need fora restoring write cycle when which are the abilityto simultaneously shift the data are to be retained unchanged. ... contents of a large group of cells, thus openingor "Insertion of new data into the localmemories closing a gap in thememory, and the ability to (writing) can be accomplishedwithin the 100- simultaneously mark strings of interestin separate nanosecond clock period of thecomputer module." parts of the memory. (Gluck, 1965,p. 663). "By properly manipulating thecell states, simple "4 52-bit wordscan be requested from a memory module and received programs for correcting errors involving missing at a computer module ina or extraneous letters, multiple mispellings,etc., 110 can be devised. Furthermore, by using the marking in a stack are cheaply and reliably interconnected capabilitiesofthememory,errorcorrection using a Iry batch interconnection technique which during retrieval can be accomplished on a selected resembles an injectionmolding process, using subset of strings which may be located at widely molten low-temperature solder. The circuit which separated parts of the cell memory." (Gaines and results is a resistor matrix where the information Lee, 1965, p. 75). stored is in the form of a connection pattern. The 6.121This Sylvania development involves the matrix may be operated as a content-addressable use of automatically preprocessed plastic sheets or associative memory, so that the entire array to affect the performance and logic behavior of a can be searched in parallel, and any word or words solenoid-transformer array. storedanswering agivendescriptioncan be "The interrogation, which may consist of a retrieved in microseconds." (Lewin et al., 1965, number of descriptors, each containing many bits p. 428). of information, causes an appropriate group of 6.123"The study by Dugan was restricted to solenoids to be driven...The solenoids interact considering an existing computer environment and simultaneously with all emolosed loops on all the theGoodyear AssociativeProcessor (GAP),a data pknes, resulting in a simultaneous voltage 2048-word associative memory with related logic on the output of each data plane that is the cross- andinstructions. A benchmark problem was correlation between the dthen input solenoids and studied in which the data base exceeded the size each individual data plane. The output of each of GAP and was stored on disc. The disc-stored plane is connected to its own detector-driver which data required transfers to the associative processor tests the output in comparison with all the other or the conventional core for further operations. data plane outp'uts to find that output The study concluded that the effectiveness of a the best correlation. Alternatively, the detector- small associative processor, such as GAP, for driver can be set to test for some pre-determined formatted file problems depended upon the inter- threshold." (Pick and Brick, 1963, p. 245). face of the associative processor with the computer Brick and Pick (1964) describe "the application systeni, the logic of the associative processor, and of the solenoid array principle to the problem of theload/unloadcharacteristicsof the memory word recognition, code recognition, and (in a limited associated with the problem. The authors showed senso), associative memory. The proposed device, that embedding the associative processor within based entirely on existing experience with a large the core memory provided the best system. It also character recognition cross correlator, is capable providedafacilityforperformingarithmetic of recognizing one of 24,000 individual English operations on data, which is ordinarily difficult words up to 16 letters long. The simultaneous for an associative processor. The study did not correlation and selectionis made in less than show any major advantages in using a system with 3 Asec. The selection can be made either on a anassociativeprocessor similarto GAP over perfect-match or a best-match basis." (Brick and one without an associative processor... . Pick, 1964, p. 57). "Gall utilized the same computer environment as "This form of semipermanent memory offers Dugan but .investigated a dictionary lookup phase many advantages to computer and memory users. of an automatic abstracting problem. He concludt,s Among these are: a) ease of contents preparation, that incorporating associative memories that do involvingautomatedpunchingofinexpensive not have the capacity to store the entire data base standardized cards; b) reliability, as a result of requires excessive data transfer and cannot compete fewelectricalconnections,loosemechanical with conventional systems that employ a pseudo- tolerances, and passive components; c) low cost, random mapping of a word onto a storage location since the cards are not magnetic and need only a and, therefore, can locate a word by content. continuous conducting path; and d) high speed, Randomizedaddressingis anothersoftware with estimated cycle time below 111 sec." (Pick simulation of but one of the facilities provided by et al., 1964, p. 35). an associative processor, namely, the so-called 6.122 "A number of associrtive memory stacks `exact-match' 'function." (Minker and Sable, 1967, of 120 resistor cards have been constructed, each pp. 130-131). stack storing 7200 bits, with each card storing 6.124TheLibrascopeAssociativeParallel one word of 60 bits length." (Lewin et al., 1965, Processor was developed for use in the extraction p. 432).. of pattern properties and for automatic classifica- "This paper describes a fixed memory consisting tion patterns. It is noted in particular that "the of one or more stacks of paper or plastic cards, para)lel search function of associative memories each of which contains an interconnected array requires that comparison logic be provided at each of printed or silk-screened film resistors. Each memory word cell. The APP, by moderate additions card is compatible with conventional key punches, to this logic, allows the contents of many cells, and information is inserted by the punching of a selected on the basis of their initial content, to pattern of holes, each of which breaks an appro- be modified simultaneously through a 'multiwrite' priateelectrical connection. All punched cards operation." (Fuller and Bird, 1965, p. 108). 111 Swanson comments as follows: "Fuller, Bird, savings in quantities of logic necessary for a CAM and Worthy recently described two machines; since one set of comparison logic can be used to an associative parallel processor programmed to compare the key register with many memory loca- Abstract properties from visual and other patterns tions. The comparison logic need only monitor the and classify the patterns from the properties; andmemory's contents as it passes through the circulat- an associativefileprocessor for rapid parallel ing system... search of large complex data bases." (1967, p. 38). "The principle disadvantage of a circulating 6.125"The ASP machineorganization... CAM is speed. M least one circulation time of the [has as its dominant element] the context-addressed memory isrequiredtointerrogatethe entire memory ...[which] stores both ASP data and memory. In the case of a magnetic drum system, programs, and ...providesthecapabilityto this time would be measured in milliseconds which identify,inparallel, unknown items (and link is much too slow for many applications. However, labels) by specifying the context of relations in with the use of glass delay lines, information can which the unknowns appear....[It] consists of be stored at very high rates, 20 MHZ and higher, a square array of identical storage cells which are and short circulation times can store large amounts interconnected both globally and locally. Each cell of data. For example, a 100 microsecond delay contains both memory and logic circuitry. The line at 20 MHZ can store 2,000 bits of information. memory circuitry stores either an item, link label, Thus 32 delay lines could store 2,000 words of 32 or a relation, plus tag bits. The main purpose ofbits each and this memory could be searched in the logic circuitry is to perform the comparison 100 ihsec." (Rux, 1967, p. 2). operations which are required to implement global 6.128 "A goal in designing and interfacing the searches of the array and local inter -cell com- associative mapping device into the System/360, munication." (Savitt et al., 1967, p. 95). See also Model 40, was to introduce no time degradation in note 5.47. the critical main memory address path. We have 6.126"In theearliestassociative memories accomplished this goal by designing the hardware all bits of all the words of the memory were simul. to perform this address translation function in taneously compared with a search word; this is 220 ns. This interrogate time through the associative called word-parallel search. For such word-parallel memory is approximately 50 ns and the remaining search, the memory has to be of the nondestructive- time is spent in wire delay and conventional logic, readout (ND110) type." (Chu, 1965, p. 600). such as the encode circuit which was designed Instead ofword-parallelsearch,bit-parallel using the 30-ns IBM SLT family... . search has been developed because of its simpler "The technology used to implement the associa- design and because word-parallel search is of less tive mapping device is the IBM SLT technology. importance in more complex searches. Bit-parallel Four special circuits were designed for the associa- search(orbit-sequentialsearch) searches one tive memory array. They are the associative memory corresponding bit of all words at one time. For a cell used for storing one bit of information, the bit word of 64 bits, a maximum of 64 bit-parallel driver, the word driver, and the common sense searches is made in succession; thus, bit-parallel amplifier used for sensing a mismatch signal in search pays a price in speed...[but] the price the word direction or a binary 'one' signal in the is a limited one. In a bit-parallel-search associative bit direction." (Lindquist, et al., 1966, p.1777). memory,nondestructivereadoutpropertyof "The mapping device which provides the dynamic memory elements is not necessarily required. This storage allocation function in the time-shared system paper describes the organization of a destructive- is a 64-word, 16 bit per word, associative memory." readout associative memory which can be imple- (Lindquist et al., 1966, p. 1776). mented by a special, very high-speed, magnetic- ..The Univac 128-word by 36 bit-per-word, corememoryusingconventionaltechnology.7 600-nsec scratch pad memory." (Pugh et al., 1967, (Chu, 1965, p. 600). p. 169).

"Because parallel-search logicis implemented "The memory...utilizes a plated-wire (Rod) for only one long-word, implementation of several memory device operating in a 512-word 36 bit per varieties of search logic is practical. In addition word memory system. The DRO mode is employed to a bit-comparison logic, other logical operations and operation at a 100-nanosecond read-write cycle (such as NAND, NOR, AND, OR) can be imple-time is achieved." (Kaufman et al., 1966, p. 293). mented relatively simply and less expensi- Jly." "The memory is word-organized witha capacity (Chu, 1965, p. 600). of 64 words each 24 bits long. Cycle time is approxi- "For these operations [bit count and bit count mately250 nanoseconds. Such memoriesare and store], each bit of the memory short-word may suitable for use as `scratchpads' operating within represent an attribute (a property or a character- thecentralprocessororinput-outputcontrol istic), and the count of attributes is a useful argu- systems of a computer." (Bialer et al., 1965, p. 109). ment for searching closeness in attributes." (Chu, "All of the memory circuitsapproximately 180 1965, p. 605). chipsplus 1,536 bits of thin-film magnetic storage 6.127"Circulating memories offer an enormous and the thin-film interconnection wiringare on a 112 glass substrate measuring 3 by 44/2 by 1/10 inches. 6.133"It has been demonstrated that1000-bit The circuitsoccupy about half the substratearea. NiFe film DRO memories with cycletimes of The extremely small physicalsize of the memory, 60 n sec andaccess tittles of about 30 it sec can be the shorter signal paths, theelimination of redun- built using existingcomponents. Experience with dant connections, which ,packagedcircuits would this model indicates that the designcan be extended have required, all contributeto an improvement in to allow a significant increase of capacityin a system speed.. memory having this same cycle time and "The 64-word access memory has a cycle time of about time; however, it is felt that to achievea marked 250 nanoseconds. Plans are to builda 256 -word increase in speed will require radicaldepartures memory that is equally fast and expectationsare that eventually 50-nanosecond from the conventional circuit andarray techniques memories can be that were employed in themodel described here." built with similar design andfabrication methods (Anacker et al., 1966,p. 50). [i.e., ultrasonic face-downbonding for interconnec- "IBM has developeda bipolar monolithic IC tion of integrated circuit chips withthin-films]." buffer memory for (Bialer et al., 1965, use on the 360/85 that is faster pp. 102-103). than any they have previouslyintroduced. Access 6.129"The sonic filmmemory represents a time to the entire contents ofthe 2K by 72-bit novel approach to thestorage of digital information. memory is 40 nsec. The buffermemory is con- Thin magnetic films and scanningstrain waves are structed of half-inchsquare building blocks com- combined to realizea memory in which information posed of two silicon chipsand their leads and is stored serially. Theremanent property of mag- insulation. Each of the chipsmeasures less than netic films is used for nonvolatilestorage. The effect an eighth of a square inch and contains 664com- of strain waveson magnetic films is used to obtain ponents (transistors, diodes, and resistors).Each serial accessing. This effect is alsoused to derive a chipprovides64distinctbutinterconnected nondestructivereadsignalforinterrogation." storage cells. The components involvedare so (Weinstein et al., 1966,p. 333). minute that 53,000can fit into a one square inch 6.130"The new [tunnel diode]memory system area. contains 64 words of 48 bits each, andtest results "The significance of themicrominiaturization is from a partially-populated cross-sectionalmodel of course little relatedto 'how many of what fit indicate a complete READ/RESTOREor a CLEAR/ where.' What IBM gains fromthis construction is WRITE cycle time of less than 25 nanoseconds." a circuit speed demonstratedon some experi- (Crawford et al., 1965,p. 627). mental chipsthat isas fast as 750 picoseconds (trillionths of 6.131"The basic cell employsa thick magnetic a second). film as the high-speed sensing element "The speed of the buffermemory (which at one to sense the time was to be called information which is storedas a pattern of magnets a 'cache', but that term has on a card. Since the magnet card is separate from apparently been dropped) isnot down to the the array, the lattercan be permanently laminated 750 pisec figure, buta 7 n sec/chip read and a or sealed and the information can be changed 12 n sec/chip write speed isn'tbad." (Datamation very I:5, No. 4, 193 (Apr. simply and reliably. The advantagesof this system 1969).) stem from a combination of several important fea- 6.134"Electronic Memories, Inc.,demonstrated its NANOMEMORY 650. . tures, namely card changeability, high speed,wide . capacity of 16,384 mechrnical and electrical tolerances,and a linear words of up to 84 bits, andan access time of drive-sense relationship which resultsin a wide 300 n sec." (Commun. ACM9, No. 6, 468 (June range of operating levels. 1966).) "Circuit costs can be minimized by 6.135"The ICM-40,a oneµ sec cycle time, using low-level 500 nsec access time, drivers, giving an additional increasein speed with core memory, available with capacities from 4K X 6 bitsto 16K X 84 bits has only a minor increase insense circuitry...For a memory containing four arrays of 256 words and been announced by ComputerControl Company, 288 bits per word,an access and cycle time of 19 Inc." (Commun. ACM 9, 316 (1966).) 6.136"International Business Machines and 45 ns respectivelywas achieved. .." (Matick Corp. et al., 1966, p. 341.) has developedan experimental thin-film computer memory that has a 120-nanosecond cvs,le time, 6.132 a These investigators suggest further that 589, 824-bit capacity and fits ina frathe 68 by 42 "the number of bits ofstorage can be increased in by 7 inchesincluding the several ways. A modular approach electronic circuits for can be used by driving and sensing." (Electronics39, No. 3, 41 connecting 64 X 8 memories in parallelor the mem- (1966). ory boards can be redesigned to accept the larger 6.137 number of bits. The modular approach "The memory hasa capacity of 8192 is particu- words, 72 bits per word, and hasa cycle time of larly applicable to the distribution ofsmall memories 110 nanoseconds and of various sizes throughout an access time of 67 nano- a large computer. It is seconds. The storage devicesare miniature ferrite possible to constructa 64 X 32 memory using either cores, 0.0075 by 0.0123 by 0.0029 inches, and of the above approaches witha cycle time of approxi- are operated in a two-core-per-bit destructive read- mately 20 nanoseconds." (Cattet al., 1966, p. 330). out mode. A planar arraygeometry with cores 113 resting on a single ground plane is used to control addressing of two different word lines is 20 ns." drive line parameters. Device switching speed and Seitzer, 1967, p. 172). bit line recovery are treatedas special problems." 6.142 "A single layer composite magneticfilm (Werner at al 1967, abstract,p. 153). isoperated ina rotational destructive-read-out 6.138"It seems a certainty that plated wire mode with two access wires. Each bitis composed memories will become a very important member of two 2 X 6 mil intersections of the wordand in the hierarchy of storage systems to be used in digit lines with a density of 12,500 bits/in, Mag- the computers of tomorrow." (McCallister and netic film structures which provide fluxclosure Chong, 1966, p. 313). in the hard, easy, or both directionswere con- "Both UNIVAC 920%9300 Systems utilizea new sidered by rejected when adequatemargins were plated-wire memory for internal storage featuring obtained with the single layer. Although theopen a non-destructive read-out mode and monolithic structure has fabrication advantages, the closed circuitry." (Commun. ACM 9, 650 (1966).) structures remain of interest for future work." 6.139"Capacity of this memory is 4096 68- (Raffel et al., 1968, pp. 259-260). bit words (278, 528 bits, to be exact) and itoperates "The access time of thememory from change with a cycle time of 200 nanoseconds andan access of address to informationoutput from the buffer time of 160 nanoseconds. It is a word-organized, flip-flops is about 450nsec. The largest contribution random-access memory. Thememory element is to this delay is the transienton the sense-line due composed of a pair of planar thin films coupled to group-switch voltage transitions. The circuit- together and read out destructively." (Meddaugh limited cycle time for read-writeor clear-write is and Pearson, 1966, p. 281). 600 nsec. Recovery from the digit-pulsetransient 6.140"The operation of this half-microsecond- limits the total cycle timeto 1psec with the digit cycle memory module representsa significant transient overlapping the group-switch transient." achievement in a program of magnetic thin-film (Raffel et al., 1968,p. 261). developmentforcomputer storage which was 6.143"The chainsare made from copper begun at these laboratories in 1955. Large numbers strips which have been plated witha NiFe film of substrates were processed and tested, and and are used tocarry word current. The bit/sense memory plane assembly and test are now routine signals are carried in wires whichpass through operations. the holes in the 'links' of the chain. Thememory "Memory frames which contain 20 substrateselement thus formed willoperate in a rotational (15,360bits) switching mode andcan be used for a word-orga- can be assembled without great nized memory." (Geldermans difficulty... et al., 1967, abstract, p. 291). "A shorter memory cyclecan be made possible 6.144"Ithas been shown thathigh-speed by reducing the totalsense delay, and by the elimina- chain memoriescan be built in very high-density tion of the bit recover pulse. The pulsetrans- arrays with minimum electromagnetic interactions. formerswillbe replaced by activesolid-state The bit/sense wirescan be treated as homogeneous devices. A reduction of 150nsec 50 nsec from a transmission lines with relatively high character- shorter sense delay and 100nsec from elimination istic impedence (100 Cl) and good signal-to-noise of the bit recover pulsemake a cycle time ofratios. The word lines are high-impedence strip 350 nsec, 3-Mc operation, possible." (Bittman, lines whose inductance is mainly determinedby 1964, p. 105). the nonlinear magnetic film. This makes evaluation "Fabrication, assembly, and operation ofthese more difficult, but implies favorable properties half-microsecond memories hasproven that large for the design of very long lines. numbers of reliable film substratesare producible "Based on the analysis of recently platedchains and that the completed memoriescan compete with smaller dimensions and better films, thechar- in both speed and price with the high-speed 2-1/2 acteristics of various possible chain memories have D-type core memories. The future for planar films been extrapolated. Straightforward design philoso- looks very brightboth larger and faster memories phy, using transistor selectioncan be applied for a are in the design stage. These memories willcom- 0.3 X 106-bit NDROmemory, a 106-bit, 100-nsec bine the economic advantages of batch fabrication DRO memory, anda 38 X 106-bit 500-nseo DRO with the fast switching properties of thin-films." memory. (Jones and Bittmann, 1967,p. 352). "These performance predictions reflect themerits 6.141"Extensive memory research aimedat of a film device with complete flux closure and high- implementing the inherent 1-ns switchingcapa- quality oriented filmsas exhibited in the chain bilitiesof thin magnetic films withina system device; they appear quite attractive for theirsize, environment hasresultedinacross-sectional speed, and circuitry requirements. Chains implya 147000-bit capacity filmmemory model with a simple semi-batchprocess and combine fast rota- nondestructive -READ -cycletimeof 20 ns,an tional switching properties of oriented filmswith access time of 30 ns, and a WRITE-READ time the larger signal capability of cores." (Abbaset al., interval of 65 ns. The shortest time interval between 1967, p. 311). 114 1.145Further,"the memory under develop- used in the unique 2.1/2 D selection organization, ment has a capacity of108 bits. This capacity is which incorporates an extra wire for sensing the achieved by stacking 107-bit modules into one interrogated bits. The total package with all electron- 72'1X 25" X 28"," unit,.,Each module has its own set of drivingics and power supplies measures circuits and sense amplifiers. This arrangement (Computer Design 7, No. 6, 70 (June 1968).) leads to a fast, random-access memory,readily "A new, duplex version of the Potter RAM, a realizable mechanically;itis justified from the magnetic tape random access memory. The new viewpoint of modularity and cost because the elec- unit has the same performance characteristics as tronic circuits are shared by a large number ofbits, its predecessor, 50.2 million bits of information All modulus share one set of auxiliary circuits, packed at 1,000 bp; and an average access speed which include the address decoders, timing cir- of less than 90 milliseconds." (Commun. ACM 8, cuits, information registers, and power supply ... 343-344 (1965).) "The plated wire used is a nondestructive read- "Available in storage capacities up to 32 million out (NDRO) element with equalword currents for bits (1,024,000 words of 32 bits), a new magnetic reading and writing. This property makes it unneces- core memory has a cycle timeof less than 4 micro- sary to have rewrite circuitryfor each stored bit," seconds. Dependent upon quantity and capacity, (Chong et al., 1967, p. 363). the price will be as low as 1.1/2 cents per bit. The 6.146 "Much attention is currently focused onModel CM-300 is said to offer true random access the development of block-oriented random-access at speeds and capacities not previously available memories. One prospectisthemagnetosonic in static storage devices. As such, this memory is delay-line memory which employs magnetic stor-expected to forge a place in the hierarchy of bulk age and block-access bysemiconductor electronics storage peripherals permitting greater programming the flexibility and increased computer throughput. A (to cause the propagation of a sonic wave in 2-wire, 2.1/2D magnetics organization and field selected line). Nondestructive read-out is derived on high reliability the digit lines in sequence by the propagation sonic proven circuitry are utilized to assure carried out by the coincidence and wide operating margins. All circuits in the wave, and write-in is system have been subjected to verifiable worst case of digit currents and the propagating wave.Another prospect is the opto-electricread-only memory,'design. Modular design permits exceptional flexi- where the stored information on high resolution bility in selecting memory interface, ease of main- block-selected by opticaltenance and low logistic support cost.Lockheed photographic plateis Electronics Company, Los Angeles, Cal." (Modern means, employing light-beamdeflection or an array of light-emitting elements. The optical readout (ofall Data Systems 1, No. 2, 74 (Apr. 1968).) the bits in the block in parallel) is converted to "The LIBRAFILE 4800 is one of a series of large electric signal by an array of photosensitive ele-capacity,high-speedhead-per-trackdiscfile ments. Holographic techniques areproposed formemories developed by the Systems Division of the implementation of the high-density photographic Librascope Group. It has a capacity in excess of 400 processing. The practicality of these block-orientedmillion data bits, with an average access time of systems are too early to be realisticallyappraised." 35 milliseconds. Additional memory modules may be (Lo, 1968, p. 1465). added to increase storage and the head-per-track design permits bit parallel data transfers to meet 6.147For example: "A new mass core memory which offers data access time of 1.5 microsecondsinterfaceand speed requirements." (Computer and capacity of up to 20 million bits has been placed Design 7, No. 6, 22 (June 1968).) on the market as a standardproduct by Ampex "Laboratory developments completed prior to Corporation. The RM, which is suitable for use with the initiation of the program described here demon- most large scale computers and data processing strated that tape speeds well in excess of 1000 systems now in production or use, will be available inches per second (ips) and packing densities of for delivery early in 1968." (Data Proc. Mag. 10, one million bits per squareinch, with high data No. 1, 58 (Jan. 1968).) reliability, were feasible. Using these developments "A randomly-addressable, low cost magnetic as a basis, a large memory systemcould therefore mass core memory system with a storagecapacity be designed. A prototype 'small' system with a total of 0.5 megabytes at a cost of I to 2 cents per bit is storage capacity of 1011 bits has been built and now available from Ferroxcube'sSystems Division. tested. The following is a description of such a The new memory offers the optimal compromise system and its major components. between cost, bit transfer rate and capacity. It has TBM system description: "The TBM (terabit a full cycle time of 2.5 AS and iscapable of operation memory, i.e.,1012 bit memory) random access in ambients to 105° F. The memory system can be memory uses magnetic tape as abasic storage organized in word capacities of from 9 to 144 bits medium. Random access is provided by using tape (in multiples of 9) per 524-K byte module. Any search speeds of 1000 ips (compared to approxi- number of modules can be connected for series or mately 300 ips used on conventional transports) parallel operation to build systems of almost infinite and by using packing den*ies of 700,000 bits per storage capacities. A total of 4.7 million cores are square inch (compared toapproximately 14,000 115 bits/in.for standard computer tape transports). made of a new material are expected to increase the Data recording is done in the transverse mode (to head life in these applications by a factor of 10 the direction of tape motion) using rotating heads, times or more thereby reducing the service costs of the technique used for video recording. A redundant users. recording scheme permits the achievement of error "The practical process for growing single-crystal rates of two errors in 10"0 information bits. The manganese-zinc ferrite has been developed by salient features of the tape transport, permitting using a technique similar to that used in producing this high search speed, and of the recording mode synthetic gem stones. This material not the con- and associated data channels, permitting this high ventionallydesignated`monocrystalline'ferrite, packing density, will be considered in detail follow- which, though composed of large-size crystals, ig ing the system description." (Damron et al., 1968, actually polycrystalline. pp. 1381-1382). "The single-crystal ferrite, as the name implies, "The NCR 353-5 Card Random Access Memory is a single, completely homogeneous crystal, with (CRAM) Fileprovideshigh-speed randomor no grain boundaries to permit crystal pullout which sequential processing of data for NCR 315 computer is frequently responsible for the familiar crumbling systems. The data recording is done on magnetic or wear of the contact face and gap edges. The cards 3.65" x 14", each containing 144 recording superior mechanical properties of this new ferrite tracks. Each track has a storage capacity of 1,500 are further enhanced by proprietary glass-bonding, 6-bitalphanumericcharacters.The removable or metal-bonding processes. Heads fabricated from cartridge houses 384 magnetic cards, providing a single-crystal ferrite and bonded by this means pre- total storage capacity of 82,944,000 6-bit alpha- sent a monolithic contact surface of extremelyhigh numeric characters. density and very low porosity in which the magnetic "Any card from a cartridge can be dropped to gap can be controlled to within .7.L. 5 mcroinches or read/write position within 125 milliseconds, provid- less and original sharp edges and machined pro- ing throughput of 5 cards per second. Data is files preserved intact through thousands of hours of transferred to the processor at the rate of 50,000 operation. alphanumeric characters per second. "Characteristics include an initial permeability "Up to sixteen CRAM Handlers may be connected (uo) of 2250 ± 250 at 100 kHz, 350 ± 50 at 5 MHz." to the processor. This provides an onlinefile (Comp. Design 8, No. 1, 30 (Jan. 1969)). capacity of over 1,327,104,000 alphanumeric char- 6.149"There are severe problems in locating acters (over 1,990,656,000 digits)." (Management and trackinginformationstoredatvery high Report: MASS RANDOM ACCESS FILES from density. Servo-techniques (also being pursued for NCR, nd., p. 1). higher track density in magnetic recording) based 6.147a"Enthusiasts of Bell Telephone Labs' upon track-seeking principles are essential for beam recently-patentedsingle-walldomainmagnetic scanning approaches." (Hoagland, 1967, p. 259). memory claim it may some day obsolete the disk. 6.150"Adapting videotape recording methods to By controlling the magnetic domains, millions ofcomputer systems may increase the capacity of bits can be stored in a diameter less than a micron. bulk, random access computer memories a thou- The action can actually be seen through a micro- sandfold. According to Dr. William A. Gross, Ampex scope, according to one source. vice president, an experimental system now in the "DevelopedbyWilliamSchockley, Andrew lab stores 50 billion bits on single 10- by 1/2-inch Bobeck, and H. E. Scoville, the memory works on reels of magnetic tape, or about 1000 times the the spin moments between electrons and the capacity of reels currently in use. Information can nucleus in a magnetoplumbite material containing be accessed and transferred in less than 10 seconds. rare earth orthoferrites." (Data Proc. Mag. 1 1 , "A finished memory based on these developments No. 8, 19 (Aug. 1969).) would enable a user to place all of his digital rec- 6.148"Magnetic recording bit and track densi- ords on line, ready for random access. This would ties, each an order of magnitude higher than those eliminate shelf storage and delays when a disk now used, have been demonstrated in the labora- pack must be located and placed in the system. tory. Twenty thousand bits per inch and one thou- "This videotape recorder increases recording sand tracks per inch have been reported. The practi- density by using four recording and playback heads calapplicationof theseexperimentsrequires mounted on a small metal disk that rotates perpen- considerabledevelopmentofmagneticheads, dicularly across the moving tape. This rotary head recording media, and track location techniques." increases tape-to-head speed by six times that of (Bonn, 1966, p. 1868). the fastest fized-head device and enables the record- "Ferroxcube Corporation has announced the ing of tv pictures or, when applied to coded data, development of a monocrystalline ferrite material increases the density." (Data Proc. Mag. 1 1 , No. 2, for use in magnetic recording heads. The new ma- 14 (Feb. 1969).) terial, with its increased wear resistant character- 6.151"A light sensitive recording process called istics, is expected to find wide use in video and high Photocharge, uses a photoelectricpotentialof density tape-recording applications where recording material in the film to produce images. It was head wear is a significant problem. Recording heads invented by Dr. Joseph Gaynor and Gordon Sewell 116 [G. E. Advanced Technology Labs.]. Its main the world's largest computer storage system is advantage lies in the fact that no external electrical only 0.000075 percent. This statistic is based on test charges or fields are required, as in conventional data in 'Error Detection and Correction in a Photo- electrophotographic processes. Digital Memory System,' an article in the November "Light and heat alone produce a completely issue of the 'IBM Journal of Research and Devel- developed picture which consists of minute defor- opment.' The journal article describes the trillion- mations on the surface of the film. bit system's error-correction techniques and shows "Recorded images can be erased by reheating how the system can correct errors rapidly enough the film.It is then ready to receive and record for real-time operation. The Photo-Digital Storage another picture.... System was built by the International Business "Since the material is limited in the range of Machines Corporation under a special contract light to which it is sensitive and requires brighter for the U.S. Atomic Energy Commission. The AEC light than conventional film, Dr. Gaynor anticipates contract specified the allowable error rate as no that early applications probably will be directed more than one 300-bit line with uncorrected errors toward memories for graphic and pictorial informa- in 2,700,000 lines read. The system has demon- tion and high resolution, inexpensiVe imago repro- strated a much better average of one line with un- duction." (Bus. Automation 12, No. 7, 50 (1965).) corrected errors in 13,500,000 lines read. Con- 6.152"A technological development prerequisite structed by the IBM Systems Development Labora- to the widespread exploitation of large, multiaccess tory, San Jose, Calif., the system functions in a data-base systemsisthelarge,low-cost,fast network of computers at the Lawrence Radiation random-accessmassstore.Kuehler & Kerby Laboratory in Livermore, Calif. A second Photo- describe an IBM. 'Photo-Digital' storage device, Digitrl System, using the same error-correction which is a step in this direction. The article gives techniques and having approximately one-third nocostinformation,which makes evaluation the storage capacity of the first, has been installed difficult, and the long random-access delay is a at the Lawrence Radiation Laboratory in Berkeley, handicap; the other properties of the device are Calif." (bema News Bull., Dec. 16, 1968, p. 6). neverthelessratherexciting.Briefly, a modest "The theory of error detection and correction configuration can store 10 12 bits of data, accepting codes is finding its way into practice. With these them at a rate of about 0.2 megabits per second and codes, machines can reconstruct data that have sequentially reading them at a rate of about 1 been destroyed. The first practical implementation megabit per second. Random access requires a of a powerful and complex code was achieved in maximum 3-second delay for acquisition of the the trillion-bit IBM Photo-Digital Storage System, proper 'cartridge.' A cartridge is roughly equivalent which uses a Reed-Solomon code for correcting up in capacity to a reel of magnetic tape; the reading to 5 error characters in a line of 50 data characters. rate after acquisition is also roughly equivalent to "With a correction facility able to correct bursts that of a high-performance tape drive.It would of errors in a line, a recorded bit need not be larger require nearly two months of continuous writing at than the average flaws in the recording medium. the maximum rate to fill a 1012 -bit configuration." In this system flaws cannot be eliminated by pre- (Mills, 1967, p. 226). testingthe medium, because the recording is Recent IBM papers describe error detection and permanent on silver halide photographic film. How- correction and data recovery techniques for this ever, since the film has high resolution, there is a system. For example: "In a high-density photo- tradeoff between the cost of the correction facility digital storage system, contamination and other and the cost savings of high-density recording." defects can easily obliterate a group of data bits. To (Griffith, 1969, p. 456). operate successfully in spite of this problem in the 6.153"The principle of the UNICON Computer IBM Photo-Digital Storage System... apowerful Mass Memory isderived fromthe UNICON error-correction code is used... Coherent Light Data Processing System to create "The problem of effectively implementing a code and detect (record and reproduce) information of this complexity has been solved by a number of elements in two dimensions by means of signal- innovations. Most important is the use of hardware modulated coherent laser radiation... for encoding, calculation of the power sums, and "Continuous readout of the UNICON system error detection, while using a control processor, on utilizes a lightguide surrounding the imaging circle a time-sharing basis, for error correction.Another of the rotating objective, carrying the laser radiation important featureisthat single-character error transmitted through the unidensity film to a central correction is tried first;if this is not sufficient, photomultiplier. Hence, any coherently illuminated further correction activities can be tried. Other information bit is photoelectrically detected within important features are use of a 'trial and recheck' few nanoseconds.... method of error correction, selection of a symmetric "Width of the information-carrying area of the code polynomial, use of a tabic of logarithms for 16 mm Unidensity film is 8 mm. Information packing multiplication and division in a Galois field of 64 density is 6.45 X 108 bits per square inch. Rate of elements, et cetera." (Oldham et al., 1968, p, 422). information processingisinthemegabits-per- "The chance for error in a line of data read by second range. Total capacity of one UNICON 117 memory system is 88 X 109 bits fora 16 mm Uni- formation, allowinga typical installation to hold density film reel of 100 feet." (Becker,1966, pp. several trillion bits of dataon line." (Data Proc. 711-712). Mag. 1 1,No. 8, 73 (Aug. 1969)). "The Laser Recording Unit designed and devel- 6.155"Thin oped by Precision Instrument Companyprovides dielectricfilmswhichexhibit a means for reliably and permanently recording and sustained electronic bombardment inducedcon- reproducing digital data. ductivity (SEBIC)appear to satisfy the control "The Laser Recording Unituses a new type of layer requirements for high sensitivity andstorage. permanent recording process which employsa "Thin films of cadmium sulfide whichexhibit laser to vaporize minute holes in the metallicsurface SEBIC were first developed by the Hughes Research of the recording medium. In this Laboratories.... manner digital "SEBIC layers information is recorded in parallel data tracksalong can store information in the the length of a recording medium strip. Thetracks form of two dimensional conductivity modulations are spaced on the order of five to ten microns (micro- with almost photographic resolution.In addition, meters), center-to-center; each track is composedthey can be excited with brief pulsesof high of bit cells three to five microns insize. For re- energy electron beams, and they are reusable because they can be erased almostinstantaneously. cording or reproducing sequential tracks of digitalIn a sense, they data, the maximum transfer rate of the LaserRe- may be thought of as a fc,rm of real cording Unit is approximately four millionbits per time photographic filmthe principal remaining second, with an average unrecoverableerror rate problem concerns the readout of information." of one in 108 to 109 bits, dependingon the data (Lehrer and Ketchpel, 1966,p. 533). density selected. 6.156"This program is devotedto the prepara- "The Laser Recording Unit includesa program- tion and investigation of novel kinds of datastorage mable Recorder Control Subsystem whichcan be elements of about micron size, and high-density designed to provide a hardware andsoftware inter- regular arrays thereof, to be addressed withan elec- face compatible witha specified computer system. tron beam of diameter comparable to the element "The major benefits offered by the LaserRecord- size. Such storage mosaicsare formed by developing Unit as amass digital-data storage unit are ing and adapting appropriate thin-film deposition summarized below: and micromachining techniques. Thelatteris (1) Permanent Storage: Datado not degrade based on the use ofan electron beam probe to over a period of time of the order ofyears. exposeanelectron-sensitiveresist. A storage (2) Compact Storage:Data are storedat a capacity of about 108 bits is believedto be realizable densityapproximately 250times.greater and accessible withan electron beam, without than that of digital magnetictape. mechanical movement of thestorage surface. Cur- (3) Unlimited Readout: Datacan be repeatedly rently we are investigatingtwo kinds of elements. readout for long periods of timewithout The first one is an electrically isolatedmicro- reduction in qualityor damage to the record. capacitor.. .at the bottom of a hole in a metal- (4) RecordingVerification:Essentiallyerror- dielectric-metal film sandwich. The otherconsists free data records result from thesimultaneous of an isolated washeror ring of metal embedded in read-while-writeverification capability that the dielectric of a multilayer metal anddielectric isuniquetothelaserhole-vaporization film sandwich. Ultimately, elements 1/4kin di- method of permanent recording. ameter orsmaller, spaced approximately 1/2pt (5) Low Error Rates: Theaverage unrecoverable center to center, are expected to be feasible,repre- error rate is approximately one in 108 to 109 sentingultimatepackingdensities up to 4 X bits. 108/cm2." (Rogers and Kelly, 1967,p. 1). (6) Economical Data Storage: Recordingof large 6.157 quantities of dataon the Laser Recording "The selection of materials and thick- Unit and permanent ne, ses for the recording media and substrates is storage of the data in on obtaining maximum sensitivity to a mini- P1 Record Strips significantly decreasesthe - di power density in the recording spot, while cost-per-bit of recording andstorage imposed maintaining adequate contrast for the byexistingmethods."(McFarlandand readout Hashiguchi, 1968,p. 1369). means selected and stability to the anticipated environment. This basis for selection impliesthat See also notes 6.18, 6.19, 6.20. an increase in the absorption efficiency of the 6.154 "A Photo-Optical Random AccessMass medium is useful only if it leadsto a more sensitive Memory (FM 390) with multibillionbit capacity, media and/or improvedcontrast. Excellent record- announced by Foto-Mem, Inc.,can be used to ing have been achieved at highrates with coatings replace or supplement magnetictape, disk or having less than 20 percent absorptionat the laser drum units. Used separatelyor combined into one wavelength. Additional coating and substratecon- system, the FM 390 uses a Photo-Data Card (PDCtm) siderations are: adequate adhesionto one another, for data storage. Advantagesover magnetic storage abrasion resistance,permanency, cost, etc., and are in cost and space saving. A typical Foto-Data special considerations,e.g., the use of a mica sub- Celltm with 100 PDC'sstores 3 billion bits of in- strate to help obtain certain unique properties in the 118 1Y, Rfi

thin films of MnBi which are used for Curie point The reduced packing density necessary for accept- magnetic recording." (Carlson and Ives, 1968, p. 1). able error rates cause these approaches to suffer 6.158"The results of the studies described in by comparison with magnetic recording." (Gross, this paper have established laser heat-mode record- 1967, p. 6). ing as a very high resolution real-time recording process capable of using a wide variety of thin 6.162"The advantages of electron beamsover film recording media. The best resultswere ob- light are a thousandfold increase inenergy density, tained with images which are compatible with easy control of intensity and position, and a sub- microscope-type optics. The signals are in elec- stantial increase in resolution. To offset these tronic form prior to recording and can receive ex- advantages,therearethecomplications of a tensive processing before the recording process demountable vacuum system." (Thornley etal., occurs. In fact, the recordings can be completely 1964, p. 36). generated from electronic input." (Carlson and Ives, 1968, p. 7). 6.163"...Some [media] like thermoplastics, 6.159"Instead of recording a bit as a hole in a involve nearly reversible changes, and the noise card, it is recorded on the file as a grating pattern content therefore rises with use." (Gross, 1967, p. 2). of a certain spacing...A number of different grating patterns with different spacings can be 6.164"The standing-wave read-only memory superposed and when light passes through, each is based on the Lippmann process... [in which] grating bends the light its characteristic amount, a panchromatic photographic emulsion is placed with the result that the pattern decodes itself... in contact with a metallic mirror...Sufficiently The new system allows for larger areas on the film coherent light passes normally through the emul- to be used and lessens dust sensitivity and the possi- sion, reflects from the mirror, and returns through bilityof dirt and scratch hazards." (Commun. the emulsion. This sets up standing waves witha ACM 9, No. 6, 467 (June 1966)). node at the metallic mirror surface. Developable 6.160"Both black-and-white and color video silver ions form in the regions of the antinodes recordings have recently been made on magnetic of the standing wave...If several anharmonic film plated discs.... waves are used to expose the same region of the "Permanent memory systems employing silver- emulsion, each will set up a separate layer struc- halide film exposed by electron or laser beams. It ture ...Conceivably, n color sources spaced is possible to record a higher density with beams. appropriately over the band of sensitivity could Readout at an acceptable error rate is the major provide n information bits, oneper color, at each problem." (Gross, 1967, p. 5). location". (Fleisher et al., 1965, p. 1). "A recentlyconceived memory whichuses Advantage would then be taken of "...the optical readout. Instead of recording bits as pulses, Bragg effect, which causes the reflected light to bits are recorded as frequencies. An electron beam, shift to shorter wavelength as the angle of incidence intensitymodulatedwiththeappropriatefre- increases...With this method, a monochromatic quencies, strikes the electron sensitile silver-halide light source, say of violet color, could read out the film moving transverse to the direction of tape mo- violet bit at normal incidence and the red bit at the tion..." (Gross, 1967, p. 8). appropriate angle from normal. Hence, a single "For recording analog information Ampex has monochromatic source, such as a laser, could be used to read out all bits.. focussed efforts on silver-halide film.. [which] ." (Fleisher et al., 1965, can be made sensitive to either electron or laser p. 2). beams...packing density is an order of magnitude Further, "random word selection requiresa sum- greater than the most dense magnetic recording." mation of various injection lasers...or the use of (Gross, 1967, p. 6). a white light source in which all colors are present. "Recent work at Ampex indicatesthatthe This source isthen deflected to the selected Kerr magneto-optic effect is likely to be practical location by the electro-optical deflector. The output for reading digital information. Recordingon a from the memory plane is then separated into the reflectiveplated tape formagneto-optic repro- various colors by means of a prism or other dis- ducing can be done by local heating with a laseror persive medium for a parallel bit readout". (Fleisher electronbeam."[Eras able,potentialdensity et al., 1965, p. 19). 1 X 108]." (Gross. 1967, p. 8). 6.165"A feature of the SWROM [standing-wave 6.161"At first glance, machining with electron read-only memory] whichappears to be unique is beams, or adding ions, appear to be suitable for itscapability of storing both digital and video recording digital information. However, problems in (analog) information. This feature, combined with obtaining sufficient linearity in the transfer func- the capability of the memory for simultaneous, tion (the dynamic range and signal-to-noise limits), multibit readout with minimalcross talk, will give and accurately positioning the electron beam for the SWROM an even wider range of application." reading make it impossible to read out the potential (Fleisher et al., 1965, p. 25). recording density with acceptableerrorrates. 6.166"Parallel word selection...could be 119 ., 1, .e. . s 7 7 7F,..11745,14.7, 41:777 .

accomplished by fiber-optic light splitting. It could is that of the Et or series of bits to be selected. This also be accomplished by flooding the area to be type of word selection would be useful for associa- read out with monochromatic light whose frequen,.:y tive word selection." (Fleisher et al., 1965, p. 21).

7 Debugging, On-Line Diagnosis, Instrumentation, and Problems of Simulation

7.1"The quantity and quality of debugging that virtually any program can be speededup 25 to must be faced up to in facility design. This is 50% without significant redesign! Unless monitored perhaps the area which has been givenmore lip and measured, a program's efficiencymay easily service and less attention than any other." (Wagner be as low as 25%. What is worse, multiprogramming, and Granholm, 1965, p. 284). multiprocessing, real time, and other present-day "Software checkout still remains an unstructured methods have created such a jumble of interactions art and leaves a lot to be desired for the production and interferences that without instrumentation it of perfect code." (Weissman, 1967,p. 31). would be impossible to know where effort applied "Debugging, regardless of the language used, is for change would yield the best return. One tries one of the most time consuming elements in the to mine the highgrade ore first, while it still exists." software process." (Rich, 1968, p. 34). (Bemer and Ellison, 1968, p. C40). "It has been suggested that... we are now 7.3For example, "another practical problem, entering an era in which computer use is 'debugging- which is now beginning to loom very large indeed limited'." (Evans and Dar ley, 1966,p. 49). and offers little prospect of a satisfactory solution, 7.2"As computing systems increase in size and is that of checking the correctness of a large pro- complexity, it becomes both more important and gram." (Gill, 1965, p. 203). moredifficulttodetermine precisely whatis "With the introduction of time-sharing systems, happening within a computer. The two sorts of the conventional tools have become almost worth- performancemeasurements whicharereadily less. This has forced a reappraisal of debugging available are not very useful; theyare the micro- procedures. It has become apparent thata new type timing information provided by the manufacturer of debugging tool is needed to handle the special (.4 microseconds/floating add) and the macro-timing problems created by a time-sharing system. These information provided by the user ("why does it special problems result from the dynamic character take three days to get my job back?"). The relation- of a time-sharing systema system in which the ship, if any, between them is obscured by the program environment is continually changing, a intricate bulk of the operating system; if it isa system in which a user is unaware of the actions of multi-programming or time-sharing system, the other users, a system in which program segments obscurity is compounded. are rolled in and out of storage locations, and a "The tools available to the average installation system in which one copy of code can be shared by for penetrating this maze are few and inadequate. many users. To debug in this dynamic environment, Simulation is not particularly helpful: the informa- the programmer needs a debugging supportsys- tion which is lacking is the very information neces- tem a set of debugging programs that operate sary for the construction of an accurate model. largely independently from the operatingsystem Trace routines interfere excessively with theopera- they service.. . tion of the system, distorting storage requirements "What is needed for time-sharing is a debugging as well as relative timing information. Hardware support system that meets the following require- monitors are not generally available, and thougha ments: wondrous future is foreseen for certain of them, The system should permit a systempro- they have yet to demonstrate their capabilities in grammer at a user terminal to debug system an operational environment; furthermore, they are programs associated with his task. When certain to be too costly for permanent installation, used in this manner, the support system should and perhaps too cumbersome for temporaryuse. operate in a time-sliced mode. The peripheral processor of the Control Data 6000 When used to debug a separate task, thesup- series computers, however, provides some installa- port system should provide the facility to tions with an easily utilized, programmable hard- modify a system program in relation to that ware monitor for temporary use at no extra cost." task, without affecting the programas executed (Stevens, 1968, p. C34). in relation to other tasks. "Without instrumentation, the user is swimming When a system program bug cannot be lo- against the tide of history. It is commonly thought cated and repaired from a user terminal, the that a good programmer naturally achieves at least support system should permit a skilled sys- 80% of the maximum potential efficiency fora tem programmer at a central location to sus- program. But while systems have increased greatly pend time-sharing activity until theerror is in size and complexity, the average expertise of located and repaired. The support system programmers has decreased. In fact, it is axiomatic should then permit time-sharing activity to be 120 resumed as though there had beenno inter-program debugging versus debugging in a batch- ruption. The support system should permita processing mode." (Evans and Darley, 1966,p. system programmer to monitor the progress of48), by 1968 Sackman et al., could report "on the any task from a remote terminal or from thebasis of the concrete results of these experiments, user's terminal. theonlineconditionsresultedinsubstantially The support system should contain the facilityand, by and large, significantly better performance to remain dormant until activated by a speci-for debug man hours than the offline conditions." fied condition. When activated by the condi- (Sackman et al., 1968, p. 8). tion, the system should be able to gather 7.6"We have, in general, merely copied the specified data automatically and then permiton-line assembly-language debugging aids, rather processing to continue. than design totally new facilities for higher-level In its dormant state, the support system should languages. We have neither creatednew graphical not impact the performance of the parent time- formats in which to present the debugging infor- sharing system. mation, nor provided a reasonable means by which The support system shoulduse a minimum of users can specify the processing required on any main storage and reside primarilyon high- available debugging data. speed external storage. "These features have been largely ignored be- The support system should be completelycause of the difficulty of their implementation. Independent of the time-sharing system (that The debugging systems for higher-level languages is, it must use none of the facilities of the are much more complex than those for assembly parent system), and it must be simple enough code. They must locate the symbol table, find to eliminate any requirement for a support the beginning and end of source-level statements, system of its own. and determine some way to extract the dynamic "An effort is currently underway to produce a information neededfordebugging aboutthe time-sharing support system thatmeets these reprogram's behavior, which is now hidden ina quirements." (Bernstein and Owens, 1968,pp. 7, 9). sequence of machine instructions rather than "There has been far too littleconcern on the part being the obvious result of one machine instruction. of the hardware system designers with the prob- Is it any wonder that, after all this effort merely lems of debugging of complex to create a minimal environment in which to per- programs. Hardware formon-linehigher-levellanguagedebugging, aids to program debugging would beamong the most little energy remained for creating new debugging important hardware aids to software production. -aids--that -would probably require an increased On-line debugging is essential. It shouAbeigo-s-Sibre-dynamic information-gathering capability. to monitor the performance of $,Dftwaie on a cathode "EXDAMS (EXtendable Debugging And Moni- ray tube console, without-hiterfering with the per- toring System) is an attempt to break this impasse formance of the .soft-w-are. It should be possibleto by providing a single environment in which examine areas-of peripheral storageas well as areas users can easily add new on-line debugging aids of core serge." (Rosen, 1968,p. 1448). Fu tier, "the error reporting rate from to the system one-at-a-time without further modify- a program ing the source-level compilers, EXDAMS, Of their .tem of several million instructions is sufficient to programs to be debugged. It is hoped that EXDAMS occupy a staff larger than most computing installa- will encourage the creation of tions possess." (Steel, 1965,p. 233). new methods of 7.4"By online debugging debugging by reducing the cost ofan attempt we mean program sufficientlyto make experimentationpractical. debugging by a programmer in direct communica- At the same time, it is similarly hoped that EX- tion with a computer (through, typically,a type- DAMS will stimulate interest in the closely related writer or teletype), making changes, testing his but largely neglected problem of monitoringa program, making further changes, etc., all with a program by providing new ways of processing reasonably short response time from thecomputer, the program's behavioral information andpresent- until a satisfactory result is achieved." (Evans and ing it to the user. Or, as a famous philosopher Dariey, 1965, pg 321). once almost said, 'Give me a suitable debugging 7.5"Another area of contact between hardware environment and a tool-building facility powerful and debugging is involved with trapping...The (and simple) enough, and I will debug the world'." user may ask for a trap on any combination of a (Balzer, 1969, p. 567). number of conditions, such as a store intoa specified 7.7"Diagnostics have been almost nonexistent register, execution of an instruction ata specified as a part of operating software and very weak as location, or execution ofany skip or jump instruc- a part of maintenance software. As a result needless tion. The debugging program handles the interrupt time is spent determining the cause of malfunctions; and reports the relevant information to the user." whether they exist in the program, the hardware, (Evans and Darley, 1966, p. 44). the subsets, the facilities or the terminals." (Dan- It is to be noted that although these authors, tine, 1966, pp. 405-406). as of 1966, were concerned that "very little data 7.8"Another advantage of computer simulation seems to exist on the relative efficiency of on-line is that it may enable a system'smanager to shrink 121 the anticipated real world life of hissystem into a degradation of the environment." (Estrinet al., relatively short span of simulation time. Thiscapa- 1967, p. 645). bility can provide the manager witha means of "The Nightwatchman' experimentsare directed examining next week's (month's, year's) production toward the maintenance problem. Attempts will be problems this week; thus he can begin to antici- made to structure a maintenance concept that will pPte the points where the operations will require allow for the remote-automatic-checkout of all the modification. Moreover, he can examine alternative computers in the network from a single point. The courses of action, prior to their being implemented concept is an extension of the `FALT' principle in the system, to determine which decision ismost mentioned previously. Diagnosticprograms will be effective. For example, the manager can increase sent over the lines, during off-use time, to check the processing load in the simulation to determine components, aggregates of components, complete where the saturation points are. Once these have modules, and the entire system. The 'Sentinel' been determined, he can hold these overloading station of the network will be responsible for the states constant and vary the other variables (e.g., gathering of statistical data concerning the data, number of service units, types of devices, methods the queries, the traffic, and the overall operations." of operations) to determine how best to increase the (Hoffman, 1965, pp. 98-100.) system's capacity." (Blunt et al., 1967,p. 76). "The Sentinel is the very heart of the experi- Mazzarese (1965) describes the Air Force Cam- mental network. It is charged with the gathering bridge DX-1 system witha "dual computer con- of the information needed for longrange planning, cept" that permits investigators to changecomputer the formulation of data automation requirements, logic and configuration inone machine without and thestructuringof prototype systems.In interference to programs whichrun on its inter- addition to the gathering of statistical data, the connected mate, especially for study of real time sentinel will be the control center for the network, data filtering operations. generating priority, policy, and operational details. 7.9"A techniqueforservicing time-shared The responsibility for the observance of security computers without shutting them down has been and proprietary procedureswillrest with the developed by Jesse T. Quatse,manager of engi- sentinel station." (Hoffman, 1965,p. 100.) neering development in the Computation Centerat "This data was taken by aprogram written to run the Carnegie Institute of Technology. The tech- as part of the CTSS Supervisory Program. The nique is called STROBES, anacronym for shared- data-taking program was entered each time the time repair of big electronic systems. It includesa Scheduling Algorithm was entered and thuswas test program to exercise the computer, and modi- able to determine the exact time ofuser state fied test gear to detect faults in the system." (Elec- changes." (Scherr, 1965, pp. 27-28). tronics 38, No. 18, 26 (1965)). 7.10"Diagnostic engineering begins in the initial "Data taken over the summer of 1964 by T. Hastings...indicates that the averageprogram phases of system design. A maintenancestrategy accesses (i.e., reads or writes) approximately 1500 is defined and the system is designedto include features necessary to meet the requirements of this disk words per interaction." (Scherr, 1965,p. 29). strategy. Special features, known as 'diagnostic "We can and will develop instrumentation which handles', are needed for testing thesystem auto- will be automatically insertedat compile time. A matically, and for providing adequateerror isola- user easily will be able to get a plot of the various tion." (Dent, 1967, p. 100). running times of hisprogram .. " "An instantaneous alarm followed bya quick and Sutherland also refers toa Stanford University correct diagnosis in a self-controlling system will program which "plots the depth of a problem tree limit down-time inmany cases to the mere time of versus time was used to trace the operation of a repair. Instruments for Kalah-playing program." (Sutherland, 1965,pp. error detection are unneces- 12-13). sary." (Steinbuch and Piske, 1963,p. 859). 7.13"The techniques of fault detectionfall 7.11Further, "when a digital system ispar- into two major categories: titioned under certain restrictions into subsystems it is possible to achieve self-diagnosis of thesystem 1. Concurrent diagnosis by the application of through the mutual diagnosis of its subsystems." error-detecting codes and special monitoring (Forbes et al., 1965, p. 1074). circuits. Detection occurs while thesystem is being used. "A diagnostic subsystem is that portion ofa digital system capable of effectively diagnosing another 2. Periodic diagnosis using diagnostic hardware portion of the digital system. It has been shown that and/or programs. Use of thesystem is inter- at least two mutually exclusive diagnost'c subsys- rupted for diagnosis." (Aviiienis, 1967,p. 734). tems are needed inself-diagnosable systems." "The four principal techniques of correctionare: (Forbes et al., 1965, p. 1081). 1. Correction of errors by theuse of error- 7.12"Systems are used to test themselves by correctingcodesandassociatedspecial generation of diagnostic programs using prede- purpose hardware and/or software (including fined data sets and by explicit controls permitting recomputation). 122 2. Replacement of the faulty element or system 7.15"A program testing method based on the by a stand-by spare. monitoring of object-program instruction addresses 3. Replacement as above, with subsequent main- (as opposed to a method dependent on,e.g., the tenance of the replaced part and its return occurrence of particular types of instruction, or to the stand-by state. the use of particular data addresses) would appear 4. Reorganization of the system into a different to be the most suitable, because the instruction fault-free configuration which can continue address is the basic variable of this monitoring the specified task." (Aviiienis, 1967, p. 734). technique. Monitoring could be made 'selective' by specifying instruction addresses at which it is "The STAR (Self-Testing and Repairing) com- to start and stop: to start it at an arbitrary instruc- puter, scheduled to begin experimental operation tion address it is only necessary to replace the at the Jet Propulsion Laboratory of the California instruction located there by the first unconditional Institute of Technology this fall, is expected to be interrupt inserted, and similarly when monitoring one of the first computers with fully automatic self- is to stop and restart later.. .. repair as one of its normal operating functions. .. "Another use in this field would be to include in There are three 'recovery' functions of the STAR the Monitor facilities for simulating any instruction, computer: (1) detection of faults; (2) recognition of anei ro supply it with details of particular instructions temporary malfunctions and of permanent failures; suspected of malfunctioning. The Monitor could and (3) module replacement by power switching. then stop any program just before one of these The occurrence of a fault is detected by applyinginstructions was to be obeyed, simulate it, allow an error-detecting code to all instructions and num- the program to execute the same instruction in bers within the computer. Temporary malfunctions the normal way, and then compare the results are corrected by repeating a part of the program. If obtained by the normal action and by simulation." the fault persists, the faulty module is replaced." (Wetherfield, 1966, p. 165). (Aviiienis, 1968, p. 13). "Of course, having achieved the aim of being 7.14"Diagnostic routines can check the operable to trace in advance the exact course of the ating of a computer for the following possible mal- object program's instructions, the Monitor is then functions: a single continuous malfunction, several able to simulate their actions to any desired degree, continuousmalfunctions, and intermittentmal- and it is here that the power of the technique can be functions. When the test routine finds an error it exploited. The contents of the store, registers, etc., can transfer program control to an appropriate mal- before the execution of any instructioncan be in- function isolation subroutine. This type of diagnostic spected by the Monitor if it temporarily replaces technique, is standard and has been well used by the that instruction by an unconditional interrupt." computer industry for large package replacement." (Wetherfield, 1966, p. 162). (Jacoby, 1959, p. 7-1). "The monitoring operation can go wrong forany "Needless to say, in order for any malfunction of the following three reasons. to be isolated by an automatic program, it is neces- "(1) In the first case one of the planteduncon- sary for a minimum amount of equipment to func- ditional interrupt instructions actually overwrites tion adequately. One of the functions of this mini- the instruction at which the object program is going mum equipment includes the ability to sequence to resume (the one at which monitoring started). from one instruction to another, and to be able to This would effectively bring things toa standstill interpret (correctly) and execute at least one trans- since the situation will recur indefinitely. If the rules fer of control instruction so that logical choices can above have been followed, this situationcan only be made. The control functions of a computer can arise when a branching instruction includes itself be defined as Boolean algebraic expressions of the among its possible destinations, i.e., there is a po- instantaneous state of the computer. If we state tential loop stop in the objectprogram. in order to that a line, or path, common to two control state- cope with this situation, if it could occur, it may be ments contains those components that are acti- necessary for the Monitor to simulate the action vated when either of the statements is true, this line of the branch instruction completely and make the is either a factor of both statements or a factor of object program bypass it. The loop stop might still terms of both statements. Similarly, if we consider occur, but it would be foreseen. circuit elements activated by one but not both of "(2) The second possible reason for a failure of two ways to accomplish the same control function, the monitoring operation occurs ifone of the planted we have a picture of two terms in the algebraic instructions overwrites part of the data of the object statement for the control function separated by program, thus affecting the latter's behaviour. This the connector OR. `data' might be a genuine instruction which As "A Boolean term will appear as a circuit which examined, as well as obeyed, by the object program. must be active for any statement, of which it is a Alternativelyitmight be genuine data which factor, to be true. Hence the location of circuit happens to be stored in a position which is, by malfunctions may be considered from the point of accident or design, a 'redundant' destination ofa view of isolating the minimal Boolean term in- branching instruction. Both of these dangerscan be volved." (Jacoby, 1959, p. 7-1). anticipated by the Monitor, at the cost ofa more 123

376-411 0 - 70 - 9 detailed examination of instructions (to find out designer and the ingenuity of the user. Digital which store references by the object program simulation can expedite the analysis of a complex involve a replaced instruction location) and more system under various stimuli if the aggregate can frequent interrupts. be divided into elements whose performance can be "The situation savours of 'trick' programming. It suitably described. If the smallest elements into is apparent that the monitoring process will be which we can divide a system are themselves un- simplified if there is some guarantee that thesepredictable (even in a probabilistic sense) digital oddities are absent from object programs." (Wether- simulation is not feasible. (Conway, et al., 1959, ,field, 1966, pp. 162-163). p. 94). This feasibility test uncovers an important 7.16"MAID (Monroe Automatic Internal limitation in today's simulation technology with Diagnosis) is a program that tells a machine how respect to information systems. In many respect., to measure itscircuitry and ttest performance some of the more important man-information-system onsample problemscomputei. hypochondria." interactions cannot now be described in a formal (Whiteman, 1966, p. 67). manner;hence,cannotbecharacterizedfor 7.17"I have used a program which interprets digital simulation. For example, one can calculate the program under test and makes a plot of the the speed and costs of processing an inquiry, but memory address of the instruction being executed cannot predict if the output will satisfy the user or versus time. Such a plot shows the time the program estimate its impact on his operations. spends doing its various jobs. In one case, it showed "This limitation, therefore, (1) restricts simulation me an error which caused a loss of time in a program applications to examining the more mechanical whichnevertheless gave correct answers.. .. aspects of data processing, or (2) forces the design "Think of the thousands of dollars saved by engineer to adopt some simplifying assumptions tightening up that one most-used program loop. concerning the effects of man's influence on the Instrumentation can identify which loop is the system. An example of the first point is a data most used." (Sutherland, 1965, pp. 12-13). flow simulation examining the rate of data proc- 7.18"On-Line Instrumentation will bring us essing without regard to the quality of the types better understanding of the interplay of the pro- and mixes of equipment and personnel- This capa- grams and data within the computer. Simple devices bility for examining the resultant effects in varying and programs to keep track, on-line, of what the parameters of the system enable the design engineer computer does will bring us better understanding to explore more alternatives in less time and at less of what our information reprocessing systems cost than ever before; e.g., he can develop cost- are actually doing." (Sutherland, 1965, p. 9). capabilitycurvesfordifferentpossible system 7.19"The process of building a pilot system configurations under both present and anticipated configuration and then evaluating it, modifying it, processing needs. Neglecting this aspect of systems and improving it is very costly both in time and analysis has sometimes led to the implementation money. Another approach is possible. Before he of a system saturated by later requirements and builds the system, the designer should be able confronted by an unnecessary high cost for modifica- to test his concepts on a simulation model of a tion or replacement." (Blunt et al., 1967, pp. 75-76). document retrieval system. One such model for "To use simulation techniques inevaluating simulatinginformationstorageandretrieval different computer systems, one must be able to systems was designed by Blunt and his co-workers specify formally the expected job mix and con- at HRB-Singer, Inc., under a contract with the straints under which the simulated system must Officeof Naval Research. Inthismodel, the operate, e.g., operating time per week. Equally im- input parameters for the simulation reflects the portant, one must carefully select a set of char- configuration of the system, the work schedule of the acteristics on which the competing systems will be system, thework scheduleof thepersonnel, judged. For different installations the most im- equipment availability, the likelihood and effect portant characteristics may well be different. Each of errors in processing and the location and availa- system under consideration is modelled, simulation bility of the system user. Simulation output pro- runs are executed, and the results are compared on vides a study of system response time (both delay the selected characteristics. time and processing time), equipment and personnel "Unfortunately, the ideal case seldom occurs. work and idle time and the location and size of Often the available information about the computer the data queues. The systems designer can thus vary system's expected job mix is very limited. Further- the inputs, use the model to simulate the inter- more, it is a well-known fact that an installation's actions among personnel, equipment, and data at job mix itself may be strongly influenced both each step of the information processing cycle, qualitatively and quantitatively by the proposed and then determine the effect on the system re- changes in the system. For example, many of the sponse time." (Borko, 1967, p. 55). difficulties with early time-sharing systems can be 7.20"Simulation is a tool for investigation and, attributed to the changes in user practices caused like any tool, is limited to its inherent potential. by the introduction of the system. When statistics Moreover, the realization of this potential is depend- on job mix are available, they are often expressed in ent upon economics, the craftsmanship of the averages. Yet, it may be most important to simulate 124 a systera's performance under extreme conditions. in the real world do not lend themselves to neat Finally, it is often difficult to show that a simulation mathematical formulations and inmost cases is valid that is, that it actually does simulate the the operations analyst is forced to reduce the system in question." (Huesmann and Goldberg, problem to simpler terms to make it tractable." 1967, p. 150). (Clapp, 1967, p. 5). 7.21"The field of information retrieval has been "Admittedly the degree to which identifiable marked by a paucity of mathematical models, and factors can be measuredcompared to the in- the basis of present operational computer retrieval fluence of unidentifiable factorsdoes help deter- systems is essentially heuristic in design." (Baker, mine whether or not an approach can be scientific. 1965, p. 150). It acts as a limit on the area where scientfic methods "The semantic and linguistic aspects of infor- can be applied. Precision in model building is mation retrieval systems also lend themselves relating to the difficulty of the problem and the poorly to the rigidity of models and model tech- stateof human knowledge concerning specific niques,for which claim:,often lack empirical techniques andtheirapplication." (Kozmetsky support." (Blunt, 1965, p. 105). and Kircher, 1956, p. 137). 7.22"There are structures which can easily 7.23"Thereisno guarantee that a model be defined but which present-day mathematics such as latent class analysis, factor analysis,or cannot handle because of the limitations of present- anything else borrowed from another fieldwill day theory." (Hayes, 1963,p. 284). meet the needs of its new context; however this "Markov models cannot, in general, be used should not dissuade one from investigating such to represent processes where other than random plausible models." (Baker, 1965,p. 150). queuing is used." (Scherr, 1965, p. 32). "Models must be used but must never be believed. "Clearly, we need some mathematical models As T. C. Chamberlain said, 'science is the holding permitting the derivation of methods which will of multiple working hypotheses'." (Tukey and Wilk, accomplish the desiredresults and for which 1966, p. 697). criteria of effectiveness can be determined. Such 7.24"System simulation or modeling was subse- models do not appear often in the literature." quently proposed as a substitute for deriving test (Bryant, 1964, p. 504). problems and is still generally accepted as sucheven "First, it will be necessary to construct mathe- though its use introduced the new difficulty of deter- matical models of systems in which content, struc- mining and building meaningful models." (Davis, ture, communication, and decision variables all 1965, p. 82). appear. For example, several cost variables are "The biggest problem in simulation modeling,as usually included in a typical operations research in all model building, is to retain all 'essential' model. These are either takenas uncontrollable detailand remove thenonessential features." or as controllable only by manipulating such other (Scherr, 1965, p.09). variables as quantity purchased or produced, time "The fundamental problem in simulation of digital of purchase or production, number and type of networks is that of economically constructinga facilities, and allocation of jobs to these facilities. mathematical model capable of faithfully replicating These costs, however, are always dependenton the real network's behavior in regard to simulation human performance, but the relevant variables objectives." (Larsen and Mano, 1965, p. 308). dealing with personnel, structure, and communica- 7.25"At this time, there exists no special- tion seldom appear in such models. To a large purpose simulation programming language specif- extent this is due to the lack of operational defini- ically for use with models of digital computer tions of many of these variables and, consequently, systems. The general-purposes languages, such as to the absence of suitable measures in terms of SIMSCRIPT, GPSS, etc.,all have faults which which they can be characterized." (Ackoff, 1961, render them unsuitable for this type of work." p. 38). (Scherr, 1965, p. 43). "Mathematicalanalysisof complex systems "Theinventionofanadequatesimulation isvery often impossible; experimentation with language promises to be the crowbar needed to actual or pilot systems is costly and timeconsum- bring the programming of operating systems to ing, and the relevant variables are not always the level of sophistication of algebraicor commercial subject to control.... programming languages." (Perlis, 1965, p.192). "The technique of input simulation... canbe "Simulation problems are characterized by being very expensive. The programs necessary to create mathematicallyintractable and having resisted the simulated inputs are far from trivial andmay solutionbyanalyticmethods.The problems well constitute a second system larger than the usually involve many variables,many parameters, operational system." (Steel, 1965, p. 232). functionswhicharenotwell-behavedmathe- 7.26"Those programs which require the simu- matically, and random variables. Thus, simula- lated computer system and job mix to be specified tion is a technique of last resort." (Teichroew and in algebraic or assembly languages have proved Lubin, 1966, p. 724). useful; but as general computer systems simulation "The complex systems generally encountered tools, they require too much difficult recording to be 125 completely satisfactory. One way to improve upon on main-storage usage. Studies ofthis nature can this situation has been to use languages specifically become very time consuming unless parameter designed to simulate systems. Teichroew and Lubin selections and variations are carefully limited. It in a recent review have listed more than twenty is no small problem to determine which are the languages, among them GPSS, SIMSCRIPT, SOL, major variations that affect the system. In this and CSL. These simulation languages allow the aspect, simulation is not as convenient as algo- modeller to specify the computer configuration rithmic methods with which many variations can and jobmix ina more convenient manner." be tabulated quickly and cheaply." (Seaman, 1966, (Huesmann, and Goldberg, 1967, p. 152). p. 177). 7.27"One of the more exotic applications of 7.28"The [SIMSCRIPT] notation is an aug- digital computers is to simulate a digital computer mented version of FORTRAN, which is acceptable; on another entirely different type of computer.Using but this organization does not take advantage of the a simulation program, application programs de- modularity of digital systems. veloped for the first computer, the source computer, "SIMSCRIPT is an event-based language. That may be executed on a second computer, the object is, the simulation is described, event by event, with computer. small programs, one per event. Each event program "Simulation obviously provides many advantages (or sub-program) must specify the times for the in situations where a computer is replaced by a events following it. Conditional schedulingof an different computer, for which the applications have event is extremely difficult." (Scherr, 1965, p.43). not yet been programmed. Simulation techniques 7.29Ewards points out that "...the prepara- enable an installation to continue solving problems tion of so-called scenarios, or sequences of events using existing programs after the new computer to occur as inputs to the simulation, is amajor has been installed and the old one removed.. .. problem, perhaps the most important one, in the "Another situation in which simulation is advan- design of simulations, especially simulations of tageous is during the development of a new com- information-processing systems." (Edwards, 1965, puter. Once specifications for the new computer p. 152). have been established, programming of applications 730"Parallelprocessescanberendered for the computer can proceed in parallel with hard- sequential, for simulation purpose; difficulties then ware developments. The use of a simulator inthis arise when the processes influence each other, situation enables the users to debug their applica- leadingperhapstoincompatibilitiesbarring a tions before the hardware is actually available." simultaneous development. Difficulties of this type (Trimble, 1965, p. 18). cannot be avoided, as a matter of principle,and the "One of the most successful applications of the system is thus not deterministic; the only way out recent microprogramming technologyisinthe would be to restore determinism through recourse to simulation of computers on computers. appropriate priority rules. This approach is justified "The microprogramcontrol and thesetof only if it reflects priorities actually inherent in the microprogram routines are in effect a simulation system." (Caracciolo di Forino, 1965, p. 18). program that simulates the programmer's instruc- 7.31"As a programming language, apart from tion set on a computer whose instruction set is simulation, SIMULA has extensive list processing the set of elementary operations. It may be equally facilities and introduces an extended co-routine possible to simulate computers with other' pro- conceptina high-level language."(Dahl and grammer instruction sets in terms of the same set Nygaard, 1966, p. 671). of elementary .operations. This, slightly oversimpli- fied perhaps, isthe idea of hardware assisted 7.32"The LOMUSS model of the Lockheed simulation that is now usually called emulation." UNIVAC on-line,time-sharing, remote terminal (Rosen, 1968, p. 1444). system simulated two critical periods...and "As a result of simulation's ability to deal with provided information upon which the design of the many details, it is a good tool for studyingextensive 1108 system configuration was based. An effort is and complicated computer systems. With simula- continuing which will monitor the level and char- tion, one may assess the interaction of several sub- acteristics of the workload, equipment utilization, systems, the performances of which are, modified turnaround time, etc., for further model validation." by internal feedback loops among the subsystems. (Hutchinson and Maguire, 1965, pp. 166-167). For instance, in a teleprocessing system where pro- "A digital computer is being used to simulate the grams are being read from drum storageand held logic, determine parts values, compute subunit temporarily in main storage, the number of mes- loading, write wiring lists, design logic boards, print sages in the processing unit depends uponthe drum checkout charts and maintenance charts. Simulating response time, which depends uponthe drum access the logic and computing the loading of subunits rate, which, in turn, depends upon the number of gives assurance that a computer design will function messages in main storage. In this case,only a properly before the fabrication starts. After the logic system-wide simulation that includes communica- equations are simulated, it is a matter of hours until tion lines, processing unit,, and I/O subsystems will all fabrication information and checkout information determine the impact of varying program priorities is available. Examples are given of the use of these 126 techniques on the design and fabrication of a large use of COMPAX, has been written for CDC 3600 scale military computer." (Malbrain, 1959, p. 4-1). systematthe Tata Instituteof Fundamental "The new EDP Machine Logic and Control Research, Bombay." (Narasimhan, 1966, p. 171). Simulator, LOCS, is designed to facilitate the simulation of data processing systems and logic on the At IBM, there has been developed a computer programforimage-forming systemssimulation IBM 7090 Data Processing System...The inputs of LOCS consist of a description of the machine to (IMSIM/1), so that the photo-optical design engineer be simulated, coded in LOCS language, and a set of can study performance before such systemS are test programs coded in either the procedure lan- actually built (Paris, 1966). guage of the test problems (e.g., FORTRAN) or in 7.37"Thefirstmodeldevelopedmatches the instruction language of the simulated ma- CTSS.. .Next, a simple, first-come, first-served chine...The outputs of LOCS consist of the per- round-robin scheduling procedurewill be sub- formancestatistics,computationresults,and stituted. Then, a model which incorporates multi- diagnostic data which are relevant to both the test programming techniques with the CTSS hardware programs and the design of the simulated machine." configuration will be developed. Finally, a simple (Zucker, 1965, pp. 403-404). continuous-time Markov model will be used to represent both single-processor and multiple-processor 7.33". ..A system of software and hardware time-shared systems.... which simulates, with a few exceptions, a multiple set of IBM System/360 computing systems (hard- "The primary result obtained is that it is possible ware and software) that are simultaneously available to successfully model users of interactive computer to many users." (Lindquist et al., 1966, p. 1775). systems and systems themselves with a good degree of accuracy using relatively simple models." (Scherr, 7.34"Another simulation program designed to simulate multiprocessor systems is being developed 1965, p. 31). by R. Goldstein at Lawrence Radiation Laboratory. "The final model to be simulated will represent a Written in SIMTRAN, this program is specifically system in which swapping and processor operation designed tosimulate the OCTOPUS computer are overlapped. While a program is being run by systemat LRL which includes an IBM 7030 the processor, the program which was running STRETCH, two IBM 7094's, two CDC 6600's, one previously is dumped and the next program to run CDC 3600, two PDP 6's, an IBM 1401, and various is loaded. Since loading and dumping cannot occur I/O devices. A parameterized input table specifies simultaneously, there must be room in the core the general multiprocessing configuration, the data memory for at least two complete user programs transmission rates, memory sizes and buffer sizes. the program being dumped or loaded. Should two Any other hardware variations and operating system programs intended to run in sequence not fit together characteristicsareintroducedwith SIMTRAN in the core memory, the processor must be stopped routines. As output the program produces figures on to complete the swapping." (Scherr, 1965, pp. actual memory utiliiation, graphs of memory access 40-41). time, graphs of overhead, graphs of response time, and graphs of several other relevant variables." Other examples of experiments in computer (Huesmann and Goldberg, 1967, p. 153). simulationof multipleaccess and time-shared systems include the following: "The development of 7.35"One of the more exotic applications of the simulation program now provides a first-come- digital computers is to simulate a digital computer first-serve queue unloading strategy. Continuing on another entirely different type of computer. Using effort, however, will provide for optional strategies, a simulation program, application programs devel- e.g., selecting the data unit with the shortest servic- oped for the first computer, the source computer, ing time, consideration of what flow will minimize may be executed on a second computer, the object idle time at the central processor etc." (Blunt, 1965, computer." (Trimble, 1965, p. 18). p. 15). 7.36"The flexibility of a digital computer enables "Project MAC is using a display to show the one totry out complicated pictureprocessing dynamicactivity of jobs withinitsscheduling schemes with a relatively small amount of effort. algorithm. Watching this display, one can see jobs To facilitate this simulation, a digital picture scanner moving to higher and lower priority queues as time and cathode-ray tube display was constructed. passes." (Sutherland, 1965, p. 13). Pictures were scanned with this system, the signal was recorded on a computer magnetic tape, and this 7.38"A great advance also is the wide applica-' tape was used as input to a program that simulated tion of digital computers for the simulation of a picture-transmitting system." (Huang and Tretiak, recognizing and adaptive systems. Digital simula- 1965, pp. 45-46). tionextraordinarilyfacilitatesandaccelerates coaducting experiments in this realm by permitting "Computer simulation of a letter generator using extremelyeffectiveexperimentalinvestigations the above grammar isa comparatively straight- without expenditure of materials and with little forward programming task. Such a program, making time spent." (Kovalevsky, 1965, p. 42), 127 "Third, computer simulation may serveas a Moreover, a program which solves problems is by heuristic in the search fer models. The effort of that sole virtue a candidate for a model and deserves getting a computer to perfoi An a given task may lead thepsychologists'attention.Afterall,proving to illuminating p.9ychological hypotheses, even iftheorems or recognizing patterns was until recently no behavioral evidence has been taken into account. uniquely human or animal." (F'rijda, 1967, p. 59).

128 Appendix B. Bibliography

Abbas, S. A., H. F, Koehler, T. C. Kwei, H. 0. Leilich and Computer Conf., Vol. 31, Anaheim, Calif., Nov. 14-16, 1967, R. H. Robinson, Design Considerations for the Chain Magnetic pp. 409-411 (Thompson Books, Washington, D.C., 1967). Storage Array, IBM J. Res. & Dev. 11, 302-311 (May 1967). Aron, J. D. Real-Time Systems in Perspective, IBM Sys. J. 6, Abraham, C., G. N. Lance and T. Pearcey, Multiplexing of Slow 494,7 (1467). Peripherals, Commun. ACM 9, No. 12, 877-878 (Dec. 1966). Aron, J. D., Information Systems in Perspective, in IBM Proc. Abrahams, P. W., J. A. Barnett, E, Book, D. Firth, S. L. Kameny, Inf. Systems Symp., Washington, D.C., Sept. 4-6, 1968, C. Weissman, L. Hawkinson, M. I. Levin and R. A. Saunders, pp. 3-37 (IBM Corp., 1968). The LISP 2 Programming Language and System, AFIPS Proc. Arora, B. M., D. L. Bitzer, H. G. Slottow and R. H. Wilson, The Fall Joint Computer Conf., Vol. 29, San Francisco, Calif., Plasma Display Panel- A New Device for Information Display Nov. 7-10, 1966, pp. 661-676 (Spartan Books, Washington, and Storage, Rept. No. R-346, 24 p. (Coordinated Science D.C., 1966). Laboratory, Univ. of Illinois, Urbana, DI., April 1967). Ackoff, R. L., Systems, Organizations, and Interdisciplinary Aschenbrenner, R. A., M. J. Flynn and G. A. Robinson, Intrinsic Research, in Systems: Research and Design, Proc. First Multiprocessing, AFIPS Proc. Spring Joint Computer Conf., Systems Symp. at Case Institute of Technology, Cleveland, 0., Vol. 30, Atlantic City, N.J., Apr. 18-20, 1967, pp. 81-86 Apr. 1960, Ed. D. P. Eckman, pp. 26-42 (Wiley, New York, (Thompson Books, Washington, D.C., 1967). 1961). Asendorf, R. H., The Remote Reconnaissance of Extraterrestrial Adams, C. W., Responsive Time-Shared Computing in Business - Environments, in Pictorial Pattern Recognition.. Proc. Symp. Its Significance and Implications, AFIPS Proc. Fall Joint on Automatic Photointerpretation, Washington, D.C., May 31- Computer Conf., Vol. 27, Pt. 1, Las Vegas, Nev., Nov. 30- June 2, 1967, Ed. G. C. Cheng et al., pp. 223-238 (Thompson Dec. 1, 1965, pp. 483-488 (Spartan Books, Washington, D.C., Book Co., Washington, D.C., 1968). 1965). Ash, W. L. and E. H. Sibley, TRAMP: An Interpretive Associa- Aines, A. A., The Convergence of the Spearheads, in Toward a tive Processor with Deductive Capabilities, Proc. 23rd National NationalInformationSystem,SecondAnnualNational Conf. ACM, Las Vegas, Nev., Aug. 27-29, 1968, pp. 143-156 Colloquium on InformationRetrieval,Philadelphia,Pa., (Brandon/Systems Press, Inc., Princeton, N.J., 1968). April 23-24, 1965, Ed. M. Rubinoff, pp. 3-13 (Spartan Books, Auerbach Corporation, Source Data Automation, Final Rept. Washington, D.C., 1965). 1392-200-TR-1, 1 v. (Philadelphia, Pa., Jan. 16, 1967). Alberga, C. N., String Similarity and Misspellings, Commun. Austin, C. J., Dissemination of Information and Problems of ACM 10, No. 5, 302-313 (May 1967). Graphic Presentation, Proc. 1965 Congress F.I.D. 31st Meeting Alt, F. L.The Standardization of Programming Languages, and Congress, Vol. II, Washington, D.C., Oct. 7-16, 1965, Proc. 19th National Conf., ACM, Philadelphia, Pa., Aug. 25-27, pp. 241-245 (Spartan Books, Washington, D.C., 1966). 1964, pp. B 2-1 to B 2-6 (Assoc. for Computing Machinery, Aviiienis, A., Design of Fault-Tolerant Computers, AFIPS Proc. 1964). Fall Joint Computer Conf., Vol. 31, Anaheim, Calif., Nov. Alt, F. L. and M. Rubinoff, Eds., Advances in Computers, Vol. 7, 14-16, 1967, pp. 733-743 (Thompson Books, Washington, 303 p. (Academic Press, New York, 1966). D.C., 1967). Amdahl, G. M., Multi-Computers Applied to On-Line Systems, Aviiienis, A., A Digital Computer with Automatic Self-Repair in On-Line Computing Systems, Proc. Symp. sponsored by and Majority Vote, Computers & Automation 17, No. 5, 13 the Univ. of California, Los Angeles, and Informatics, Inc., (May 1968). Los Angeles, Calif., Feb. 2-4, 1965, Ed. E. Burgess, pp. 38-42 Avram, H. D., R. S. Freitag and K. D. Guiles, A Proposed Format (American Data Processing, Inc., Detroit, Mich. 1965). for a Standardize(' Machine-Readable Catalog Record, ISS Ammon, G. J. and C. Neitzert, An Experimental 65-Nanosecond Planning Memo. ,Preliminary Draft, 110 p. (Library of Thin Film Scratchpad Memory System, AFIPS Proc. Fall Congress, Washingwii, D.C., June 1965). Joint Computer Conf., Vol. 27, Pt. 1, Las Vegas, Nev., Nov. 30- Dec. r, 1965, pp. 649-659 (Spartan Books, Washington, D.C., Baker, C. E.' and A. D. Rugari, A Large-Screen Real-Time Display 1965). Technique, Inf. Display 3, No. 2, 37-46 (Mar./Apr. 1966). Anacker, W., G. F. Bland, P. Pleshko and P. E. Stuckert, On the Baker, F. B., Latent Class Analysis as an Association Model for Design and Performance of a Small 60-Nsec Destructive Read- Information Retrieval, in Statistical Association Methods for out Magnetic Film Memory, IBM J. Res. & Dev. 10, No. 1, Mechanized Documentation, Symp. Proc. Washington, D.C., 41-50 (Jan. 1966). March 17-19, 1964, NBS Misc. Pub. 269, Ed. M. E. Stevens Anderson, R., E. Marden and B. Marron, File Organization for a et al., pp. 149-155 (U.S. Govt. Print. Off. Washington, D.C., Large Chemical Information System, NBS Tech. Noto 285, Dec. 15, 1965). 17 p. (U.S. Govt. Print. Off., Washington, D.C., Apr. 1966). Baker, F. T. and W. E. Triest, Advanced Computer Organization, Andrews, D. H., An Array Processor Using Large Scale Integra. Rept. No. RADC- -TR -66 -148, 1 v. (Rome Air Development tion, Computer Design 8, No. 1, 34-43 (Jan. 1969). Center, Griffiss Air Force Base, N.Y., May 1966). Andrews, M. C., Multifont Print Recognition, in Optical Character Baker, J. D., From the Diet of Worms to the Bucket of Worms: Recognition, Ed. G. L. Fischer, Jr., et al., pp. 287-304 (Spartan A Protest Concerning Existing Display Dogma for Information Systems, in Second Congress on the Information System Books, Washington, D.C., 1962). Sciences, The Homestead, Hot Springs, Va., Nov. 1964, Applebaum, E. L., implications of the National Register of Micro- Ed. J. Spiegel and D. E. Walker pp. 429-433 (Spartan Books, film Masters, as Part of a National Preservation Program, Lib. Washington, D.C., 1965). Res. & Tech. Serv. 9, 489-494 (1965). Baker, J. U. and W. A. Kane, Make EDP Controls Work for Yr- -I, Armstrong, J: A., Fresnel Holograms: Their Imaging Properties in Data Processing, Vol. X, Proc. 1966 International Data and Aberrations, IBM J. Res. & Dev. 9, No. 3, 171-178 (May Processing Conf., Chicago, DI., June 21 -24, 1966, pp. 97-100 1965). (Data Processing Management Assoc. 1966). Armstrong, R., H. Conrad, P. Ferraiolo and P. Webb, Systems Balzer, R. M., EXDAMS-EXtendable Debugging and Monitoring Recovery from Main Frame Errors, AFIPS Proc. Fall Joint System, AFIPS Proc. Spring Joint Computer Conf., Vol. 34, 129 Boston, Mass., May 14-16, 1969, pp. 567-580 (AFIPS Press, Bialer, M., A. A. Hastbacka and T. J. Matcovich, Chips Are Montvale, N.J., 1969). Down in New Way to Build Large Microsystems, Electronics Baran,P., On DistributedCommunications:IV.Priority, 38, No. 20, 102-109 (Oct. 4, 1965). Precedence, and Overload, Rept. No. RM-3638-PR, 63 p, Bilous, 0., I. Feinberg and J. L. Langdon, Design of Monolithic (The RAND Corp., Santa Monica, Calif., Aug. 1964). Circuit Chips, IBM J. Res. & Dev. 10, No. 5, 370-376 (Sept. Baran, P., On Distributed Communications: V. History, Alterna- 1966). tive Approaches, and Comparisons, Rept. No. RM-3097-PR, Birch, R. L., Packaging, Labeling, and Finding Evaluated Tech- 51 p. (The RAND Corp., Santa Monica, Calif., Aug. 1964). nical Data, in Data/Information Availability, Ed. R. I. Cole, Baran, P., On Distributed Communications: VII. Tentative pp. 163-175 (Thompson Book Co., Washington, D.C., 1966). Engineering Specifications and Preliminary Design for a High- Birmingham, D. J., Planning for Data Communications, Data Data-Rate Distributed Network Switching Node, Rept. No. Proc. Mag. 6, No. 10, 36-38 (Oct. 1964). RM-3763-PR, 85 p. (The RAND Corp., Santa Monica, Calif., Bittman, E. E., A 16k-Word, 2-MC Magnetic Thin-Film Memory, Aug. 1964). AFIPS Proc. Fall Joint Computer Conf., Vol. 26, San Francisco, Baran,P., On Distributed Communications:IX.Security, Calif., Oct. 1964, pp. 93-106 (Spartan Books, Baltimore, Md., Secrecy, and Tamper-Free Considerations, Rept. No. RM- 1964). 3765-PR, 39 p. (The RAND Corp., Santa Monica, Calif., Bleier, R. E. and A. H. Vorhaus, File Organization in the SDC Aug. 1964). Time-Shared Data Management System (TDMS), Proc. IFIP Baran, P., On Distributed Communications: VIII. The Multi- Congress 68, Edinburgh, Scotland, Aug. 5-10, 1968 Booklet F, plexing Station, Rept. No. RM-3764-PR, 103 p. (The RAND pp. F 92-F 97 (North-Holland Pub. Co., Amsterdam, 1968).

Corp., Santa Monica, Calif., Aug. 1964). Bloembergen, N., New Horizons in Quantum Electronics., IEEE Barnum, A. R., Reliability Central Data Management System, Spectrum 4, 83-86 (July 1967). in Toward a National Information System, Second Annual Bloom, A. L., Gas Lasers, Proc. IEEE 54, 1262-1276 (Oct. 1966). National Colloquium on Information Retrieval, Philadelphia, Blunt, C. R., An Information Retrieval System Model, Rept. No. Pa., April 23-24, 1965, Ed. M. Rubinoff, pp. 45-61 (Spartan 352. 14-R-1, 150 p. (HRB-Singer, Inc., State College, Pa., Books, Washington, D.C., 1965). Oct. 1965). Barton, R. S., A Critical Review of the State of the Programming Blunt, C., R. Duquet and P. Lnekie, Simulation of Information Art, AFIPS Proc. Spring Joint Computer Conf., Vol. 23, Detroit, Systems -Some Advantages, Limitations and An Example of a Mich., May 1963, pp. 169-177 (Spartan Books, Baltimore, Md., General Information Systems Simulator, in Levels of Inter- 1963). action Between Man and Information, Proc. Am. Doc. Inst. Baruch, J. J., A Medical Information System: Some General Annual Meeting, Vol. 4, New York, N.Y., Oct. 22-27, 1967. Observations, in Information System Science and Technology, pp. 75-79 (Thompson Book Co., Washington, D.C., 1967). papers prepared for the Third Cong., Scheduled for Nov. 21-22, Bobrow, D. G., Problems in Natural Language Communication 1966, Ed. D. E. Walker, pp. 145-150 (Thompson Book Co., with Computers, Rept. No. Scientific-5, BBN-1439, AFCRL Washington, D.C., 1967). 66-620, 19 p. (Bolt, Beranek and Newman, Inc., Cambridge, Bauer,/ W. F., On-Line Systems -Their Characteristics and Mass., Aug. 1966). Motl,vations, in On-Line Computing Systems, Proc. Symp. Bonin, E. L. and J. R. Baird, What's New in Semiconductor Los/ Angeles, Calif., Feb. 2-4, 1965, Ed. E. Burgess, pp. 14-24 Emitters and Sensors, Electronics 38, 98-104 (Nov. 1965). (American Data Processing, Inc., Detroit, Mich., 1%5). Bonn, T. H., Mass Storage; A Broad Review, Proc. IEEE 54, Becker, C. H., UNICON Computer Mass Memory System, 1861-1870 (Dec. 1966). AMPS Proc. Fall Joint Computer Conf., Vol. 29, San Francisco, Borko, H., Design of Information Systems and Services, in Calif., Nov. 7-10, 1966, pp. 711-716 (Spartan Books, Washing- Annual Review of Information Science and Technology, Vol. 2, ton, D.C., 1966). Ed. C. A. Cuadra, pp. 35-61 (Interscience Pub., New York, Becker, J. and R. M. Hayes, Information Storage and Retrieval: 1967). Tools, Elements, Theories, 448 p. (John Wiley & Sons, New Borko, H., Ed., Automated Language Processing, 386 p. (Wiley, York, N.Y., 1963). New York, 1967). Berner, R. W. and A. L. Ellison, Software Instrumentation Sys- Bowers, D. M., W. T. Lennon, Jr., W. F. Jordan, Jr., and D. G. tems for Optimum Performance, Proc. IFIP Congress 68, Benson, TELLERTRON - A Real-Time Updating and Trans- Edinburgh, Scotland, Aug. 5-10, 1968, Booklet C, pp. C 39- action Processing System for Savings Banks, 1962, IRE Int. C 42 (North-Holland Pub. Co., Amsterdam, 1968). Cony. Rec., Pt. 4, pp. 101-113. Bennett, E., Flexibility of Military Information Systems, in Mili- Boyd, R., Monitor Controlled Computer Processing of Bookwork, tary Information Systems -The Design of Computer-Aided in Institute of Printing, Computer Typesetting Conf., Rept. of Systems for Command, Ed. E. Bennett et al., pp. 88-109 Proc., London Univ., July 1964, pp. 152, 153, 157 (Pub. London, (Frederick A. Praeger, Pub., N.Y., 1964). 1965). Brick, D. B. and G. G. Pick, Microsecond Word-Recognition Bennett, E., J. Degan and J. Spiegel, Eds., Military Information Systems -The Design of Computer-Aided Systems for Com- System, IEEE Trans. Electron. Computers EC-13, 57-59 i (Feb. 1964). ' mand, 180 p. (Frederick A. Praeger, Pub., N.Y., 1964). Briley, B. E., Picoprogramming: A New Approach to Internal Bennett, E., E. C. Haines and J. K. Summers, AESOP: A Proto- ComputerControl,AFIPSProc.FallJointComputer type for On-Line User Control of Organizational Data Storage, Conf., Vol. 27, Pt. 1, Las Vegas, Nev., Nov. 30 -Dec. 1, 1965, Retrieval and Processing, AFIPS Proc. Fall Joint Computer pp. 93-98 (Spartan Books, Washington, D.C., 1965). Conf., Vol. 27, Pt. 1, Las Vegas, Nev., Nov. 30 -Dec. 1, 1965, Brookner, E., M. Kolker and R. M. Wilmotte, Deep-Space Optical pp. 435-455 (Spartan Books, Washington, D.C., 1965). Communications, IEEE Spectrum 4, 75-82 (Jan. 1967). Bennett, S. J. and J. W. C. Gates, Holography of Diffusely Re- Brooks, F. P., Jr., The Future of Computer Architecture, in Information Processing 1965, Proc. IFIP Congress 65, Vol. 1, " fleeting Objects Using a Double Focus Lens, Nature 221, 1234-1235 (Mar. 29, 1969). New York, N.Y., May 24-29, 1965, Ed. W. A. Kalenich, pp. 87-91 (Spartan Books, Washington, D.C., 1965). Bernstein, M. I. and T. G. Williams, An Interactive Programming System for the Casual User, Computers and Automation 17, Brown, B. R. and A. W. Lohmann, Computer-Generated Binary 26-29 (Feb. 1968). Holograms, IBM J. Res. & Dev. 13, No. 2, 160-168 (Mar. 1969). Brown, G. W., J. G. Miller and T. A. Keenan, Eds., EDUNET- Bernstein, M. I. and T. G. Williams, A Two Dimensional Pro- Report of the Summer Study on Information Networks Con- gramming System, Proc.IFIP Congress 68, Edinburgh, ductedbytheInteruniversityCommunications Council Scotland, Aug. 5-10, 1968, Booklet C, pp. C 84-C 89 (North- (EDUCOM), 440 p. (Wiley, New York, 1967). Holland Pub. Co., Amsterdam, 1968). Brown, R. M., An Experimental Study of an On-Line Man- Bernstein, W. A. and J. T. Owens, Debugging in a Time-Sharing Computer System, IEEE Trans. Electron. Computers EC-14, Environment, AFIPS Proc. Fall Joint Computer Conf., Vol. 33, 82-85 (1965). Pt. 1, San Francisco, Calif., Dec. 9-11, 1968, pp. 7-14 (Thomp- Bryant, E. C., Redirection of Research into Associative Retrieval, son Book Co., Washington, D.C., 1968). in Parameters of Information Science, Proc. Am. Doc. Inst. 130 Annual Meeting, Vol. 1, Philadelphia, Pa., Oct. 5-8, 1964, Catt, I., E. C. Garth and D. E. Murray, A High-Speed Integrated pp. 503-505 (Spartan Books, Washington, D.C., 1964). Circuit Scratchpad Memory, AFIPS Proc. Fall Joint Computer Bucci, W.,-PCM: A Global Scramble for Systems Compatibility, Conf., Vol. 29, San Francisco, Calif., Nov. 7-10, 1966, pp. Electronics 42, No. 13, 94-102 (June 23, 1969). 315-33,1 (Spartan Books, Washington, D.C., 1966). Buck, C. P., R. F. Pray, III and G. W. Walsh, Investigation and Chang, J. T., J. F. Dillon, Jr. and U. F. Gianola, Magneto-Optical Study of Graphic-Semantic Composing Techniques, Rept. Variable Memory Based upon the Properties of a Transparent No. RADC-TR-61-58, Final Rept. Contract AF 30(602)2091, Ferrimagnetic Garnet at its Compensation Temperature, Proc. v, (Syracuse Univ., Research Inst., June 1961). 10th Conf. on Magnetism and Magnetic Materials, Minneapolis, Buckland, L. F., Machine Recording of Textual Information Minn., Nov. 16-19, 1964, Ed. I. S. Jacobs and E. G. Spencer, During the Publication of Scientific Journals, Research Pro- Appeared as Part 2 of the Journal of Applied Physics 36, posal to the National Science Foundation, 1 v. (Inforonics, Inc., No. 3, 1110-1111 (Mar. 1965). Maynard, Mass., Dec. 16, 1963). Chapman, R. E. and M. J. Fisher, A New Technique for Removable Bujnoski, F., On-Line Memory Integrity Evaluation and Improve- Media, Read-Only Memories, AFIPS Proc. Fall Joint Com- ment, Computer Design '7, No. 12, 31-34 (Dec. 1968). puter Conf., Vol. 31, Anaheim, Calif., Nov, 14-16, 1967, Burdick, D. S. and T. H. Naylor, Design of Computer Simulation pp. 371-379 (Thompson Books, Washington, D.C., 1967). Experiments forIndustrialSystems, Commun. ACM 9, Chase, E. N., Computer-Driven Line-Drawing Displays- A User's 329-339 (May 1966). View, Data Proc. Mag. 10, 24-27 (Jan. 1968). Burge, W. H., A Reprogramming Machine, Commun. ACM 9, Cheng, G. C., R. S. Ledley, D. K. Pollock and A. Rosenfeld, Eds., 60-66 (Feb. 1966). S Pictorial Pattern Recognition,,Proc. Symp.on AUtOMElde Photo- Burger, J. B., High Speed Display of Chemical Nomenclature, interpretation, Washington, D.C., May 31-June 2, 1967, 521 p. Molecular Formula and Structural Diagram, Rept. No. C105- (Thompson Book Co., Washington, D.C., 1968). R-4, 19 p. (General Electric Co., Huntsville, Ala., Dec. 3L Cheydleur, B. F., SHIEF: A Realizable Form of Associative 1964): Memory, Am. Doc. 14, No. 1, 56-57 (Jan. 1963). Burgess, E., Ed., On-Line Computing Systems, Proc. Symp. Cheydleur, B. F., Ed., Colloquium on Technical Preconditions sponsored by the Univ. of California, Los Angeles, and for Retrieval Center Operations, Proc. National Colloquium on Informatics, Inc., Los Angeles, Calif., Feb. 2-4, 1965, 152 p. Information Retrieval, Philadelphia, Pa., April 24-25, 1964, (American Data Processing, Inc., Detroit, Mich., 1965). 156 p. (Spartan Books, Washington, D.C., 1965). Burkhardt, W. H., Universal Programming Languages and Chong, C. F., R. Mosenkis and D. K. Hanson, Engineering Design Processors: A Brief Survey and New Concepts, AFIPS Proc. of a Mass Random Access Plated Wire Memory, AFIPS Proc. Fall Joint Computer Conf., Vol. 27, Pt. 1, Las Vegas, Nev., Fall Joint Computer Conf., Vol. 31, Anaheim, Calif., Nov. 14-16, Nov. 30-Dec. 1, 1965, pp. 1-21 (Spartan Books, Washington, 1967, pp. 363-370 (Thompson Books, Washington, D.C., 1967). D.C., 1965). Christensen, C. and E. N. Pinson, Multi-Function Graphics for a Bush, G. P., Cost Considerations, in Electronics in Management, Large Computer System, AFIPS Proc. Fall Joint Computer Ed. L. H. Hattery and G. P. Bush, pp. 101-111 (The University Conf., Vol. 31, Anaheim, Calif., Nov. 14-16, 1967, pp. 697-711 Press of Washington, Washington, D.C., 1956). (Thompson Books, Washington, D.C., 1967). Butler, E. J.Microfilm and Company Survival Plans, Proc. Chu, Y., A Destructive-Readout Associative Memory, IEEE Eleventh Annual Meeting and Convention, Vol. XI, Washing- Trans. Electron. Computers EC-14, No. 4, 600-605 (Aug. 1965). ton, D.C., Apr. 25-27, 1962, Ed. V. D. Tate, pp. 57-68 (The Clancy, J. C. and M. S. Fineberg, Digital Simulation Languages: National Microfilm Assoc., Annapolis, Md., 1962). A Critique and A Guide, AFIPS Proc. Fall Joint Computer Cain, A. M. and I. H. Pizer, The SUNY Biomedical Communica- Conf., Vol. 27, Pt. 1, Las Vegas, Nev. Nov. 30-Dec. 1, 1965, tion Network: Implementation of an On-Line, Real-Time, pp. 23-36 (Spartan Books, Washington,D.C,., 1965). User-Oriented System, in Levels of Interaction Between Man Clapp, L. C., Some Brainware Problems in Information Systems and Information, Proc. Am. Doc. Inst. Annual Meeting, Vol. 4, and Operations Analysis, in Information System Science and New York, N.Y., Oct. 22-27 1967, pp. 258-262 (Thompson Technology, papers prepared for the Third Cong., Scheduled Book Co., Washington, D.C., 1967). for Nov. 21-22, 1966, Ed. D. E. Walker, pp. 3-6 (Thompson Cameron, S. H., D. Ewing and M. Liveright, DIALOG: A Con- Book Co., Washington, D.C., 1967). versational Programming System with a Graphical Orientation, Clem, P. L., Jr., AMTRAN- A Conversational-Mode Computer Commun. ACM 10, 349-357 (June 1967). System for Scientists and Engineers, in Proc. IBM Scientific Campbell, D. J. and W. J. Heffner, Measurement and Analysis of Computing Symp. on Computer-Aided Experimentation, York- Large Operating Systems During System Development, AFIPS town Heights, N.Y., Oct. 11-13, 1965, pp. 115-150 (IBM Corp., Proc. Fall Joint Computer Conf., Vol, 33, Pt. 1, San Francisco, White Plains,, N.Y., 1966). Calif., Dec. 9-11, 1968, pp. 903-914 (Thompson Book Co., Climenson, W. D., File Organization and Search Techniques, in Washington, D.C., 1968). Annual Review of Information Science and Technology, Cantrell, H. N. and A. L. Ellison, Multiprogramming System Vol. 1, Ed. C. A. Cuadra, pp. 107-135 (Interscience Pub., Performance Measurement and Analysis, AFIPS Proc. Spring New York, 1966). Joint Computer Conf., Vol. 32, Atlantic City, N.J., April 30- Clippinger, R. F., Programming Implications of Hardware Trends, May 2, 1968, pp. 213-221 (Thompson Book Co., Washington, IFIP Congress 65, Vol. 1, New York, N.Y., May 24-29, 1965, D.C., 1968). Ed. W. A. Kalenich, pp. 207-212 (Spartan Books, Washington, Caracciolo di Forino, A., Linguistic Problems in ProgramMing D.C., 1965). Theory, in Information Processing 1965, Proc. IFIP Congress Coffman, E. G., Jr., and 0. Kleinrock, Computer Scheduling 65, Vol. 1, New York, N.Y., May 24-29, 1965, Ed. W, A. Methods and Their Countermeasures, AFIPS Proc. Spring Kalenich, pp. 223-228 (Spartan Books, Washington, D.C., Joint Computer Conf., Vol. 32, Atlantic City, N.J.Apr. 30- 1965). May 2, 1968, pp. 11-21 (Thompson Book Co., `Washington, Caracciolo diForino, A., Special Programming Languages, D.C., 1968). Centro Studi Calcolatrici Elettroniche, Universita di Pisa, Italy, 1965, 21 p. Coggan, B. B., The Design of a Graphic Display System, Rept. Carlson, C. 0. and H. D. Ives, Some Considerations in the Design No. 67-36, 135 p. (Dept. of Engineering, Univ. of California, of a Laser Thermal Microimage Recorder, 1968 WESCON Los Angeles, Aug. 1967). Technical Papers, 16/1, Aug. 1968, 8 p. Cohen, J., A Use of Fast and Slow Memories in List-Processing Carlson, G., Techniques for Replacing Characters That Are Languages, Commun. ACM 10, No. 2, 82-6 (Feb. 1967). Garbled on Input, AFIPS Proc. Spring Joint Computer Conf., Cohen, M. I.B. A. Unger and J. F. Milkosky, Laser Machining Vol. '8, Boston, Mass., April 1966, pp. 189-192 (Spartan of Thin Films and Integrated Circuits, Bell Sys. Tech. J. 47, Boor- ,, Washington, D.C., 1966). 385-405 (Mar. 1968). Carr, J. W., III and N. S. Prywes, Satellite Computers as Inter- Cohler, E. U. and H. Rubenstein, A Bit-Access Computer in a preters, Electronics 38, No. 24, 87-89 (1965). Communication System, AFIPS Proc. Fall Joint Computer Castleman, P. A., An Evolving Special-Purpose Time-Sharing Conf., Vol. 26, San Francisco, Calif., Oct. 1964, pp. 175-185 System,' Computers & Automation 16, 1.6-18 (Oct. 1967). (Spartan Books, Baltimore, Md., 1964). 131 Cole, R. I., Ed., Data/Information Availability, 183 p. (Thompson Daly, J., R. D. Joseph and P. M. Kelly, Self-OrganizingLogic Book Co., Washington, D.C., 1966). L;ystems, 27 p. (Astropower, Inc., Costa Mesa, Calif., Jan. 1962). Collier, R. J., Some Current Views on Holography, IEEESpec- Damron, S., J. Lucas, J. Miller, E, Salbo and M. Wilmnann, trum 3, 67-74 (July 1966). A Random Access Terabit Magnetic Memory, AFIPS Proc. Col lila, R. A., Time-Sharing and MultiprocessingTerminology, Fall Joint Computer Conf., Vol. 33, Pt. 2, San Francisco, Calif., Datamation 12, No. 3, 49-51 (Mar. 1966). Dec. 9-11, 1968, pp. 1381-1387 (Thompson Book Co., Washing- Connolly, R., Itek EDP System Utilizes Laser, EmulsionTech- ton, D.C., 1968). niques, Electronic News 10, No. 475, 4 (Feb. 15.:x.965). Dantine, D. J., Communications Needs of the User for Manage- Constantine, L. L., Integral Hardware/Software Design,Part 1: ment Information Systems, AFIPS Proc. Fall Joint Computer The Characteristics of a Modern Data System, ModernData Conf. Vol. 29, San Francisco, Calif., Nov. 7-10, 1966,pp. 403 Systems 1, No. 2, 50-59 (Apr. 1968). 411 (Spartan Books, Washington, D.C., 1966). Conway, M. E. and L. M. Spandorfer, A Computer SystemDe- Davenport, W. C. CAS Computer-Based Information Services, signer's View of Large Scale Integration, AFIPS Proc. Fall Datamation 14:No. 3, 33-39 (Mar. 1968). Joint Computer Conf., Vol. 33, Pt. 1, San Francisco,Calif., Davenport, W. P. Efficiency and Error Control in Data Communi- Dec. 9-11, 1968, pp. 835-845 (Thompson Book Co., Washing- cations, Pt. 2, Elam Proc. Mag. 8,30-35 (Sept. 1966). ton, D.C., 1968). David, E. E., Jr., Prologue Computers for All Seasons,in The Conway, R. W., B. M. Johnson and W. I. Maxwell, SomeProblems Human Use of Computing Machines, Proc. Symp. concerned of Digital Systems Simulation, Management Sci. 6,No. 1, with Diverse Ways of Enhancing Perception and Intuition, 92-110 (Jan. 1959). Bell Telephone Laboratories, Murray Hill, N.J., June 20-21, Conway, R. W., J. J. Delfausse, W. L. Maxwell and W. E.Walker, 1966, pp. 1-3 (Bell Telephone Labs., 1966). CLP The Cornell List Processor, Commun. ACM8, No. 4, 215-216 (Apr. 1965). David, E. E., Jr., Sharing a Computer, Int. Sci. Tech.54, 38-47 (1966). Cooper, B., Optical Communications in the Earth'sAtmosphere, IEEE Spectrum 3, 83-88 (July 1966). Davis, R. M., Information Control in Command-Control Systems, in New Perspectives in Organization Research, Ed. W. W. Cooper, W. W., H. J. Leavitt and M. W. Shelly, II, Eds.,New Cooper et al., pp. 464-478 (Wiley, New York, 1964). Perspectives in Organization Research, 606p. (Wiley, New York, 1964). Davis, R. M., Military Information Systems Design Techniques, Corbato, F. J, and V. A. Vyssotsky, Introduction and Overview in Military Information Systems The Design of Computer- Aided Systems for Command, Ed. E. Bennettet al., pp. 19-28 of the Multics System, AFIPS Proc. Fall Joint ComputerConf., (Frederick A. Praeger, Pub, New York, 1964). Vol. 27, Pt. 1, Las Vegas, Nev., Nov. 30Dec. 1, 1965,pp. 185 196 (Spartan Books, Washington, D.C., 1965). Davis, R. M., Classification and Evaluation of Information System Design Techniques, in Second Cong.on the Informa- Cornew, R. W., A Statistical Method of Spelling Correction,Inf. & Control 12, 79-93 (Feb. 1968). tion System Sciences, The Homestead, Hot Springs, Va., Nov. 1964, Ed. J. Spiegel and D. E. Walker,pp. 77-83 (Spartan Craig, J. A., S. C. Berezner, H. C. Carney and C. R.Longyear, Books, Washington, D.C., 1965). DEACON: Direct English Access and Control, AFIPS Proc. Fall Joint Computer Conf., Vol. 29, San Francisco, Calif., Davis, R. M., Man-Machine Communication, in Annual Review Nov. 7-10, 1966, pp. 365-380 (Spartan Books, Washington, of Information Science and Technology, Vol. 1, Ed. C. A. D.C., 1966). Cuadra, pp. 221-254 (Interscience Pub., New York,1966). Davis, R. M., Information Control In An informationSystem, Crawford, D. J., R. L. Moore, J. A. Parisi, J. K. Piccianoand draft of lecture delivered to the Washington, D.C., Chapter, W. D. Pricer, Design Considerations fora 25-Nanosecond The Institute of Management Sciences, Oct. 18, 1967, 49p. Tunnel Diode Memory, AFIPS Proc. Fall Joint Computer Davis, R. M., Communication Technology and Planning Conf., Vol. 27, Pt. 1, Las Vegas, Nev., Nov. 30Dec.1, 1965, in Support of Network Implementation, unpublishedmemo- pp. 627-636 (Spartan Books, Washington, D.C., 1965). randum, 1968, 6 p. Crosby, W. S., A Systems Design for the Integrated Bibliography Denning, P. J., The Working Set Model for Program Behavior, Pilot Study, in Integrated Bibliography Pilot StudyProgress Commun. ACM 11, No. 5, 323-333 (May 1968). Report, Ed. L. Sawin and R. Clark,pp. 33-49 (Univ. of Colorado, Denning, P. J., Thrashing: Its Causes and Prevention, AFIPS Boulder, June 1965). Proc. Fall Joint Computer Conf., Vol. 33, Pt. 1, SanFrancisco, Croxton, F. E., Identification of Technical Reports, in The Calif., Dec. 9-11, 1968,pp. 915-922 (Thompsoct Book Co., Production and Use of Technical Reports, Proc. Workshopon Washington, D.C., 1968). the Production and Use of Technical Reports, conductedat the Dennis, J. B., Segmentation and the Design of Multiprogrammed Catholic Univ. of America, April 13-18, 1953, Ed. B. M. Fry Computer Systems, J. ACM 12,589-602 (Oct. 1965). and J. J. Kortendick, pp. 121-135 (The Catholic Univ. of Dennis, J. B., A Position Paperon Computing and Communi- America Press, Washington, D.C., 1955). cations, Commun. ACM 11, No. 5, 370-377 (May 1968). Crutcher, W. C., Common Creation and Ownership of Massive Dennis, J. B. and E. L. Glaser, The Structure of On-LineInforma- Data Banks, Computers & Automation 18, No. 5,24-26 tion Processing Systems, in Second Cong.on the Information (May 1969). System Sciences, Hot Springs, Va., Nov. 1964, Ed. J. Spiegel and D. E. Walker, pp. 5-14 (Spartan Books, Washington,D.C., Cuadra, C. A., Ed., Annual Review of Information Scienceand 1965). Technology, Vol. 1, 389 p. (Interscience Pub., New York,1966). Dennis, J. B. and E. C. Van Horn, Programmed Semantics for Cuadra, C. A., Ed., Annual Review of Information Science and Multiprogrammed Computations, Rept. No. MACTR-23,46 p. Technology, Vol. 2, 484 p. (Interscience Pub., New York,1967). (Massachusetts Institute of Technology, Cambridge, Dec.1965). Cunningham, J. F., The Need for ADP Standards in the Federal Dent, J., Diagnostic Engineering, IEEE Spectrum 4,99-104 Community, Datamation 15, No. 2, 26-28 (Feb. 1969). (July 1967): Cutrona, L. J.J., Recent Developments in Coherent Optical Tech- DeParis, J. R., Random Access, Data Proc. Mag. 7, No.2, 30-31 nology, in and Electro-Optical Information Processing, (Feb. 1965). Ed. J. R. Tippett et al., pp. 83-123 (M.I.T., Press, Cambridge, DeParis, J. R., Desk Top Teleprinter, Data Proc. Mag. 7,No. 10, Mass., 1965). 48-49 (Oct. 1965). Dahl, 0. J. and K. Nygaard, SIMULA An ALGOL-Based Dertouzos, M. L., PHASE PLOT: An On-Line GraphicalDisplay Technique, IEEE Trans. Electron. Computers EC-16, Simulation Language, Commun. ACM 9, No. 9, 671-678(Sept. 203-209 1966). (April 1967). Dahm, D. M., F. H. Gerbstadt and M. M. Pacelli,A System Dillon, et al., 1964See Chang et a1.,1964. Organization for Resource Allocation, Commun.ACM 10, Dimeff, J., W. D. Gunter, Jr., and R. J. Hruby, SpectralDependence No. 12, 772-779 (Dec. 1967). of Deep-Space Communications Capability, IEEESpectrum 4, Daley, R. C. and J. B. Dennis, Virtual Memory,Processes, and No. 9, 98-104 (Sept. 1967). Sharing in MULTICS, Commun. ACM 11, No.5, 306-312 Dodd, G. G., APL A Language for Associative Data Handling (May 1968). in PL/1, AFIPS Proc. Fall Joint Computer Conf., Vol.29, 132 San Francisco, Ca lif.,, Nov. 7-10, 1966, pp. 677-684 (Spartan Fano, R. M. and F. J. Corbato, Time-Sharing on Computers, Books, Washington, D.C., 1966). Scient. American 215, No. 3, 129-140 (1966). Dorff, E. K., Computers and Communications: Complementing Fedde, G. A., Plated Magnetic Cylindrical Thin Film Main Technologies, Computers & Automation 18, No 5, 22-23 Memory Systems, AMPS Proc. Fall Joint Computer Conf., (May 1969). Vol. 31, Anaheim, Calif., Nov. 14-16, 1967, pp. 594-596 Dorion, G. H., R. W. Roth, J. J. Stafford and G. Cox, CRT (Thompson Books, Washington, D.C., 1967). Phosphor Activation of Photochromic Film, Inf. Display 3, Feldman, J. A., Aspects of Associative Processing, Tech. Note 56-58 (Mar./Apr. 1966). 1965-13, 47 p. (Lincoln Laboratory, M.I.T., Lexington, Mass Drew, D. L., R. K. Summit, R. I. Tanaka and R. B. Whitely, Apr. 21, 1965). An On-Line Technical Library Reference Retrieval System, Fernbach, S., Computers in the U.S.A.-Today and Tomorrow, in Information Processing 1965, Proc. IFIP Congress 65, in Information Processing 1965, Proc. IFIP Congress 65, Vol. 2, New York, N.Y., May 24-29, 1965, Ed. W. A. Kalenich, Vol. 1, New York, N.Y., May 24-29, 1965, Ed. W. A. Kalenich, pp. 341-342 (Spartan Books, Washington, D.C., 1966). Also pp. 77-85 (Spartan Books, Washington, D.C., 1965). in Am. Doc. 17, No. 1, 3-7 (Jan. 1966). Fife, D. W, and R. S. Rosenberg, Queuing in a Memory Shared Dreyer, L., Principles of a Two-Level Memory Computer, Com- Computer, Proc. 19th National Conf., ACM, Philadelphia, Pa., puters & Automation 17, No. 5,40 -42 (May 1968). Aug. 25-27, 1964, pp. H1-1 to H1-12 (Assoc. for Computing Duffy, G. F. and W. D. Timberlake, A Business-Oriented Time- Machinery, New York, 1964). Sharing System, AFIPS Proc. Spring Joint Computer Conf., Fine, G. H., C. W. Jackson and P. V. McIsaac, Dynamic Program Vol. 28, Boston, Mass., April 1966, pp. 265-275 (Spartan Book9, Behavior Under Paging, Proc. 21st National Conf. ACM, Washington, D.C., 1966). Los Angeles, Calif., Aug. 30 -Sent,1, 1966, pp. 223-228 Duggan, M. A., Software Protection, Datamation 15, No. 6, (Thompson Book Co., Washington, D.C., 1966). Also in Rept. 113, 116 (June 1969). No, SP-2397, 19 p. (System Development Corp., Santa Monica, Duncan, C. J., General Comment, in Advances in Computer Calif., Jerre 1966). Typesetting, Proc. Int. Computer Typesetting Conf., Sussex, Fischer, G. L., Jr., D. K. Pollock, B. Radack and M. E. Stevens, England, July 14-18, 1966, Ed. W. P. Jaspert, pp. x-xi (The Eds., Optical Character Recognition, 412 p. (Spartan Books, Institute of Printing, London, 1967). Washington, D.C., 1962). Duncan, C. J., Advanced Computer Printing Systems, The Pen- Fischler, M. A. and A. Reiter, Variable Topology Random Access rose Annual 59, 254-267 (1966). Memory Organization, AFIPS Proc. Spring Joint Computer Dunn, E. S., Jr., The Idea of a National Data Center and the Conf., Vol. 34, Boston, Mass., May 14-16, 1969, pp. 381-391 Issue of Personal Privacy, The Amer. Statistician 21, 21-27 (AFIPS Press, Montvale, N.J., 1969). (Feb. 1967). Fisher, D. L., Data, Documentation and Decision Tables, Commun. Dunn, R. S., The Case for Bipolar Semiconductor Memories, ACM 9, No. 1, 26-31 (Jan. 1966). AFIPS Proc. Fall Joint Computer Conf., Vol. 31, Anaheim, Fleet, J. J., How Far Away Are Cheap Mass Memories, Data & Calif., Nov. l4 -16, 1967, pp. 596-598 (Thompson Books, Control 3, No. 8, 28-29 (Aug. 1965). Washington, D.C., 1967). Fleisher, A., P. Pengelly, J. Reynolds, R. Schools and G. Sincer- box, An Optically Accessed Memory Using the Lippman Ebersole, J. L., North American Aviation's National Operating Process for Information Storage, in Optical and Electro- System, in Toward a National Information System, Second Optical Information Processing, Ed. J. R. Tippett et al., pp. Annual National Colloquium on Information Retrieval, Phila- 1-30 (M.I.T. Press, Cambridge, Mass., 1965). delphia, Pa., April 23-24, 1965, Ed. M. Rubinoff, pp. 169-198 Flynn, M. J., A Prospectus on Integrated Electronics and Com- (Spartan Books, Washington, D.C., 1965). puter Architecture, AFIPS Proc. Fall Joint Computer Conf., Ebersole, J. L., An Operating Model of a National Information Vol. 29, San Francisco, Calif Nov. 7-10, 1966, pp. 97-103 System, Am. Doc. 17, No. 1, 33-40 (Jan. 1966). (Spartan Books, Washington, D.C., 1966). Eckman, D. P., Ed., Systems: Research and Design, Proc. First Forbes, R. E., D. H. Rutherford, C. B. Stieglitz and L. H. Tung, Systems Symp. at Case Institute of Technology, Cleveland, A Self-Diagnosable Computer, AFIPS Proc. Fall Joint Com- Ohio, April 1960, 310 p. (Wiley, New York, 1961). puter Conf., Vol. 27, Pt. 1, Las Vegas, Nev., Nov. 30-Dec. 1, Edwards, A. W. and R. L. Chambers, Can A Priori Probabilities 1965, pp. 1073-1086 (Spartan Books, Washington, D.C., 1965). Help in Character Recognition, J. ACM 11, No. 4, 465-470 Forgie, J. W., A Time- and Memory-Sharing Executive Program (Oct. 1964). for QuickResponse On-Line Applications, AFIPS Proc. Fall Edwards, W., Probabilistic Information Processing System for Joint Computer Conf., Vol. 27, Pt. 1, Las Vegas,'Nev., Nov. 30- Diagnosis and Action Selection, in Second Cong. on the Dec. 1, 1965, pp. 599-609 (Spartan Books, Washington, D.C., Information System Sciences, Hot Springs, Va., Nov. 1964, 1965). Ed. J. Spiegel and D. E. Walker, pp. 141-155 (Spartan Books, Fossum, E. G. and G. Kaskey, Optimization of Information Washington, D.C., 1965). Retrieval Language and Systems, Final Rept. under Contract Emerson, M., The 'Small' Computer versus Time-Shared Systems, AF 49(638)1194, 87 p. (UNIVAC, Blue Bell, Pa., Jan. 28, 1966). Computers & Automation 14, No. 9, 18-20 (Sept. 1965). Foster, J. M., List Processing, 54 p. (American Elsevier Pub., Estrin, G., D. Hopkins, B. Coggan and S. D. Crocker, SNUPER New York, 1967). COMPUTER - A Computer in Instrumentation Automation, Fox, R. S., R. L. McDaniel, C. N. Mooers and W. J. Sanders, A AFIPS Proc. Spring Joint Computer Conf., Vol. 30, Atlantic Technique for Overcoming Lack of Standardization in In- City, N.J., April 18-20, 1967, pp. 645-656 (Thompson Books, formation Network Operation, in Progress in Information Washington, D.C., 1967). Science and Technology, Proc. Am. Doc. Inst. Annual Meeting, Vol. 3, Santa Monica, Calif., Oct. 3-7, 1966, pp. 157-165 Estrin, G. and L. Kleinrock, Measures, Models and Measurements (Adrianne Press, 1966). for Time-Shared Computer Utilities, Proc. 22nd National Conf., Frank, A. L., B-Line, Bell Line Drawing Language, AFIPS Proc. ACM, Washington, D.C., Aug. 29-31, 1967, pp. 85-96 (Thomp- Fall Joint Computer Conf., Vol. 33, Pt. 1, San Francisco, son Book Co., Washington, D,C., 1967). Calif., Dec. 9-11, 1968, pp. 179-191 (Thompson Book Co., Evans, G. J., Jr., Experience Gained from the American Airlines Washington, D.C., 1968). SABRE System Control Program, Proc. 22nd National Conf., Frank, W. L., On-Line CRT Displays: User Technology and ACM, Washington, D.C., Aug. 29-31, 1967, pp. 77-83 (Thomp- Software, inOn-Line Computing Systems, Proc. Symp. son Book Co., Washington, D.C., 1967). sponsored by the Univ. of California, Los Angeles, and in- Evans, T. G. and D. L. Darley, DEBUG -An Extension to Current formatics, Inc., Feb. 1965, Ed. E. Burgess, pp. 50-62 (American Online Debugging Techniques, Commun. ACM 8, No. 5, Data Proc., Inc., Detroit, Mich., 1965). 321-326 (May 1965). Franks, E. W., A Data Management System for Time-Shared Evans, T. G. and D. L. Darley, On-Line Debugging Techniques: File Processing Using a Cross-Index File and Self-Defining A Survey, AFIPS Proc. Fall Joint Computer Conf., Vol. 29, Entries, AFIPS Proc. Spring Joint Computer Conf., Vol. 28, San Francisco, Calif Nov. 7-10, 1966, pp. 37-50 (Spartan Boston, Mass., April 1966, pp. 79-86 (Spartan Books, Wash- Books, Washington, D.C., 1966). ington, D.C., 1966). 133 French, M. B., Disk and Drum Memories, Modern Data 2, No. 5, Goettel, H. J Bell System Business Communications Seminar, 42-45, 48-50 (May 1969). in Data Processing, Vol. X, Proc. 1966 International Data Frijda, N. H., Problems of Computer Simulation, Behay. Sci. Processing Conf., Chicago, Ill., June 21-24, 1966, pp. 188-197 12, 59-67 (Jan. 1967). (Data Processing Management Assoc., 1966). Fry, B, M. and J, J, Kortendick, Eds., The Production and Use Gordon, J. P., Optical Communication, Int. Sci. & Tech. 44, of Technical Reports, Proc. Workshop on the Production and 60-69 (Aug. 1965). Use of Technical Reports, conducted at the Catholic Univ. of Gorn, S., Advanced Programming and the Aims of Standardiza- America, April 13-18, 1953,, 175 p. (The Catholic Univ. of tion, Commun. ACM 9, No. 3, 232 (Mar. 1966), America Press, Washington, D.C., 1955). Gould, R. L., GPSS/360- An Improved General Purpose Simu- Fubini, E. G., The Opening Address, in Second Cong. on the lator, IBM Sys. J. 8, No. 1, 16-27 (1969). Information System Sciences, Hot Springs, Va., Nov. 1964, Grocer, F, and R, A. Myers, Graphic Computer-Assisted Design Ed. J. Spiegel and D. E. Walker, pp. 1-4 (Spartan Books, of Optical Filters, IBM J. Res. & Dev. 13, No, 2, 172-178 Washington, D.C., 1965). (Mar. 1969). Fuller, R. H., Content-Addressable Memory Systems, Rept. No, Graham, R. F., Semiconductor Memories; Evolutionor Revolu- 63:725, 2 v. (Dept. of Engr., Univ. of Calif., Los Angeles, Calif., tion?, Datamation 15, No. 6, 99-101, 103-104 (June 1969). 1963). Graham, R. M., Protection in an Information Processing Utility, Fuller, R. H. and R. M. Bird, An Associative Parallel Processor Commun. ACM 11, No. 5, 365-369 (May 1968). with Application to Picture Processing, AFIPS Proc. Fall Greenberger, C. B., The Automatic Design of a Data Processing Joint Computer Conf., Vol. 27, Pt. 1, Le,'4 Vegas, Nev., Nov. 30- System, in Information Processing 1965, Proc. IFIP Congress Dec. 1, 1965, pp. 105-116 (Spartan Books, Washington, D.C., 65, Vol. 1, New York, N.Y., May 24-29, 1965, Ed. W. A. 1965). Kalcnich, pp. 277-282 (Spartan Books, Washington, D.C., Fulton, R. L., Visual Input to Computers, Datamation 9, No. 8, 1965). 37-40 (Aug. 1963). Greenberger, M., A Multipurpose System for the Structuring On-Line of Information and Decision Processes, in Informa- Gabor, D., Associative Holographic Memories, IBM J. Res. & tion Processing 1965, Proc. IFIP Congress 65, Vol. 2, New Dev. 13, No. 2, 156-159 (Mar. 1969). York, N.Y., May 24-29, 1965, Ed. W. A. Kalenich, pp. 347-348 Gaines, R. S. and C. Y. Lee, An Improved Cell Memory, IEEE (Spartan Books, Washington, D.C., 1966). Trans. Electron. Computers EC-14, No. 1, 72-75 (Feb. 19651. Griffith, R. L., Data Recovery in a Photo-Digital Storage System, Gall, R. G., Hybrid Associative Computer Study, Rept. No. IBM J. Res. & Dev. 13, No. 4, 456-467 (July 1969). RADC-TR-65-445, Vol. 1, 109 p. (Rome Air Development Gross, W. A., Information Storage and Retrieval, A State-of-the- Center, Griffiss Air Force Base, N.Y., July 1966). Art Report, Ampex READOUT, special issue, 9 p. (Ampex Gallenson, L, and C. Weissman, Time-Sharing Systems: Real and Corp., Redwood City, Calif., 1967). Ideal, Rept. No. SP-1272- 20 p. (System Development Corp., Giinther, A., Microphotography in the Library, UNESCO Bull. Santa Monica, Calif., Mar. 1965). Lib. 16,1 -22. (1962). Gamblin, R. L., Some Problems of a High Capacity Read-Only Gurk, H. M. and J. Minker, The Design and Simulation ofan Digital Information Storage System, 1968 WESCON Technical Information Processing System, J. ACM 8, 260-270 (1961). Papers, 16/3, Aug. 1968, 6 p. Garvin, P. L. Ed., Natural Language and the Computer, 398 p. (McGraw-Hill Book Co., Inc.,New York, 1963). Hagan, T. G., R. J. Nixon and L. J. Schaefer, The Adage Graphics Gayer, D. P., Jr., Probability Models for Multiprogramming Terminal, AFIPS Proc. Fail Joint Computer Conf., Vol. 33, Computer Syst ems, J. ACM 14, No. 3, 423-438 (July 1967). Pt. 1, San Francisco, Calif., Dec. 9-11, 1968, pp. 747-755 Geldermans, P., H. 0. Leilich and T. R. Scott, Characteristics (Thompson Book Co., Washington, D.C., 1968). of the Chain Magnetic Film Storage Element, IBM J. Res. & Halpern, M., The Case for Natural-Language Programming, Dev. 11, 291-301 (May 1967). Proc. Fall Joint Computer Conf., Vol. 29, San Francisco, Calif., Gelernter, H. L., Realization of a Geometry Theoreir Proving Nov. 7-10, 1966, pp. 639-649 (Spartan Books, Washington, Machine, in UNESCO Information Processing, Proc. Int. Conf., D.C., 1966). Paris, June 15-20, 1959, pp. 273-282 (Oldenbourg, Munich; Halpern, M., The Foun rations of the Case for Natural-Language Butterworths, London, 1960). Programming, AFIPS Proc. Fall Joint Computer Conf., Vol. 29, Gentle, E. C., Jr., Data Communications in Risiness, 163 p. San Francisco, Calif., Nov. 7-10, 1966, pp. 639-649 (Spartan (American 1 elephone and Telegraph Co., New York, 1965). Books, Washington, D.C., 1966). Another version appears in Gibbs, G. and J. MacPhail, Philco Corporation, session on optical IEEE Spectrum 4, 140-149 (Mar. 1967). character page readers, in Research and Engineering Council Hanlon, A. G., W. C. Myers and C. 0. Carlson, Photochromic of the Graphic Arts Industry, Inc., Proc. 14th Annual Conf., Micro-Image Technology, A Review, paper present d at the Rochester, N.Y., May 18-20, 1964, pp. 95-106 (Washington, Equipment Manual Symp., jointly sponsored by the American D.C., 1964). Ordnance Assoc., and the Army Material Command, Detroit, Gibson, D. H., Considerations in Block-Oriented Systems Design, Mich., Nov. 30-Dec. 2, 1965, 35 p. (The National Cash Register AFIPS Proc. Spring Joint Computer Conf., Vol. 30, Atlantic Co., Hawthorne, Calif., 1965). City, N.J., April 18-20, 1967, pp. 75-80 (Thompson Books, Hansen, M. H. and J. L. McPherson, Potentialities and Problems Washington, D.C., 1967). of Electronic Data Processing, in Electronics in Management, Gill, S., The Changing Basis of Programming, in Information Ed. L. H. Hattery and G. P. Bush, pp. 53-66 (The Univ. Press Processing 1965, Proc. IFIP Congress 55, Vol. 1, New York, of Washington, Washington, D.C., 1956). N.Y., May 24-29, 1965, Ed. W. A. Kalenich, pp. 201-206 Harder, E. L., The Expanding World of Computers, Commun. (Spartan Books, Washington, D.C., 1965). ACM 1 1, 231-239 (Apr. 1968). Haring, D. R., The Beam Pen: A Novel High Speed Input/Output Glaser, E. L., J. F. Couleur and G. A. Oliver, System Design of a Device for Cathode-Ray-Tube Display Systems, AFIPS Proc. Computerfor Time Sharing Applications, AFIPS Proc. Fall Joint Computer Conf., Vol. 27, Pt. 1, Las Vegas, Nev., Fall Joint Computer Conf., Vol. 27, Pt. 1, Las Vegas, Nev., Nov. 30-Dec. 1, 1965, pp. 847-855 (Spartan Books, Washington, Nov. 30-Dec. 1, 1965, pp. 197-202 (Spartan Books, Wash- D.C., 1965). ington, D.C., 1965). Haring, D. R., A Display Console for an Experimental Computer- Glaser, E., D. Rosenblatt and M. K. Wood, The Design of a Based Augmented Library Catalog, Proc. 23rd National Conf., Federal Statistical Data Center, The American Statistician ACM, Las Vegas, Nev., Aug. 27-29, 1968, pp. 35-43 (Brandon/ 21, No. 1, 12-20 (Feb. 1967). Systems Press, Inc., Princeton, N.J., 1968). Gluck, S. E., Impact of Scratchpads in Design: Multifunctional Haring, D. R., Computer-Driven Display Facilities foran Experi- Scratchpad Memories in the Burroughs B8500, AFIPS Proc. mental Computer-Based Library, AFIPS Proc.Fall Joint Fall Joint Computer Conf., Vol. 27, Pt. 1, Las, Vegas, Nev., Computer Conf., Vol. 33, Pt. 1, San Francisco, Calif., Dec. 9- Nov. 30 -Dec. 1, 1965, pp. 661-666 (Spartan Books, Wash- 11, 1968, pp. 255-265 (Thompson Book Co., Washington, D.C., ington, D.C., 1965). 1968). 134 Hart, U.L., Photochromic Materials for Data Storage and Horvath, V. V., J, M. Holeman and C. Q. Lemmond, Fingerprint Display, Second Annual Report on Contract Nonr 4583(00), Recognition by Holographic Techniques, in Law Enforcement 36 p. (UNIVAC Defense Systems Division, Sperry Rand Corp, Science and Technology, Vol. 1, Proc. First National Symp. St. Paul, Minn., May 15, 1966). on Law Enforcement Science and Technology, Chicago, Hartsuch, P. J., Graphics Arts Progress. , .in 1967, Graphic March 1967, Ed. S. A. Yefsky, pp. 485-492 (Thompson Book Arts Monthly 40, 52-57, 61 (Jan. 1968). Co., Washington, D.C., 1967). Haskell, J. W., Design of a Printed Card Capacitor Read-Only Housden, R. J. W., The Definition and Implementation (,)f LSIX Store, IBM J. Res. & Dev. 10, No. 2, 142-157 (Mar. 1966). in BCL, The Computer J.12, No. 1, 15-23 (Feb. 1969). Hattery, L. H. and G. P. Bush, Eds., Electronics in Management, Howe, W., High-Speed Logic Circuit Considerations, AFIPS 207 p. (The University Press of Washington, Washington, D.C., Proc. Fall Joint Computer Conf., Vol. 27, Pt. 1, Las Vegas, 1956). Nev., Nov. 30-Dec. 1, 1965, pp. 505-510 (Spartan Books, Hayes, R. M., Mathematical Models for Information Retrieval, Washington, D.C., 1965). in Natural Language and the Computer, Ed. P. L. Garvin, Huang, T. S. and 0. J. Tretiak, Research in Picture Processing, pp. 268-309 (McGraw-Hill, New York, 1963). in Optical and Electro-Optical Information Processing, Ed. Hayes, R. M., Information Retrieval: An Introduction, Data- J. R. Tippett et al., pp. 45-57 (M.I.T. Press, Cambridge, Mass., mation 14, No. 3, 22-26 (Mar. 1968). 1965). Head, R. V., The Programming Gap in Real-Time Systems, Hudson, D. M., The Applications and Implications of Large- Datamation 9, 39-41 (Feb. 1963). Scale Integration, Computer Design 7, 38-42, 47-48 (June Heiner, J. A., Jr., and R. 0. Leishman, Generalized Random 1968). Extract Device, Proc. 21st National Conf., ACM, Los Angeles, Huesmann, L. R. and R. P. Goldberg, Evaluating Computer Calif., Aug. 30-Sept. 1, 1966, pp. 339-345 (Thompson Book Co., Systems Through Simulation, The Computer J. 10, 150-156 Washington, D.C., 1966). (Aug. 1967). Hellerman, H., Some Principles of Time-Sharing Scheduler Hughes Dynamics, Inc., Methodologies for System Design, Final Strategies, IBM Sys. J. 8, No, 2, 94-117 (1969). Report for Contract No, AF30(602)-2620, Rept. No. RADC- Henle, R. A. and L. 0. Hill, Integrated Computer Circuits - TDR-63-486, Vol. 1, 1 v. (Los Angeles, Calif., Feb. 24, 1964). Past, Present, and Future, Proc. IEEE 54, 1849-1860 (Dec. Hughes Dynamics, Inc., The Organization of Large Files, 1966). Part 1-VI, (Advance Inf. System Div., Hughes Dynamics, Inc., Hennis, R. B., Recognition of Unnurtured Characters in a Multi- Sherman Oaks, Calif., Apr. 1964). font Application, Tech. Pub. 07.212, 13 p. (IBM Systems Huskey, H. D., On-Line Computing Systems: A Summary, in Development Div., Lab., Rochester, Minn., May 9, 1967). On-Line 1.omputing Systems, Proc. Symp. sponsored by the Henry, W. R., Hierarchical Structure for Data Management, Univ, of California, Los Angeles, and Informatics, Inc., Feb. IBM Sys. J. 8, No. 1, 2-15 (1i969). 1965, Ed. E. Burgess, pp. 139-142 (American Data Proc., Inc., Hershey, A. V., Calligraphy for Computers, NWL Rept. No. Detroit, Mich., 1965). 2101, 1 v. (U.S. Naval Weapons Lab., Dahlgren, Va., Aug. 1, Hutchinson, G. K. and J. N. Maguire, Computer Systems Design 1967). and Analysis Through Simulation, AFIPS Proc. Fall Joint Hickey, P. R., The A-1200: A Narrow Band System Test Vehicle, Computer Conf., Vol. 27, Pt. 1, Las Vega;;, Nev., Nov. 30- in Data Processing, Vol. X, Proc. 1966 Int. Data Proc. Conf., Dec. 1, 1965, pp. 161-167 (Spartan Books, Washington, D.C., Chicago, Ill., June 21-24, 1966, pp. 175-187 (Data Proc. 1965). Management Assoc., 1966). Hillegass, J. R., NCR's New Bid: The Century Series, Data Proc. IBM System/360, Data Proc. 7, No. 5, 290-302 (Sept.-Oct. 1965). Mag. 10, No. 4, 96-49, 52-54 (Apr. 1968). The Impact of integrated Circuits on the Computer Field, Hillegass, J. R. and L. F. Melick, A Survey of Data Collection Computers & Automation 14, No. 7, 9-10 (1965). Systems, Data Proc. Mag. 9, No. 6, 50-56 (June 1967). Impact of LSI on the Next Generation of Computers (Panel Hirsch, P. M., J. A. Jordan, Jr. and L. B. Lesem, Digital Con- Discussion),Participants:J.P.Eckert,G.Hollander, struction of Holograms, Proc. IFIP Congress 68, Edinburgh, M. Palevsky and B. Pollard, Computer Design 8, No. 6, 48, Scotland, Aug. 5-10, 1r...68,.Booklet H, pp. H 104.-H 109 (North- 50-59 (June 1969). Holland Pub. Co., Amsterdam, 1968). Investigation of Inorganic Phototropic Materials as a Bi-Optic Hoagland, A. S., Storing Computer Data, Int. Sci. & Tech. No. Element ApplicableinHigh Density Storage Computer 37, 52-58 (Jan. 1965). Memories, Rept. No. ASD-TDR-62-305, 50 p. (Aeronautical Hoagland, A. S., Mass Storage Revisited, Proc. 22nd National Systems Div., Electronic Technology Lab., Wright-Patterson Conf., ACM, Washington, D.C., Aug. 29-31, 1967, pp. 255-260 AFB, Ohio, Apr. 1962). (Thompson Book Co., Washington, D.C., 1967). Israel, D. R., System Engineering Experience with Automated Hobbs, L. C., The Impact of Hardware in the 1970's, Datamation Command and Control Systems, in Information System Science 12, No. $, 36-44 (Mar. 1966). and Technology, papers prepared for the Third Cong., sched- Hobbs, L, C., Display Applications and Technology, Proc. uled for Nov. 21-22, 1966, Ed. D. E. Walker, pp. 193-213 IEEE 54,1870-1884 (Dec. 1966). (Thompson Book Co., Washington, D.C., 1967). Hoffman, A., The Information and Data Exchange Experimental Activities (IDEEA) Program and Its Relation to the National Interests, in Toward a National Information System, Second Jacobellis, B. R., Impact of Computer Technologyon Communi- AnnualNationalColloquiumonInformationRetrieval, cations, Proc. ACM 19th National Conf., Philadelphia, Pa., Philadelphia, Pa., April 23-24, 1965, Ed. M. Rubinoff, pp. Aug. 25-27, 1964, pp. N2. 1-1 to N2. 1-4 (Assoc. for Computing 87-104 (Spartan Books, Washington, D.C., 1965). Machinery, New York, N.Y., 1964). Holland, G., Reactive Terminal Service from ITT, Data Proc. Jacobs, I. S. and E. G. Spencer, Eds., Proc. 10th Conferenceon 11, No. 1, 65-67 (Jan.-Feb. 1969). Magnetism and Magnetic Materials, Minneapolis, Minn., Nov. Holland, J., A Universal Computer Capable of Executing an 16-19, 1964. Appeared as Part 2 of the Journal of Applied Arbitrary Number of Sub-Programs Simultaneously, Poc. Physics 36, No. 3, 877-1280 (Mar. 1965). Eastern ,Joint Computer Conf., Vol. 16, Boston, Mass., Dec. Jacobs, J. F., Communication in the Design of Military Informa- 1-3, 1959, pp. 108-113 (Eastern Joint Computer Conf., 1959). tion Systems, in Military Information Systems, Ed. E. Bennett Hormann, A. M., Introduction to ROVER, an Information Proc- et al., pp. 29-45 (Frederick A. Praeger, Pub., New York, 1964). essor, Rept. No. FN-3487, 51 p. (System Development Corp., Jacoby, K. Isolation of Control Malfunctions ina Digital Com- Santa Monica, Calif., Apr. 25, 1960). puter, preprints of summaries of papers presented at the 14th Horne, W. H., A. S. Sabin and J. D. Wells, Random Access National Meeting, ACM, Cambridge, Mass., Sept. 1-3, 1959, Communications for the Safety Services, in Law Enforcement pp. 7-1 to 7-4 (Assoc. for Computing Machinery, New York, Science and Technology, Vol. 1, Proc. First National Symp. 1959). on Law Enforcement Science and Technology, Chicago, Ill., Jaspert, W. P., Ed., Advances in Computer Typesetting, Proc. March 1967, Ed. S. A. Yefsky, pp. 115-123 (Thompson Book Int. Computer Typesetting Conf., Sussex, England, July 14-18, Co., Washington, D.C., 1967). 1966, 306 p. (The Institute of Printing, London, 1967). 135 Jensen, P. A., A Graph Decomposition Technique for Structuring King, G. W., The Library of Congress Project, Lib. Res. & Tech. Data, Rept. No. 106-3, 1 v. (Computer Command and Control Serv. 9, No. 1, 90-93 (Winter 1965). Co., Washington, D.C., Sept. 1, 1967). Knowlton, K. C., A Programmer's Description of Lo, Commun. Johnson, H. R., Computers and the Public Welfare, Law Enforce- ACM 9, No. 8, 616-625 (Aug. 1966). merit, Social Services and Data Banks, preprint, paper for Knuth, D. E., and J. E, McNeley, SOL: A Symbolic Language for UCLA Conf. on Computers and Communications, 21p., General Purport' Systems Simulation, IEEE Trans. Electron. (OfficeofTelecommunicationsManagement,Executive Computers EC-13, No. 4, 401 (Aug. 1964). Office of the President, Washington, D.C., 1967). Kochen, M., Ed., Some Problems in Information Science, 309 p. Jones, A. H., Time - Sharing -An Accepted Technique, Data (The Scarecrow Press, Inc., New York, 1965). Proc. 11, No. 1, 68-69 (Jan.-Feb. 1969). Kohn, G., Future of Magnetic Memories, in Information Process- Jones, J. D., Electronic Data Communications, in Data Processing ing 1965, Proc. IFIP Congress 65, Vol. 1, New York, N.Y., May Yearbook 1965, pp. 63-66 (American Data Processing, Inc., 24-29, 1965, Ed. W. A. Kalenich, pp. 131-136 (Spartan Detroit, Mich., 1964). Books,, Washington, D.C., 1965). Jones, R. D The Public Planning Information System and the Kohn, G., W. Jutzi, Th. Mohr and D. Seitzer, A Very-High- Computer Utility, Proc. 22nd National Conf., ACM, Washing- Speed, Nondestructive-Read Magnetic Film Memory, IBM ton, D.C., Aug. 29-31, 1967, pp. 553-564 (Thompson Book Co., J. Res. & Dev. 1 1,No. 2, 162-168 (Mar. 1967). Washington, D.C., 1967). Kosonocky, W. F. and R. H. Cornely, Lasers for Logic Circuits, Jones, R. H. and E. E. Bittmann, The B8500-Microsecond Thin- 1968 WESCON Technical Papers, 16/4, Aug. 1968, 10 p. Film Memory, AFIPS Proc. Fall Joint Computer Conf., Vol. Kovalevsky, V. A., Present and Future of Pattern-Recognition 31, Anaheim, Calif. Nov. 14-16, 1967, pp. 347-352 (Thompson Theory, in Information Processing 1965, Proc. IFIP Congress Books, Washington, D.C., 1967). 65, Vol. 1, New York, N.Y., May 24-29, 1965, Ed. W. A. Joseph, E. C., Computers: Trends Toward the Future, Proc. Kalenich, pp. 37-43 (Spartan Books, Washington, D.C., 1965). IFIP Congress 68, Edinburgh, Scotland, Aug. 5-10, 1968, Kozmetsky, G. and P. Kircher, Electronic Computers and Man- Vol. 1, Invited Papers, pp. 145-157 (North-Holland Pub. Co., agement Control, 296 p. (McGraw-Hill, New York, 1956). Amsterdam, 1968). Kroger, M. G., Introduction to Tactical Information Systems, Joyce, J. D. and M. J. Cianciolo, Reactive Displays: Improving in Second Cong. on the Information System Sciences, Hot Man-Machine Graphical Communication, AFIPS Proc. Fall Springs, Va., Nov. 1964, Ed. J. Spiegei and D. E. Walker, pp. Joint Computer Conf., Vol. 31, Anaheim, Calif., Nov. 14-16, 267-273 (Spartan Books, Washington, D.C., 1965). 1967, pp. 713-721 (Thompson Books, Washington, D.C., 1967). Kump. H. J. and P. T. Chang, Thermostrictive Recording on Justice, B. and F. B. Liebold, Jr., Photochromic Glass- A New Permailoy Films, IBM J. Res. & Dev. 10, No. 3, 255-260 Tool for the Display System Designer, Inf. Display 2, 23-28 (May 1966). (Nov.-Dec. 1965). Kuney, J. H., Analysis of the Role of the Computer in the Repro- duction and Distribution of Scientific Papers, in Automation and Scientific Communication, Short Papers, Pt. 2, papers Kalenich, W. A., Ed Information Processing 1965, Proc. IFIP contributed to the Theme Sessions of the 26th Annual Meeting, Congress 65, Vol. 1, New York, N.Y., May 24-29, 1965, 304 p. Am. Doc. Inst., Chicago, Ill., Oct. 6-11, 1963, Ed. H. P. Luhn, (Spartan Books, Washington, D.C., 1965). pp. 249-250 (Am. Doc. Inst., Washington, D.C., 1963). Kalenich, W. A., Ed., Information Processing 1965, Proc. IFIP Kuney, J.H. and B. G. Lazorchak, Machine-Set Chemical Congress 65, Vol. 2, New York, N.Y., May 24-29, 1965. pp. Structures, in Parameters of Information Science, Proc. Am. 305-648 (Spartan Books, Washington, D.C., 1966). Doc. Inst. Annual Meeting, Vol. 1, Philadelphia, Pa., Oct. Kaplan, D. C. and C. F. Kooi, Magnetostatic Echoes, Tech. Rept. 5-8, 1964, pp. 303-305 (Spartan Books, Washington, D.C., No. 8 (Lockheed Palo Alto Research Las., Palo Alto, Calif., 1964). Aug. 1, 1966). Karush, W., On the Use of Mathematics in Behavioral Research, Lang, C. A., R. B. Polansky and D. T. Ross, Some Experiments in Natural Language and the Computer, Ed. P. L. Garvin, with an Algorithmic Graphical Language, Rept. No. ESL- pp. 67-83 (McGraw-Hill, New York, 1963). TM-220, 55 p. (Massachusetts Institute of Technology, Cam- Kaufman, B. A., P. B. Ellinger and H. J. Kuno, A Rotationally bridge, Aug. 1965). Switched Rod Memory with a 100-Nanosecond Cycle Time, Larsen, R. P. and M. M. Mano, Modeling and Simulation of AFIPS Proc. Fall Joint Computer Conf., Vol. 29, San Francisco, Digital Networks, Commun. ACM 8, No. 5, 308-312 (May Calif., Nov. 7-10, 1966, pp. 293-304 (Spartan Books, Washing- 1965). ton, D.C., 1966). "Laser Applied to Molecular Kinetics Studies", NBS Tech. Kautz, W. H., K. N. Levitt and A. Waksman, Cellular Intercon- News Bull. 52, No. 11, 242-257 (Nov. 1968). nection Arrays, IEEE Trans. Computers C-17, 443-451 (May Laurance, N., A Compiler Language for Data Structures, Proc. 1968). 23rd National Conf., ACM, Las Vegas, Nev., Aug. 27-29, Kay, M., F. Valadez and T. Ziehe, The Catalog Input/Output 1968, pp. 387-394a (Brandon/Systems Press, Inc., Princeton, System, Memo. RM-4540-PR, 64 p. (The RAND Corp., Santa N.J., 1968). Monica, Calif., Mar. 1966). The Lavish Laser, Vectors (Hughes Aircraft Co.) VIII, 15-17, Kay, R. H., The Management and Organization of Large Scale Fourth Quarter, 1966. Software Development Projects, AFIPS Proc. Spring Joint Lawson, H. W., Jr., PL/I List Processing, Commun. ACM 10, Computer Conf., Vol. 34, Boston, Mass., May 14-16, 1969, 358-367 (June 1967). pp. 425-433 (AFIPS Press, Montvale, N.J., 1969). Ledley, R. S., J. Jacobsen and M. Belson, BUGSYS: A Program- Kent, A., Ed., Information Retrieval and Machine Translation, ming System for Picture Processing -Not for Debugging, Pt. I, Proc. Int. Conf. for Standards on a Common Language Commun. ACM 9, 79-84 (Feb. 1966). for Machine Searching and Translation, Western Reserve Lee, R. W., and R. W., Worral, Eds., Electronic Composition in Univ., Cleveland, Ohio, Sept. 6-12, 1959, 686 p. (International Printing, Proc. Symp. Gaithersburg, Md., June 15-16, 1967, Pub. Inc., New York, 1960). NBS Special Pub. 295, 128 p. (U.S. Govt. Print. Off., Washing- Kesselman, M. L., How Do We Stand on the Big Board?, AFIPS ton, D.C., Feb. 1968). Proc. Fall Joint Computer Conf., Vol. 31, Anaheim, Calif., Lehrer, N. H. and R. D. Ketchpel, Recent Progress on a High- Nov. 14-16, 1967, pp. 161-167 (Thompson Books, Washington, Resolution, Meshless, Direct-View Storage Tube, AFIPS Proc. D.C., 1967). Fall Joint Computer Conf., Vol. 29, San Francisco, Calif., Kessler, M. M., An Experimental Communication Center for Nov. 7-10, 1966, pp. 531-539 (Spartan Books, Washington, Scientific and Technical Information, 19 p. (Lincoln Lab., D.C., 1966). Massachusetts Institute of Technology, Lexington, Mar. 31, Leith, E. N., A. Kozma and J. Upatnieks, Coherent Optical 1960). Systems for Data Processing, Spatial Filtering, and Wavefront King, G. W., Ed., Automation and the Library of Congress, a Reconstruction, in Optical and Electro-Optical Information survey sponsored by the Council on Library Resources, Inc., Processing, Ed., J. R. Tippett et al., pp. 143-158 (M.I.T. Press, 88 p. (Library of Congress, Washington, D.C., 1963). Cambridge, Mass., 1965). 136 Lesem, L. B., P. M. Hirsch and J. A. Jordan, Jr., Holographic System, Second Annual National Colloquium on Information Display of Digital Images, Proc. 22nd National Conf., ACM, Retrieval, Philadelphia, Pa., April 23-24, 1905, Ed. M. Rubinoff, Washington, D.C., Aug. 29-31, 1967, pp. 41-47 (Thompson pp. 199-215 (Spartan Books, Washington, D.C., 1965). Book Co., Washington, D.C., 1967). Mahan, R. E., A State of the Art Survey of the Data Display Field, Lesem, L. B., P. M. Hirsch and J. A. Jordan, Jr., The Kinoform: A Rept. No. BNWL-725, AEC Research and Development New Wavefront Reconstruction Device, IBM J. Res. & Dev. Report, Battelle Northwest, Richland, Wash., May 1968, 1 v. 13, No. 2, 150-155 (Mar. 1969). Maiman, T. H., Stimulated Optical Radiation in Ruby, Nature Levien, R. and M. E. Maron, Relational Data File: A Tool for 187, 493-494 (Aug. 1960). Mechanized Inference Execution and Data Retrieval, Rept. Malbrain, J. P., Automated Computer Design, in Preprints of No. RM-4793-PR, 89 p. (The RAND Corp., Santa Monica, Summaries of Papers Presented at the 14th National Meeting, Calif., Dec. 1965). ACM, Cambridge, Mass., Sept. 1-3, 1959, pp. 4-1 to 4-2 Lewin, M. H., H. R. Beelitz and J. Guarracini, Fixed Resistor- (Assoc. for Computing Machinery, New York, 1959). Card Memory, IEEE Trans. Electron. Computers EC-14, Marill, T. and L. G. Roberts, Toward a Cooperative Network of 428-434 (1965). Time-Shared Computers, AFIPS Proc. Fall Joint Computer Licklider, J. C. R., Artificial Intelligence, Military Intelligence, Conf., Vol. 29, San Francisco, Calif., ,Nov. 7-10, 1966, pp. and Command and Control, in Military Information Systems, 425-431 (Spartan Books, Washington, D.C., 1966). Ed. E. Bennett et al., pp. 118-133 (Frederick A. Praeger, Pub., Markus, J. V., State of the Art of Published Indexes, Am. Doc. New York, 1964). 13, No. 1, 15-30 (Jan. 1962). Licklider, J. C. R., Man-Computer Interaction in Information Markuson, B. E., Ed., Libraries and Automation, Proc. Conf. Systems, in Toward a National Information System, Second held at Airlie Foundation, Warrenton, Va., May 26-30, 1963, Annual National Colloquium on Information Retrieval, Phila- 368 p., under sponsorship of the Library of Congress, the delphia, Pa., April 23-24, 1965, Ed. M. Rubinoff, pp. 63-75 National Science Foundation, and the Council on Library (Spartan Books, Washington, D.C., 1965). Resources (Library of Congress, Washington, D.C., 1964). Licklider, J. C. R., Libraries of the Future, 219 p. (M.I.T. Press, Markuson, B.E., Automation inLibraries and Information Cambridge, Mass., 1965). Centers, in Annual Review of Information Science and Tech- Licklider, J. C. R., Man-Computer Partnership, Int. Sci. & Tech. nology, Vol. 2, Ed. C.A. Cuadra, pp. 255-284 (Interscience 41, 18-26 (1965). Pub., New York, 1967). Licklider,J.C.R., Interactive Information Processing, in Matick, R. E., P. Pleshko, C. Sie and L. M. Terman, A High- Computer and Information Sciences-II, Proc. 2nd Symp. on Speed Read-Only Store Using Thick Magnetic Films, IBM Computer and Information Science;;, Columbus, Ohio., Aug. J. Res. & Dev. 10, No. 4, 333-342 (July 066). 22-24, 1966, Ed. J. T. Tou, pp. 1-13, (Academic Press, New Mazzarese, N. J., Experimental System Provides Flexibility and York, 1967). Continuity, Data Proc. Mag. 7, No. 5, 68-70 (1965). Licklider, J. C. R. and W. E. Clark, On-Line Man-Computer McCallister, J. P. and C. F. Chong, A 500-Nanosecond Main Communications, AFIPS Proc. Spring Joint Computer Conf., Computer Memory Utilizing Plated-Wire Elements, AFIPS Vol. 21, San Francisco, Calif., May 1-3, 1962, pp. 113-128 Proc. Fall Joint Computer Conf., Vol. 29, San Francisco, Calif., National Press, Palo Alto, Calif., 1962). Nov. 7-10, 1966, pp. 305-314 (Spartan Books, Washington, Linde, R. R. and P. E. Chaney, Operational Management of D.C., 1966). Time-Sharing Systems, Proc. 21st National Conf., ACM, Los McCamy, C. S., Photographic Standardization and Research at Angeles, Calif., Aug. 30-Sept. 1, 1966, pp. 149-159 (Thomp- the National Bureau of Standards, Applied Optics 6, No. 1, son Book Co., Washington, D.C., 1966). 27-30 (Jan. 1967). Lindquist, A. B., R. R. Seeber and L. W. Comeau, A Time- McCarthy, J., Problems in the Theory of Computation, in Infor- Sharing System Using an Associative Memory, Proc. IEEE 54, mation Processing 1965, Proc. IFIP Congress 65, Vol. 1, New 1774-1779 (Dec. 1966). York, N. Y., May 24-29, 1965, Ed. W. A. Kalenich, pp. 219- Lo, A. W., High-Speed Logic and Memory -Past, Present and 222 (Spartan Books, Washington, D.C., 1965). Future, AFIPS Proc. Fall Joint Computer Conf., Vol. 33, Pt. 2, McCarthy, J., S. Boilen, E. Fredkin and J. C. R. Licklider, A San Francisco, Calif., Dec. 9-11, 1968, pp. 1459-1465 (Thomp- Time-Sharing Debugging System for a Small Computer, AFIPS son Book Co., Washington, D.C., 1968). Proc. Spring Joint Computer Conf., Vol. 23, Detroit, Mich., Lock; K., Structuring Programs for Multiprogram Time-Sharing May 1963, pp. 51-57 (Spartan Books, Baltimore, Md., 1963). On-Line Applications, AFIPS Proc. Fall Joint Computer Conf., McClure, R. M., TMG- A Syntax Directed Compiler, Proc. 20th Vol. 27, Pt. 1, Las Vegas, Nev., Nov., 30-Dec. 1, 1965, pp. National Conf., ACM, Cleveland, Ohio, Aug. 24-26, 1965, pp. 457-472 (Spartan Books, Washington, D.C., 1965). 262-274 (Lewis Winner, New York, 1965). Loomis, H. H., Jr., Graphical Manipulation Techniques Using McDermid, W. L. and H. E. Petersen, A Magnetic Associative the Lincoln TX-2 Computer, Rept. No. 51-G-0017, 27 p. Memory System, IBM J. Res. & Dev. 5, No. 1, 59-62 (Jan. (Linco1'n Lab., Massachusetts Institute of Technology, Lex- 1961). ington, Nov. 10, 1960). McFarland, K. and M. Hashiguchi, Laser Recording Unit for High Luhn, H. P. Ed., Automation and Scientific Communication, Density Permanent Digital Data Storage, AFIPS Proc. Fall Short Papers, Pt. 2, papers contributed to the Theme Sessions Joint Computer Conf., Vol. 33, Pt. 2, San Francisco, Calif., of the 26th Annual Meeting, Am. Doc. Inst., Chicago, Ill., Dec. 9-11, 1968, pp. 1369-1380 (Thompson Book Co., Wash- Oct. 6-11, 1963, pp. 129-352 (Am. Doc. Inst., Washington, ington, D.C., 1968). D.C., 1963). McGee, W. C., File Structures for Generalized Data Management, Lynch, W. C., Description of a High Capacity, Fast Turnaround Proc. IFIP Congress 68, Edinburgh, Scotland, Aug. 5-10, University Computing Center, Commun. ACM 9, No. 2, 117- 1968. Booklet F. pp. F 68-F 73 (North-Holland Pub. Co., 123 (Feb. 1966). Amsterdam, 1968). McGee, W. C. and H. E. Petersen, Microprogram Control for MacDonald, N., A Time Shared Computer System -The Dis- the Experimental Sciences, AFIPS Proc. Fall Joint Computer advantages, Computers and Automation 14, No. 9, 21-22 Conf., Vol. 27, Pt. 1, Las Vegas, Nev., Nov. 30-Dec. 1, 1965, (Sept. 1965). pp. 77-91 (Spartan Books, Washington, D.C., 1965). Machover, C., Graphic CRT Terminals- Characteristics of Commercially Available Equipment, AFIPS Proc. Fall Joint McMains, H. J.,Electrical Communications in the Future, Computer Conf., Vol. 31, Anaheim, Calif., Nov. 14-16, 1967, Datamation 12, No. 11, 28-30 (Nov. 1966). pp. 149-159 (Thompson Books, Washington, D.C., 1967). Meadow, C. T., The Analysis of Information Systems: A Pro- Madnick, S. E., Multi-Processor Software Lockout, Proc. 23rd grammer's Introductionto Information Retrieval, 301p. National Conf., ACM, Las Vegas, Nev., Aug. 27-29, 1968, Wiley, New York, 1967). pp. 19-24 (Brandon/Systems Press, Inc., Princeton, N.J., Meddaugh, S. A. and K. L. Pearson, A 200-Nanosecond Thin 1968). Film Main Memory System, AFIPS Proc. Fall Joint Computer Magnino, J. J., Jr IBM Technical Information Retrieval Center- Conf., Vol. 29, San Francisco, Calif., Nov. 7-10, 1966, pp. Normal Text Techniques, in Toward a National Information 281-292 (Spartan Books, Washington, D.C., 1966). 137 Me lick, L. F., The Computer Talks Back, Data Proc. Mag. 8, Newell, A., The Search for Generality, in Information Processing 58-62 (Oct. 1966). 1965, Proc. IFIP Congress 65, Vol. 1, New York, N.Y., May Menkhaus, E, J., The Ways and Means of Moving Data, Bus. 24-29, 1965, Ed. W. A. Kalenich, pp. 17-24 (Spartan Books, Automation 14, 30-37 (Mar. 1967). Washington, D.C., 1965), Merryman, J. D., Making Light of the Noise Problem, Electronics Newell, A. and H. A. Simon, The Simulation of Human Thought, 3 8, No. 15, 52-56 (1965). Rept. No. RM-2506, 40 p. (The RAND Corp., Santa Monica, Miller, H. S., Resolving Multiple Responses in an Associative Calif., Dec, 28, 1959). Memory, IEEE Trans. Electron. Computer EC-13, No. 5, Newman, S. M., R. W. Swanson and K. C. Knowlton, A Notation 614-616 (Oct. 1964). System for Transliterating Technical and Scientific Texts for Miller, L., J. Miuker, W. G. Reed and W. E. Shindle, A Multi- Use in Data Processing Systems, in Information Retrieval Level File Structure for Information Processing, Proc. Western and Machine Translation, Pt. I, Proc. Int. Conf. for Standards Joint Computer Conf., Vol. 17, San Francisco, Calif., May on a Common Language for Machine Searching and Transla- 3-5, 1960, pp. 53-59 (Pub. by Western Joint Computer Conf., tion, Western Reserve Univ Cleveland, Ohio, Sept. 6-12, 1959, San Francisco, Calif., 1960). Ed. A. Kent, pp. 345-376 (International Pub., Inc., New York, Mills, R. G., Communications Implication of the Project MAC 1960). Multiple-Access Computer System, 1965 IEEE International Newman, W. M., A System for Interactive Graphical Program. Cony. Rec., Pt. 1, pp. 237-241. ming, AFIPS Proc. Spring Joint Computer Conf., Vol. 32, Mills, R. G. Man-Computer Interaction-Present and Future, Atlantic City, N.J., Apr. 30-May 2, 1968, pp. 47-54 (Thompson 1966 IEEE International Cony. Rec., Pt. 6, pp. 196-198. Book Co., Washington, D.C., 1968). Mills, R. G. Man-Machine Communication and Problem Solving, Ninke, W. H., Graphic 1- A Remote Graphical Display Console, in Annual Review of Information Science and Technology, AFIPS Proc. Fall Joint Computer Conf., Vol. 27, Pt. 1, Las Vol. 2, Ed. C. A. Cuadra, pp. 223-254 (Interscience Pub., New Vegas, Nev., Nov. 30-Dec. 1, 1965, pp. 839-855 (Spartan Books, York, 1967). Washington, D.C., 1965). Minker, J. and J. Sable, File Organization and Data Management, Nisenoff, N., Hardware for Information Processing Systems: in Annual Review of Information Science and Technology, Today and in the Future, Proc. IEEE 54, 1820-1835 (Dee. Vol. 2, Ed. C. A. Cuadra, pp. 123-160 Interscience Pub., 1966). New York, 1967). Nolan, J. and L. Yarbrough, An On-Line Computer Drawing and Minnick, R. C., A Survey of Microcellular Research, J. ACM 14, Animation System, Proc. IFIP Congress 68, Edinburgh, No. 2, 203-241 (Apr. 1967). Scotland, Aug. 5-10, 1968, Booklet C, pp. C 103-C 108 (North- Mooers, C. N., Information Retrieval Selection Study, Part II: Holland Pub. Co., Amsterdam, 1968). Seven System Models, Rept. No. ZTB-133, Part II, 39 p. North,A.,Typewriter-to-Computer Roster Publication and (Zator Co., Cambridge, Mass., Aug. 1959). Maintenance, in Electronic Composition in Printing, Proc. Mooers, C. N., TRAC, A Procedure-Describing Language for the Symp., Gaithersburg, Md., June 15-16, 1967, NBS Special Reactive Typewriter, Commun. ACM 9, No. 3, 215-219 Pub. 295, Ed. R. W. Lee and R. W. Worral, pp. 107-111 (U.S. (Mar. 1966). Govt. Print. Off., Washington, D.C., Feb. 1968). Mooers, C. N. and L. P. Deutsch, TRAC, A Text, Handling Nugent, W. R., A Machine Language for Documentation and Language, Proc. 20th National Conf., ACM, Cleveland, Ohio, Information Retrieval, in Preprints of Summaries of Papers Aug. 24-26, 1965, pp. 229-246 (Lewis Winner, New York, Presented at the ,14th National Meeting, ACM, Cambridge, 1965). Mass., Sept. 1-3, 1959, pp. 15-1 to 15-4 (Assoc. for Computing Moravec, A. F.,F. Basic Concept.; for Planning an Electronic Data Machinery, New York, 1959). Processing AFIPS Proc. Fall Joint Computer Conf., Nugent, 'W. R., A Machine Language for Document Translitera- Vol. 27, Pt. 1, Las Vegas, Nev., Nov. 30-Dec. 1, 1965, pp. tion, Preprint of paper presented at the 14th National Meeting, 169-184 (Spartan Books, Washington, D.C., 1965). ACM, Cambridge, Mass., Sept. 1-3, 1959, 33 p. Morenoff, E. and J. B. McLean, A Code for Non-numeric Informa- tion Processing Applications in Online Systems, Commun. Oettinger,, A. G., Automatic Processing of Natural and Formal ACM 10, No. 1, 19-22 (Jan. 1957). Languages,inInformation Processing1965, Proc. IFIP Morenoff, E. and J. B. McLean, On the Standardization of Com- Congress 65, Vol. 1, New York, N.Y., May 24-29, 1965, Ed. puter Systems, Rept. No. RADC-TR-67-165, 14 p. (Rome Air W. A. Kalenich, pp. 9-16 (Spartan Books, Washington, D.C., Development Center, Griffiss AFB., New York, Mar. 1967). 1965). Morris, D., F. H. Sumncr and M. T. Wyld, An Appraisal of the Ohlman, H., State-of-the-Art: Remote Interrogation of Stored Atlas Supervisor, Proc. 22nd National Conf., ACM, Wash- Documentary Material, in Automation and Scientific Com- ington, D.C., Aug., 29-31, 1967, pp. 67-75 (Thompson Book munication, Short Papers, Pt. 2, papers contributed to the Co., Washington, D.C., 1967). Theme Sessions of the 26th Annual Meeting, Am. Doc. Inst., Moulton, P. G. and M. E. Muller, DITRAN- A Compiler Empha- Chicago, Ill., Oct. 6-11, 1963, Ed. H. P. Luhn, pp. 193-194 sizing Diagnostics, Commun. ACM 10, No. 1, 45-52 (Jan. (Am. Doc. Inst., Washington, D.C., 1963). 1967). Ohringer, L., Accumulation of Natural Language Text for Muckier, F. A. and R. W. Obermayer, Information Display, Int. Computer Manipulation, in Parameters of Information Science, Sci. & Tech. 44, 34-40 (Aug. 1965). Proc. Am. Doc. Inst., Annual Meeting, Vol. 1, Philadelphia, Murrill, D. P., Microfilming and Encoding Laboratory Notebooks Pa., Oct. 5-8, 1964, pp. 311-313 (Spartan Books, Washington, at the Philip Morris Research Center, in Progress in Informa- D.C., 1964). tion Science and Technology, Proc. Am. Doc. Inst. Annual Oldham, I. B., R. T. Chien and D. T. Tang, Error Detection and Meeting, Vol. 3, Santa Monica, Calif., Oct. 3-7, 1966, pp. Correction in a Photo-Digital Storage System, IBM J. Res. 51-56 (Adrianne Press, 1966). & Dev. 12, No. 6, 422-430 (Nov. 1968). Olsen, T. M., Philco/IBM Translation at Problem-Oriented, Narasimhan, R., Syntax-Directed Interpretation of Classes of Symbolic and Binary Levels, Commun. ACM 8, No. 12, Pictures, Commun. ACM 9, No. 3, 166 -173 (March 1966). 762-768 (Dec. 1965). Nathan, M. I., Semiconductor Lasers, Proc. IEEE 54,1276-1290 Ophir, D., B. J. Shepherd and R. J. Spinrad, Three-Dimensional (Oct. 1966). Computer Display, Commun. ACM 12, No. 6, 309-310 Naur, P., The Place of Programming in a World of Problems, (June 1969). Tools and People, in Information Processing 1965, Proc. IFIP Opler, A., Dynamic Flow of Programs and Data Through Hier- Congress 65, Vol. 1, New York, N.Y., May 24-29, 1965, Ed. archical Storage, in Information Processing 1965, Proc. IFIP W. A. Kalenich, pp. 195-199 (Spartan Books, Washington, Congress 65, Vol. 1, New York, N.Y., May 24-29, 1965, Ed. D.C., 1965). W. A. Kalenich, pp. 273-276 (Spartan Books, Washington, Newberry, S. P., An Electron Optical Technique for Large- D.C., 1965). Capacity Random-Access Memories, AFIPS Proc. Fall Joint Opler, A., Procedure-Oriented Language Statements to Facilitate Computer Conf., Vol. 29, San Francisco, Calif Nov. 7-10, Parallel Processing, Commun. ACM 8, No. 5, 306-307 (May 1966, pp. 717-728 (Spartan Books, Washington, D.C., 1966). 1965). 138 Op ler, A., Requirements for Real-Time Languages, Commun. Petschauer, R. J., Magnetics-Still the Best Choicefor Com- ACM 9, No. 3, 196-199 (Mar. 1966). puter Main Memory, AFIPS Proc. Fall Joint Computer Conf. Op ler, A., New Directions in Software 1960-1966, Proc. IEEE Vol. 31, Anaheim, Calif., Nov. 14-16, 1967,pp.598-600 54, 1757-1763 (Dec. 1966). (Thompson Books, Washington, D.C., 1967). Oppenheimer, G. and N. Weiler, Resource Managementfor a Pfaltz, J. L., J. W. Snively, Jr. and A. Rosenfeld,Local and Medium Scale Time-Sharing Operating System, Commun. Global Picture Processing by Computer,in Pictorial Pattern ACM 11, No. 5, 313-322 (May 1968). Recognition, Proc. Symp. on Automatic Photointerpretation, Orthard-Hays, W., Operating Systems for Job-to-JobRunning Washington, D.C., May 31-June 2, 1967, Ed. G. C. Cheng and for Special Applications: Differences andSimilarities, in et al., pp. 353-371 (Thompson Book Co., Washington, D.C., Information Processing 1965, Proc. 1FIP Congress65, Vol. 1, 1968). New York, N.Y., May 24-29, 1965, Ed. W. A. Kalenich,pp. Pick, G. G. and D. B. Brick, A Read-Only Multi-MegabitParallel 237-242 (Spartan Books, Washington, D.C., 1965). Search Associative Memory, in Automation andScientific O'Sullivan, T. C., Exploiting the Time-SharingEnvironment, Communication, Short Papers, Pt. 2,papers contributed to the Proc. 22nd National Conf., ACM,Washington, D.C., Aug. Theme Sessions of the 26th Annual Meeting, Am. Doc.Inst. 29-31, 1967, pp. 169-175 (Thompson Book Co.,Washington, Chicago, III., Oct. 6-11, 1963, Ed. H. P. Lulu'',pp.245-246 D.C., 1967). (Am. Doc. Inst., Washington, D.C., 1963). O'Toole, G., Preparing for Data Communications,Computers & Pick, G. G., S. B. Gray and D. B. Brick, The SolenoidArray- A Automation 18, No. 5, 33-35 (May 1969). New Computer Element, IEEE Trans. Electron.Computers EC-13, 27-35 (Feb. 1964). Pankhurst, R. J., GULP -A Compiler-Compilerfor Verbal and Porter, J. W. and L. E. Johnson, United Air Lines'Electronic Graphic Languages, Proc. 23rd National Conf.,ACM, Las Information System (EIS), in Data Processing, Vol. X,Proc. Vegas, Nev., Aug. 27-29, 1968,pp. 405-421 (Brandon/Systems 1966 Int. Data Processing Conf., Chicago, 111., June Press, Inc., Princeton, N.J., 1964 21-24, 1966, pp. 74-82 (Data Processing ManagementAssoc.,966). Paris, D. P., Digital Simulation of Image-FormingSystems, IBM Potter, R. J. and A. A. Axelrod, Optidal Input/Output J. Res. & Dev. 10, No. 5, 407-411 (Sept. 1966). gystems for Computers, 1968 WESCON Technical Papers,16/2, Parker, D. C. and M. F. Wolff, Remote Sensing,Int. Sci. Tech. Aug. 1968, 6 p. 43, 20-31 (1965). Pravikoff, V. V., Program Documentation, Data Proc.Mag. 7, Parnas, D. L., A Language for Describingthe Functions of 44-45 (Oct. 1965). Synchronous Systems, Commun. ACM 9, No. 2,72-76 (Feb. Press, L. I. and M. S. Rogers, IDEA-A Conversational,Heuristic 1966). Program for Inductive Data Exploration Patrick, R. L., So You Want to Go Online?, and Analysis, Proc. Datatnation 9, No. 10, 22nd National Conf., ACM, Washington, D.C.,Aug. 29-31, 25-27 (Oct. 1963), 1967, pp. 35-40 (Thompson Book Co., Washington, D.C., Patrick, R. L. and D. V. Black, Index Files: 1967). Their Loading and Prince, M. D., Man-Computer Graphics forComputer-Aided Organization for Use, in Libraries and Automation, Proc.Conf. Design, Proc. IEEE 54, 1698-1708 (Dec. 1966). on Libraries and Automation, Airlie Foundation, Warrenton, Probst,L.A., Communications Data Processing Systems: Va., May 26-30, 1963, Ed. I. E. Markuson,pp 29-48 (Library Design Considerations, Computers & Automation17, No. of Congress, Washington, D.C., 1964). 5, 18-21 (May 1968). Pedler, C. S., New Variables in the Data ProcessingEquation, Prywes, N. S., A Storage and RetrievalSystem for Real-Time Computers & Automation 18, No. 5, 28-30 (May1969). Penner, A. R., Problems of Basic Research Problem Solving, Rept. No. 66-05, 47p. (Moore School of in the Integrated Engineering, Univ. of Pennsylvania, Philadelphia,June 1, Bibliography Pilot Study, in Integrated BibliographyPilot 1965). Study Progress Report, Ed. L. Sawin and R.Clark, pp. 18-22 Pugh, E. W., V. T. Shahan and W. T. Siegle, Device and (Univ. of Colorado, Boulder, June 1965),. Array Design for a 120-Nanosecond Magnetic Film MainMemory, Perlis, A. J., Construction of ProgrammingSystems Using Re- IBM J. Res. & Dev. 11, 169-178 (Mar. 1967). mote Editing Facilities, Abstract in Information Processing Pyke, T. N., Jr. Computer Technology: A 1965, Proc. IFIP Congress 65, Vol. 1, New York, Forward Look, NBS N.Y., May Tech. News Bull. 51, No. 8, 161-163 (Aug.1967). Also in Yale 24-29, 1965, Ed. W. A. Kalenich,p. 229 (Spartan Books, Scientific, pp. 14-15, 28, Oct, 1967. Washington, D.C., 1965). Perlis, A. J., Procedural Languages,in Second Cong. on the Information System Sciences, Hot Springs, Va.,Nov. 1964, Raffel, J. I.A. H. Anderson, T. S. Crowther, T. 0. Herndonand Ed. J. Spiegel and D. E. Walker,pp. 189-210 (Spartan Books, C E. Woodward, A Progress Reporton Large Capacity Washington, D.C., 1965). Magnetic Film Memory Development, AFIPS Proc.Spring Perlis, A. J., The Synthesis of AlgorithmicSystems, Proc. 21st Joint Computer Conf., Vol. 32, Atlantic City, N.J.Apr. 30- National Conf., ACM, Los Angeles, Calif., Aug.30-Sept. 1, May 2, 1968, pp. 259-265 (Thompson Book Co., `Washington, 1966, pp. 1-6 (Thompson Book Co., Washington,D.C., 1966). D.C., 1968). Also in J. ACM 14, No. 1, 1-9 (Jan. 1967). Rajchman, T. A., Integrated Magnetic andSuperconductive Perlman, J. A., Digital Data Transmission:The User's View, Memories- A Survey of Techniques, Results andProspects, Proc. Eastern Joint Computer Conf., Vol.20, Washington, in Information Processing 1965, Proc. IFIP Congress65, 'D.C., Dec 12-14, 1961, pp. 209-212 (MacmillanCo., New York, Vol. 1, New York, N.Y., May 24-29, 1965, Ed. W. A.Kalenich, 19611. pp. 123-129 (Spartan Books, Washington, D.C., 1965). Ramsey, K. and J. C. Strauss, A Real Time Priority Perry, M. N., Handling Very Large Programs,in Information Schedule, Processing 1965; Proc. IFIP Congress 65, Vol. Proc. 21st National Conf., ACM, Los Angeles,Calif., Aug. 1, New York, 30-Sept. 1, 1966, pp. 161-166 (Thompson Book Co., N.Y., May 24-29, 1965, Ed. W. A. Kalenich,pp. 243-247 Wash- (Spartan Books, Washington, D.C., 1965). ington, D.C., 1966). Raphael, B., The Structure of Programming Languages, Commun. Petersen, H. E. and R. Turn, System Implicationsof Information ACM 9, No. 2, 67-71 (Feb. 1966). Privacy, AFIPS Proc. Spring Joint ComputerConf., Vol. 30, Rath, G. J. and D. J. Werner, Infosearch: Studying theRemote Atlantic City, N.J., April 18-20, 1967,pp. 291-300 (Thompson Use of Libraries by Medical Researchers,in Levels of Inter- Books, Washington, D.C., 1967). action Between Man and Information, Proc, Am. Doc.Inst. Petritz, R. L., Technological Foundations and FutureDirections Annual Meeting, Vol. 4, New York, N.Y., Oct.22-27, 1967, of Large-Scale Integrated Electronics, AFIPS Proc. Fall Joint pp. 58-62 (Thompson Book Co., Washington, D.C., 1967). Computer Conf., Vol. 29, San Francisco, Calif.,Nov. 7-10, Ray, L. C., Keypunching Instructions for Total TextInput, in 1966, pp. 65-87 (Spartan Books, Washington D.C., 1966). Machine Indexing:Progress and Problems, Proc. Third Petritz, R. L., Current Status of Large Scale Institute on Information 'Storage and Retrieval,Washington, Integration Tech- D.C., Feb. 13-17, 1961,pp. 50-57 (American Univ., Wash- nology, Proc. 22nd National Conf., ACM, Washington,D.C., ington, D.C., 1962). Aug. 29-31, 1967, pp. 65-85 (Thompson Book Co.,Washington, "R and D for Tomorrow's Computers", Data Systems,52-53 D.C., 1967). (Mar. 1969). 139

376-411 0 - 70 - 10 Reagan, F. H., Jr., Data Communications -What It's All About, Ross, D. T. and C. G. Feldman, Verbal and Graphical Language Data Proc. Mag. 8, 20-24, 26, 66-67 (Apr. 1966). for the AED System: A Progress Report, Rept. No. MAC-TR-4, Reagan, F. H., Jr., Viewing the CRT Display Terminals, Data 26 p. (Massachusetts Institute of Technology, Cambridge, Proc. Mag. 9, 32-37 (Feb. 1967). Mass., May 6, 1964). Reed, D, M. and D. J. Hillman, Document Retrieval Theory, Rothery, B., The Check Digit, Data Proc. Mag. 9, No. 6, 58-59 Relevance, and the Methodology of Evaluation, Rept. No. 4, (June 1967). Canonical Decomposition, 33 p. (Center for the Information Rothman, S., Centralized Government Information Systems and Sciences, Lehigh Univ., Bethleham, Pa., Aug. 1966). Privacy, paper prepared for the President's Crime Commission, Reich, A. and G. H. Dorion, Photochromic, High-Speed, Large TRW Systems, Sept. 22, 1966, 18 p. Capacity, Semirandom Access Memory, in Optical and Electro- Rubinoff, M., Ed., Toward a National Information System, Optical Information Processing, Ed., J. R. Tippett et al., pp. Second Annual National Colloquium on Information Retrieval, 567-580 (M.I.T. Press, Cambridge, Mass., 1965). Philadelphia, Pa., April 23 -24,, 1965, 242p. (Spartan Books, Reitnann, 0. A., On All-Optical Computer Techniques, in Optical Washington, D.C., 1965), and Electro-Optical Information Processing. Ed. J. R. Tippett Rux, P. T., Evaluation of Three Content-Addressable Memory et al., pp. 247-252 (M.I.T. Press, Cambridge, Mass., 1965). Systems Using Glass Delay Lines, Rept. No. CC-67-9, 58p. Reimann, 0. A. and W. F. Kosonocky, Progress in Optical Com- (Computer Center, Oregon State Univ., Corvallis, Oregon, puter Research, IEEE Spectrum 2, No. 3, 181-195 (Mar. 1965). July 13, 1967). The (R)evolution in Book Composition, Special Report, Part I, Computers Are Here -What Now? Book Production Mag. 79, Sackman, H., Time-Sharing versus Batch Processing: The 54 00 (Feb. 1964); Part II, Computers in '64. Year of Trans- Experimental Evidence, AFIPS Proc. Spring Joint Computer ition from Theory to Practice, 79, 44-50 (Mar. 1964); Part III, Conf., Vol. 32, Atlantic City, N.J., Apr. 30-May 2, 1968,pp. What's Ahead for Computers, 79, 55-61 (Apr. 1964); Part IV, 1-10 (Thompson Book Co., Washington, D.C., 1968). The Systems Concept -Key to Computer Profits, 79,67-73 Sackman, H., Current Methodological Research, [positionpaper (May 1964). for sessions on managing the economics of computerprogram. Rich, D. S., Multiplying the Software Dollar, Software Age 2, ming], Proc. 23rd National Cong., ACM, Las Vegas, Nev., Aug. 30-35 (Mar. 1968). 27-29,1968,pp. 349-352 (Brandon/Systems Press,Inc., Rider, B., Data Integrity in Communications Circuits, in Law Princeton, N.J., 1968). Enforcement Science and Technology, Vol. 1, Proc. First Sackman, H., W. J. Erikson and E. E. Grant, Exploratory Experi- National Symp. on Law Enforcement Science and Technology, mental Studies Comparing Online and Offline Programming Chicago, III., March 1967, Ed. S. ,A. Yefsky, pp. 133-138 Performance, Commun. AGM 11, No. 1, 3-11 (Jan. 1968). (Thompson Book Co., Washington, D.C., 1967). Salton, G., Data Manipulation and Programming Problems in Rifenburgh, R. P., A Glimpse at the Future of Data Communica- Automatic Information Retrieval, Commun. ACM 9, No. 3, tions, Computers & Automation 18, No. 5, 36, 53 (May 1969). 204,210 (Mar. 1966). Riley, W. B., Time-Sharing: One Machine Serving Many Masters, Saltzer, J. H., Traffic Control in a Multiplexed Computer System, Electronics 38, No. 24, 72-78 (1965). Rept. No. MAC-TR-30, 79 p. (Massachusetts Institute of Ring, E. M., H. L. Fox and L. C. Clapp, A Quantum Optical Technology, Cambridge, Mass., July 1966). Phenomenon: Implications for Logic, in Optical and Electro- Samuel, A. L., Time-Sharing on a Multiconsole Computer, Rept. Optical Information Processing, Ed. J. R. Tippett et al., pp. No. MAC-TR-17, 23 p. (Massachusetts Institute of Technology, 31-43 (Iv.I.T. Press, Cambridge, Mass., 1965). Cambridge, Mass., Mar. 1965). Rippy, D. E., D. E. Humphries and J. A. Cunningham, MAGIC- Sass, A. R. W. C. Stewart and L. S. Cosentino, Cryogenic A Machine for Automatic Graphics Interface to a Computer, Random-Access Memories,IEEE Spectrum 4, 91-98 (July AFIPS Proc. Fall Joint Computer Conf., Vol. 27, Pt. 1, Las 1967). Vegas, Nev., Nov. 30-Dec. 1, 1965, pp. 819-830 (Spartan Savitt, D. A., H. H. Love, Jr., and R. E. Troop, ASP: A New Books, Washington, D.C., 1965). Concept in Language and Machine Organization, AFIPS Proc. Roberts, L. G., Graphical Communication and Control Languages, Spring Joint Computer Conf., Vol. 30, Atlantic City, N.J., in Second Cong. on the Information System Sciences, Hot April 18-20, 1967, pp. 87-102 (Thompson Books, Washington, Springs, Va., Nov. 1964, Ed. J. Spiegel and D. E. Walker, pp. D.C 1967). 211-217 (Spartan Books, Washington, D.C., 1965). Sawin, L., The Integrated Bibliography Pilot Study: A Progress Roberts, L. G., A Graphical Service System with Variable Report,inIntegrated Bibliography Pilot Study Progress Syntax, Commun. ACM 9, No. 3, 173-176 (Mar. 1966). Report, Ed. L. Sawin and R. Clark, pp. 87-105 (Unlv. of Rogers, K. T. and J. Kelly, High-Information-Density Storage Colorado, Boulder, 1965). Surfaces, Ninth Quarterly Report, Rept. No. ECOM-01261-9, Sawin, L. and R. Clark, Eds., Integrated Bibliography Pilot 26 p. (Stanford Research Institute, Menlo Park, Calif., Oct. Study Progress Report, 153 p. (Univ. of Colorado, Boulder, 1967). June 1965). Roos, D., An Integrated Computer System for Engineering Sayer, J. S., Do Present Information Services Serve the Engineer? Problem Solving, AFIPS Proc. Fall Joint Computer Conf., Data Proc. Mag. 7, 24-25, 64-65 (Feb. 1965). Vol. 27, Pt. 1, Las Vegas, Nev, Nov., 30-Dec. 1, 1965, pp. Sayer, J. S., The Economics of a National Information System, in 423-433 (Spartan Books, Washington, D.C., 1965). Toward a National Information System, Second Annual Rosa, J., Command Control Display Technology Since 1962, in National Colloquium on Information Retrieval, Philadelphia, Second Cong. on the Information System Sciences, Hot Pa., April 23-24, 1965,, Ed. M. Rubinoff, pp. 135-146 (Spartan Springs, Va., Nov. 1964, Ed. J. Spiegel and D. E. Walker, pp. Books, Washington, D.C., 1965). 411-414 (Spartan Books, Washington, D.C., 1965). Scarrott, G. G., The Efficient Use of Multilevel Storage, in Rose, G. A., "Light-Pen" Facilities for Direct View Storage Information Processing 1965, Proc. IFIP Congress 65, Vol. 1, Tubes -An Economical Solution for Multiple Man-Machine New York, N.Y., May 24-29, 1965, Ed. W. A. Kalenich, pp. Communication, IEEE Trans. Electron. Computers EC-14, 137-141 (Spartan Books, Washington, D.C., 1965). 637-639 (Aug. 1965). Schatzoff, M., R. Tsao and R. Wiig, An Experimental Comparison Rosen, S., Programming Systems and Languages: A Historical of Time Sharing and Batch Processing, Commun. ACM 10, Survey, AFIPS, Proc. Spring Joint Computer Conf., Vol. 25, No. 5, 261-265 (May 1967). Washington, D.C., April 1964, pp. 1-15 (Spartan Books, Schecter, G., Ed., Information Retrieval- A Critical View, 282p. Baltimore, Md., 1964). (Thompson Book Co., Washington, D.C., 1967). Rosen, S., Hardware Design Reflecting Software Requirements, Scherr, A. L., An Analysis of Time-Shared Computer Systems, AFIPS Proc. Fall Joint Computer Conf., Vol. 33, Pt. 2, San Ph. D. Dissertation, Rept. No. MAC-TR-18, 178p. (Mass. Francisco, Calif., Dec. 9-11, 1968, pp. 1443-1449 (Thompson Inst. of Tech., Cambridge, June 1965). Book Co., Washington, D.C., 1968). Schon, D. A.A., The Clearinghouse for Federal Scientific and Rosin, R. F., An Approach to Executive System Maintenance in Technical in Toward a National Information Disk-Based Systems, The Computer J. 9, 242-247 (Nov. 1966). System, Second Annual National Colloquiumon Information 140 Retrieval, Philadelphia, Pa., April 23-24, 1,965, Ed. M. Rubinoff, Philadelphia, Pa., Dec. 1962, pp. 97-107 (Spartan Books, pp. 27-34 (Spartan Books, Washington, D.C., 1965). Washington, D.C., 1962). Schultz, L., Language and the Computer, in Automated Language Smith, F. R., and S. 0. Jones, Five Years in Focus -The Douglas Processing, Ed. H. Borko, pp. 11-31 (Wiley, New York, 1967). Aircraft Company Mechanized Information System, in Progress Schwartz, J. I., E. G. Coffman and C. Weissman. Potentials of a in Information Science and Technology, Proc. Am. Doc. Inst., Large-Scale Time-Sharing System, in Second Cong. on the Annual Meeting, Vol. 3, Santa Monica, Calif., Oct. 3-7, 1966, Information System Sciences, Hot Springs, Va., Nov. 1964, pp. 185-191 (Adrianne Press, 1966). Ed. J. Spiegel and D. E. Walker, pp. 15-32 (Spartan Books, Smith, G. P., Chameleon in the Sun- Photochromic Glass, IEEE Washington, D.C., 1965). Spectrum 3, No. 12, 39-47 (Dec. 1966). Schwartz, J.I.and C. Weissman, The SDC Time-Sharing Smith, M. G. and W. A. Notz, Large-Scale Integration from the System Revisited, Proc. 22nd National Conf., ACM, Washing- User's Point of View, AFIPS Proc. Fall Joint Computer Conf., ton, D.C., Aug, 29-31, 1967, pp. 263-271 (Thompson Book Co., Vol. 31, Anaheim, Calif., Nov. 14-16, 1967, pp. 87-94 (Thomp- Washington, D.C., 1967). son Books, Washington, D.C., 1967). Seaman, P. H., On Teleprocessing System Design, Part VI- The Smith, W. V., Computer Applications of Lasers, Proc. IEEE Role of Digital Simulation, IBM Sys. J. 5, No. 3, 175-189 (1966). 54, 1295-1300 (Oct. 1966). Seltzer, D., An Experimental Word Decode and Drive System Soref, R. A. and D. H. McMahon, Bright Hopes for Display for a Magnetic Film Memory with 20-ns READ-Cycle Time, Systems: Flat Panels and Light Deflectors, Electronics, 38, IEEE Trans. Electron. Computers EC-16, 172-179 (Apr. 1967). No. 24, 56-62 (1965). Senzig, D. N. and R. V. Smith, Computer Organization for Array Sparks, D. E., M. M. Chodrow, G. M. Walsh and L. L. Laine, A Processing, AFIPS Proc. Fall Joint Computer Conf., Vol. 27, Methodology for the Analysis of Information Systems, Rept. Pt. 1, Las Vegas, Nev., Nov. 30-Dec. 1, 1965, pp. 117-128 No. R-4003-1, Final Rept. on Contract NSF-C-370, 1v. (Spartan Books, 'Washington, D.C., 1965). (Information Dynamics Corp., Wakefield, Mass., May 1965), Serchuk, A., Laser, Hologram: Storage Team, Electronic News, Appendices, May 1965, 1 v. April 17, 1967, p. 34. Spiegel, J. and D. E. Walker, Eds., Second Congress on the Shafritz, A. B., The Use of Computers in Message Switching Information System Sciences, Hot Springs, Va., Nov. 1964, Networks, Proc. 19th National Conf., ACM, Philadelphia, 525 p. (Spartan Books, Washington, D.C., 1965). Pa., Aug. 25-27, 1964, pp. N2.3-1 to N2.3-6 (Assoc. for Steel, T. B., Jr., The Development of Very Large Programs, in Computing Machinery, New York, 1964). Information Processing 1965, Proc. IFIP Congress 65, New Shah, B. R. and K. L. Konnerth, Optical Interconnections in York, N.Y., May 24-29, 1965, Vol. 1, Ed. W. A. Kalenich, pp. Computers, 1968 WESCON Technical Papers, 16/5, Aug. 231-235 (Spartan Books, Washington, D.C., 1965). 1968, 6 p. Steinbuch, K. and U. A. W. Piske, Learning Matrices and Their Sharp, D. S. and J. E. McNulty, On a Study of Information Applications, IEEE Trans.Electron.Computers EC-12, Storage and Retrieval, 18 p. (The Moore School of Electrical 846-862 (Dec. 1963). Engr., Univ. of Pennsylvania, Philadelphia, May 15, 1964). Stephens, E. D., Application of High-Speed Holographic Photo- Sharp, J. R., Content Analysis, Specification, and Control, in microscopy, Research Rev., Office of Aerospace Research Annual Review of Information Science and Technology, Vol. 2, 6, No. 8, 25-26 (Aug. 1967). Ed. C.A. Cuadra, pp. 87-122 (Interscience Pub., New York, Stephenson, A., Planning Interactive Communication Systems, 1967). Modern Data 1, No. 9, 54-56 (Nov. 1968). Shaw, A. C., A Formal Picture Description Scheme as a Basis Stevens, D. F., System Evaluation on the Control Data 6600, for Picture Processing Systems, Inf. & Control 14, No. 1, Proc. IFIP Congress 68, Edinburgh, Scotland, Aug. 5-10, 9-52 (Jan. 1969). 1968, Booklet C, pp. C34-C38 (North-Holland Pub. Co., Shaw, R. R., Parameters for Machine Handling of Alphabetic Amsterdam, 1968). Information, Am. Doc. 13, No. 3, 267-269 (July 1962). Stevens, M. E., Nonnumeric Data Processing in Europe: A Field Shively, R., A Silicon Monolithic Memory Utilizing a New Storage Trip Report, August-October 1966, NPS Tech. Note 462, 63 p. Element, AFIPS Proc. Fall Joint Computer Conf., Vol. 27, (U.S. Govt. Print. Off., Washington, D.C., Nov. 1968). Pt. 1, Las Vegas, Nev., Nov. 30 -Dec. 1, 1965, pp. 637-647 Stevens, M. E., V. E. Giuliano and L. B. Heilprin, Eds., Statistical (Spartan Books, Washington, D.C., 1965). Association Methods for Mechanized Documentation, Symp. Short, R. A., The Attainment of Reliable Digital Systems Through Proc. Washington, D.C., March 17-19, 1964, NBS Misc. Pub. the Use of Redundancy- A Survey, Computer Group News, 269, 261 p. (U.S. Govt. Print. Off., Washington, D.C., Dec. 15, IEEE 2, 2-17 (Mar. 1968). 1965). Sibley, E. H., R. W. Taylor and D. G. Gordon, Graphical Systems Stevens, M. E. and J. L. Little, Automatic Typographic-Quality Cc mmunication: An Associative Memory Approach, AFIPS Composition Techniques: A State-of-the-Art Report, NBS Proc. Fall Joint Computer Conf., Vol. 33, Pt. 1, San Francisco, Monograph 99, 98 p. (U.S. Govt. Print. Off., Washington, D.C., Calif., Dec. 9-11, 1968, pp. 545-555 (Thompson Book Co., Apr. 7,, 1967). Washington, D.C., 1968). Stone, H.S., Associative Processing for General Purpose Com- Simkins, Q. W., Planar Magnetic Film Memories, AFIPS Proc. puters Through the Use of Modified Memories, AFIPS Proc. Fall Joint Computer Conf., Vol. 31, Anaheim, Calif., Nov. 14- Fall Joint Computer Conf., Vol. 33, Pt. 2, San Francisco, Calif., 16, 1967, pp. 593-594 (Thompson Books, Washington, D.C., Dec. 9-11, 1968, pp. 949-955 (Thompson Book Co., Washing- 1967). ton, D.C., 1968). Simmons, R. F., Storage and Retrieval of Aspects of Meaning Stotz, R. H., Directions in Time-Sharing Terminals, Computer Group News, IEEE 2, 12-18 (May 1968). in Directed Graph Structures, Commun. ACM 9, No. 3, Stratton, W. D., Investigation of an Analog Technique to De- 211-215 (March 1966). crease Pen-Tracking Time in Comuter Displays, Rept. No. Simms, R. L., Jr., Trends in Computer/Communication Systems, MAC-TR-25, 74 p. (Mass. Inst. of Tech., Cambridge, Mass., Computers & Automation 17, No. 5, 22-25 (May 1968). Simpson, W. D., Design of a Small Multiturn Magnetic Thin March 1966). Film Memory, AFIPS Proc. Fall Joint Computer Conf., Vol. Stroke, G. W., Lensless Photography, Int. Sci. & Tech. 41, 33, Pt. 2, San Francisco, Calif., Dec. 9-11, 1968, pp. 1291- 52-60 (1965). 1223 (Thompson Book Co., Washington, D.C., 1968). Strom, R., Methodology for Research in Concept-Learning, in Sklar, B. and R. Shively, IBP- A Tool for Memory System Evalu- Some Problems in Information Science, Ed., M. Kochen, pp. ation, Computer Design 8, No. 1, 54, 56-57 (Jan. 1969). 105-116 (The Scarecrow Press, Inc., New York, 1965). Slade, A. E. and H. 0. McMahon, A Cryotron Catalogue Memory Sutherland, I. E., The Future of On-Line Systems, in On-Line System, Proc. Eastern Joint Computer Conf., Vol. 10, Theme: Computing Systems, Proc. Symp. Sponsored by the Univ. of New Developments in Computers, New York, N.Y., Dec. 10-12, California, Los Angeles, and Informatics, Inc., Los Angeles, 1956, pp. 115-120 (American Inst. of Electrical Engineers, Calif., Feb. 2-4., 1965, ,Ed. E. Burgess, pp. 9-13 (American New York, 1957). Data Processing, Inc., Detroit, Mich., 1965). Slotnick, D. L., W. C. Borck and R. C. McReynolds, The Solomon Su. torland, W. R., Language Structure and Graphical Man- Computer, AFIPS Proc. Fall Joint Computer Conf., Vol. 22, Machine Communication, in Information System Science and 141 Technology, papers prepared for the Third Cong., scheduled Trueswell, R. W., A Quantitative Measure of User Circulation for Nov. 2122, 1966, Ed. D. E. Walker, pp. 29-31 (Thompson Requirements and Its Possible Effect on Stack Thinning and Bodc Co., Washington, D.C., 1967). Multiple Copy Determination, Am. Doc. 16, No. 1, 20-25 Sutherland, W. R., J. W. Forgie and M. V. Morello, Graphics in (Jan. 1965). Time-Sharing: A Summary of the TX-2 Experience, AFIPS Tukey, J. W. and M. B. Wilk, Data Analysis and Statistics: An Proc. Spring Joint Computer Conf., Vol. 34, Boston, Mass., Expository Overview, AFIPS Proc.FallJoint Computer May 14-16, 1969, pp. 629-636 (AFIPS Press, Montvale, N.J., Conf., Vol. 29, San Francisco, Calif., Nov. 7-10, 1966, pp. 1969). 695-709 (Spartan Books, Washington, D.C., 1966). Swanson, R. W., Move the Information...A Kind of Mis- Turoff, M., Immediate Access and the User Revisited, Datama- sionary Spirit, Rept. No. AFOSR 67-1247, 912 p. (Office of tion 15, No. 5, 65-67 (May 1969). Aerospace Research, USAF, Arlington, Va., June 1967). Swanson, R. W., Information System Networks -Let's Profit Uber, G. T., P. E. Williams, B. L. Hisey and R. G. Siekert, The from What We Know, in Information Retrieval - A Critical Organization and Formatting of Hierarchical Displays for the View, Ed. G. Schecter, pp. 1-152 (Thompson Book Co., On-Line Input of Data, AFIPS Proc. Fall Joint Computer Conf., Washington, D.C., 1967). Vol. 33, Pt. 1, San Francisco, Calif., Dec. 9-11, 1968, pp. System Development Corporation, Research & Technology 219-226 (Thompson Book Co., Washington, D.C., 1968). Division Report for 1967, Rept. No. TM- 530/011/00,1 v. (Santa Monica, Calif., 1968). Valassis, J. G., Modular Computer Design With Picoprogrammed Control, AFIPS Proc. Fall Joint Computer Conf., Vol. 31, Tang, D. T. and R. T. Chien, Coding for Error Control, IBM Sys. Anaheim, Calif., Nov. 14-16, 1967, pp. 611-619 (Thompson J. 8, No. 1, 48-86 (1969). Books, Washington, D.C., 1967). Tate, V. D., Ed., Proceedings of the Eleventh Annual Meeting Van Dam, A., Computer Driven Displays and Their Use in Man/ and Convention, Vol. XI, Washington, D.C., Apr. 25-27, 1962, Machine Interaction, in Advances in Computers, Vol. 7, Ed. 360 p. (The National Microfilm Assoc., Annapolis, Md., 1962). F. L. Alt and M. Rubinoff, pp. 239-290 (Academic Press, Tauber, A. S., Document Retrieval System Analysis and Design, New York, 1966). in Progress in Information Science and Technology, Proc. Am. Van Dam, A. and J. C. Michener, Hardware Developments and Doc. Inst., Annual Meeting, Vol. 3, Santa Monica, Calif., Product Announcements, in Annual Review of Information Oct. 3-7, 1966, pp. 273-281 (Adrianne Press, 1966). Science and Technology, Vol. 2, Ed. C. A. Cuadra, pp. 187-222 Tauber, A. S. and W. C. Myers, Photochromic Micro-Images: (Interscience Pub., New York, 1967). A Key to Practical Microdocument Storage and Dissemination, Van Geffen, L. M. H. J., A Review of Keyboarding Skills, in Proc. Eleventh Annual Meeting and Convention, Vol. XI, Advances in Computer Typesetting, Proc.Int. Computer Washington, D.C., Apr. 25-27, 1962, Ed. V. D. Tate, pp. Typesetting Conf., Sussex, England, July 14-18, 1966, Ed. 257-269 (The National Microfilm Assoc., Annapolis, Md., W. P. Jaspert, pp. 2-11 (The Institute of Printing, London, 1962). Also in Am. Doc. 13, No. 4, 403-409 (Oct. 1962). 1967). Teichroew, D. and J. F. Lubin, Computer Simulation-Discussion Veaner, A. B., Developments in Copying Methods and Graphic of the Technique and Comparison of Languages, Commun. Communication 1965, Lib. Res. & Tech. Serv. 10, 199-210 ACM 9, No. 10, 723-741 (Oct. 1966). (Spring 1966). Teitelman, W., PILOT: A Step Toward Man-Computer Symbiosis, Vilkomerson, D. H R., R. S. Mezrich and D. I. Bostwick, Holo- Project MAC, Rept.No. MAC-TR-32, 193 p. (Mass. Inst. of graphic Read-Only Memories Accessed by Light-Emitting Tech., Cambridge, Sept. 1966). Diodes, AFIPS Proc. Fall Joint Computer Conf., Vol. 33, Pt. 2, Tell,B. V., Auditing Procedures for information Retrieval San Francisco, Calif., Dec. 9-11, 1968, pp. 1197-1204 (Thomp- Systems, Proc. 1965 Congress F.I.D., 31st Meeting and son Book Co., Washington, D.C., 1968). Congress, Vol. II, Washington, D.C., Oct. 7-16, 1965, pp. Vlahos,P., The Three-Dimensional Display:Its Cues and 119-124 (Spartan Books, Washington, D.C., 1966). Techniques, Inf. Display 2, No. 6, 10-20 (Nov.-Dec. 1965). Terlet, R. H., The CRT Display Subsystem of the IBM 1500 Vlannes, P., Requirements for Information Retrieval Networks, InstructionalSystem, AFIPS Proc.FallJoint Computer in Colloquium on Technical Preconditions for Retrieval Center Conf., Vol. 31, Anaheim, Calif., Nov. 14-16, 1967, pp. 169-176 Operations,Proc.NationalColloquiumonInformation (Thompson Books, Washington, D.C., 1967). Retrieval, Philadelphia, Pa., April 24-25, 1964, Ed. B. F. Thomas, R. B. and M. Kassler, Advanced Recognition 'Tech- Cheydleur, pp. 3-6 (Spartan Books, Washington, D.C., 1965). niques Study, Rept. No. 1, First Quarterly Progress Report 1 Vollmer, J., Applied Lasers, IEEE Spectrum 4, 66-70 (June Jan. 1963 to 31 Mar. 1963, Contract No. DA 36-039 AMC-00112 1967). (E), 48 p., plus Appendix (RCA, Data Systems Center, Bethesda, Vorthmann, E. A. and J. T. Maupin, Solid State Keyboard, Md., 1963). AFIPS Proc. Spring Joint Computer Conf., Vol. 34, Boston, Thomas, R. B. and M. Kassler, Character Recognition in Context, Mass., May 14-16, 1969, pp. 149-159 (AFIPS Press, Montvale, Inf. & Control 10, No. 1, 43-64 (Jan. 1967). N.J., 1969). Thompson, F. B., English for the Computer, AFIPS Proc. Fall Vossler, C. M. and N. M. Branston, The Use of Context for Joint Computer Conf., Vol. 29, San Francisco, Calif., Nov. Correcting Garbled English Text, Proc. 19th National Conf., 7-10, 1966, pp. 349-356 (Spartan Books, Washington, D.C., ACM, Philadelphia, Pa., Aug. 25-27, 1964, pp. D2.4-1 to 1966). D2.4-13 (Assoc. for Computing Machinery, New York, 1964). Thornley, R. F. M., A. V. Brown and A. J. Speth, Electron Beam Recording of Digital Information, IEEE Trans. Electron. Computers EC-13, 36-40 (Feb. 1964). Wade, R. D., G. P. Cawsey and R. A. K. Verber, A Teleprocessing Tinker, J. F., Imprecision in Indexing. Part II. Am. Doc. 19, Approach Using Standard Equipment, IBM Sys. J. 8, No. 1, No. 3, 322-330 (July 1968). 28-47 (1969). Tippett, J. R., D. A. Berkowitz, L. C. Clapp, C. J. Koester and Wagner, F. V. and J. Granholm, Design of a General-Purpose A. Vanderburgh, Jr., Eds., Optical and Electro-Optical Infor- Scientific Computing Facility, in Information Processing 1965, mation Processing, 780 p. (M.I.T. Press, Cambridge, Mass., Proc. IFIP Congress 65, Vol. 1, New York, N.Y., May 24-29, 1965). 1965, Ed, W. A. Kalenich, pp. 283-289 (Spartan Books, Wash- Tou, J. T., Ed. Computer and Information Sciences-II, Proc. ington, D.C., 1965). 2nd Symp. on Computer and Information Sciences, Columbus, Waldo, W. H. and M. DeBacker, Printing Chemical Structures 0., Aug. 22-24, 1966, 368 p, (Academic Press, New York, Electronically: Encoded Compounds Searched Generically 1967). with IBM-702, in National Academy of Sciences-National research Council, Proc. Int. Conf. on Scientific Information, Travis, L. E., Analytic Information Retrieval, in Natural Language Vol.1, Washington, D.C., Nov. 16-21, 1958, pp. 711-730 and the Computer, Ed., P. L. Garvin, pp. 310-353 (McGraw- (NAS-NRC, Washington, D.C.,1959). Preprints, Area 4, Hill Book Co., New York, 1963), 1958, pp. 49-68. Trimble, G. R., Jr., Using a Computer to Simulate a Computer, Walker, D. E., Ed., Information System Science and Technology, Data Proc. Mag. 7, 18-23 (Oct. 1965). Papers prepared for the Third Congress, Scheduled for Nov. 142 =I.*. to ,- ....------.-.. ')

21-22, 1966, 406 p. (Thompson Book Co., Washington, D.C., Wilkes, M. V., Lists and Why Theyare Usdul, Proc. 19th 1967). National Conf., ACM, Philadelphia, Pa., Aug. 25-27, 1964, Waiter, C. J., A. B. Walter and M. J. Bohl, Setting Characteristics pp. F1-1 to F1-3 (Assoc. for Computing Machinery, New for Fourth Generation Computer Systems, Part III- LSI, Com- York, 19,64). puter Design 7 , No. 10, 48-55 (Oct. 1968). Wilkes, M. V., Slave Memories and Dynamic Storage Allocation, Ward, J. E., Systems Engineering Problems in Computer-Driven IEEE Trans. Electron. Computers EC-14, No. 2, 270-271 CRT Displays for Man-Machine Communication, WEE Trans. (April 196$). Systems Sci. & Cybernetics SSC-3, 47-54 (June 1967). Wilkes, M. V., The Design of Multiple-Access Computers Weaver, W., Science and Complexity, Amer. Scientist 36, Systems,, The Computer J. 10, No. 1, 1-9 (May 1967). 536-544 (Oct. 1948). Wilkes, M. V., Computers Then and Now, J. ACM 15, No. 1, Weber, J. H. and L. A. Gimpelson, UNISIM- A Simulation Pro- 1-7 (Jan. 1968). , gramfor Communications Networks, AFIPS Proc. Fall Joint Wilkes, M. V. and R, M. Needham, The Design of Multiple- Computer Conf., Vol. 261, San Francisco, Calif., Oct. 1964, Access Computer Systems: Part 2, The Computer J. 10, , pp. 233-249 (Spartan Books, Baltimore, Md., 1964). 315-320 (Feb. 1968). Wegner, P., Machine Organization for Multiprngramming, Proc. Williams, J. H., Jr., Results of Classifying Documents with Multi- 22nd National Conf., ACM, Washington, D.C., Aug. 29 -31, ple Discriminant Functions, in Statistical Association Methods

1967, pp. 135-150 (Thompson Book Co., Washington, D.C., fnr Mechanized Documentation, Symp. Proc., Washington, . 1967). D.C., March 17-19, 1964, NBS Misc. Pub.. 269, Ed. M. E. Weinstein, H., L. Onyshkevych, K. Karstad and R. Shahbender, Stevens et al., pp. 217-224 (U.S. Govt, Print. Off Washington, Sonic Film Memory, AFIPS Proc. Fall Joint Computer Conf., D.C., Dec. 15, 1965). Vol. 29, San Francisco, Calif.,, Nov. 7-10, 1966,pp. 333-347 Wilson, D. M. and D. J. Moss., CAT: A 7090-3600 Computer- (Spartan, Books, Washington, D.C., 1966). Aided Translation, Commun. ACM 8, No. 12, 777-781 Weissman, Cr, PROGRAMMING PROTECTION: What Do You (Dec, 1965). Want to Pay?, SDC Mag. 10, No. 7 & 8, 30-31 (July-Aug. Witt, B. I., M65MP: An Experiment in OS/360 Multiprocessing, 1967). Proc. 23rd National Conf., ACM, Las Vegas, Nev., Aug. 27-29, Weizenhaum, J., ELIZA-A Computer Program for the Study of 1968; pp. 691-703 (Brandon/Systems Press, Inc., Princeton, Natural Language Communication Between Man and Machine, N.J., 1968). Commun. ACM 9, No. 1, 36-45 (Jan. 1966). Wooster, II., Long Range Research in the Information Sciences, Werner, G. E., R. M. Whalen, N. F. Lockhart and R. C. Maker, Rept. No. AFOSR-1571, 24 p. (U.S. Air Force, Office of Scien- A 110-Nanosecond Ferrite Core Memory, IBM J. Res. & Dev. tific Research, presentation at the Science and Engineering 11, 153-161 (Mar. 1967). Symp., San Francisco, Calif., Oct. 3-4, 1961). Wetherfield, M. R., A Technique for Program Monitoring by Wunderlich, M. C., Sieving Procedures ona Digital Computer, Interruption, The Computer J. 9, 161-166 (Aug. 1966). J. ACM 14, No. 1, 10-19 (Jan 1967). Whiteman, I. R., New Computer Languages, Int. Sci. & Tech. Wyle, H., and G. J. Burnett, Some Relationships Between Failure 52, 62-68 (1966). Detection Probability and Computer System Reliability, AFIPS Wieder, H., R. V. Pole and P. F. Heindrich, Electron Beam Writ- Proc. Fall Joint Computer Conf., Vol. 31, Anaheim, Calif., ing of Spatial Filters, IBM J. Res. & Dev. 13, No. 2, 169-171 Nov. 14-16, 1967, pp. 745-756 (Thompson Books, Washington, (Mar. 1969). D.C., 1967). Wiesner, J. B., Communication Sciences ina University Environ- ment, IBM ,J. Res. & Dev. 2, 268-275 (Oct. 1958). Yefsky, S. A., Ed., Law Enforcement Science and Technology, Vol.1, Proc. First National Symp. on Law Enforcement Wigington, R. L., A Machine Organization fora General Purpose List Processor, IEEE Trans. Electron. Computers EC-12, Science and Technology, Chicago, Pl., March 1967, 985p. (Thompson Book Co., Washington, D.C., 1967). , 707-714 (Dec. 1963). Yourdon, E., An Approach to 7.4easuring a Time-Sharing System, Wigington, R. L., Graphics as Computer Input and Output, in Datamation 15, No. 4, 124 -126 (Apr. 1969). 1966 IEEE Int. Cony. Record, Vol. 14, Pt. 3, Computers, IEEE Int. Cony., New York, N.Y., Mar. 21-25, 1966,pp. 86-90 Zucker; M. S., LOCS: An EDP Machine Logic and Control Simu- (The Inst. of Electrical and Electronics Engineers, New York, lator, IEEE Trans, Electron. Computers EC-14, No. 3, 403- 1966). 416 (June 1965).

143 U.S. GOVERNMENT PRINTING OFFICE : 1970 01.-376-411 THE NATIONAL ECONOMIC GOAL osslr01,

Sustained maximum growth in a free 4 16 a mark et economy, withoutinflation, under conditions offull employment J and equal opportunity P

MISSION AND THE DEPARTMENT OF COMMERCE The historic mission of the Department FUNCTIONS is "to foster, promote and develop the foreign and domestic commerce" of the OF THE United States.This has evolved, as a result of legislative and administrative DEPARTMENT OF additions, to encompass broadly the re- COMMERCE sponsibility to foster, serve and promote the nation's economic development and technological advancement. The Depart- "to foster, serve and ment seeks to fulfill this mission through promote the nation's these activities: economic development and technological advancement"

Participating with Promoting progressive Assisting states, Strengthening Assuring effective Acquiring, analyzing other government business policies and communities and the international use and growth of theand disseminating agencies in the growth. individuals toward economic position nation's scientific information concern- creation of national economic progress. of the United and technical ing the nation and policy, through the Business and States. resources. the economy to help President's Cabinet Defense Services Economic achieve increased and its subdivisions. Administration Development Bureau of Environmental social and economic Administration International Science Services benefit. Cabinet Committee Office of Field Commerce Administration on Economic Policy Services Regional Planning Bureau of Commissions Office of Foreign Patent Office the Census Urban Affairs Commercial Council Office of Minority Services National Bureau ofOffice of Business Business Enterprise Standards Economics Environmental Office of Foreign Quality Council Direct Investments Office of Telecommunications United States Travel Service

Maritime NOTE: This schematic is neither an organization chart nor a program outline for budget purposes.It is a general statement Administration of the Department's mission in relation to the national goal of economic development. NBS TECHNICAL PUBLICATIONS

PERIODICALS NONPERIODICALS

JOURNAL OF RESEARCH reports National Applied Mathematics Series. Mathematical tables, Bureau of Standards research and development in manuals, and studies. physics,mathematics, chemistry, and engineering. Comprehensive scientific papers give complete details Building Science Series. Researchxesults,test of the work, including laboratory data, experimental methods, and performance criteria of building ma- procedures, and theoretical and mathematical analy- terials, coniponents, systems, and structures. ses. Illustratedwith photographs, drawings, and Handbooks. Recommended codes of engineering charts. and industrial practice (including safety codes) de- Published in three sections, available separately: veloped in cooperation withinterestedindustries, professional organizations, and regulatory bodies. Physics and Chemistry Special Publications.Proceedings of NBS confer- ences, bibliographies, annual reports,wall charts, Papers of interesc primarily to scientists working in pamphlets, etc. these fields.This section covers a broad range of physical and chemical research, with major emphasis Monographs.Major contributions to the technical on standards of physical measurement, fundamental literature on various subjects related to the Bureau's constants, and properties of matter. Issued six times scientific and technical activities. a year. Annual subscription :Domestic, $9.50; for- National StandardReferenceDataSeries. eign, $11.75*. NSRDS provides quantitive data on the physical Mathematical Sciences and chemical properties of materials, compiled from the world's literature and critically evaluated. Studies and compilations designed mainly for the Product Standards. Provide requirements for sizes, mathematician and theoretical physicist.Topics in types, quality and methods for testing various indus- mathematical statistics, theory of experiment design, trial products. These standards are developed coopera- numerical analysis, theoretical physics and chemis- tively with interested Government and industry groups try, logical design and programming of computers and provide the basis for common understanding of andcomputersystems.Shortnumericaltables. product characteristics for both buyers and sellers. Issuedquarterly.Annual subscription:Domestic, Their use is voluntary. $5.00; foreign, $6.25*. Technical Notes. This series consists of communi- Engineering and Instrumentation cations and reports (covering both other agency and NBS-sponsored work) of limited or transitory interest. Reporting results of interest chiefly to the engineer and the applied scientist. This section includes many Federal Information Processing Standards Pub- of the new developments in instrumentation resulting lications. This series is the official publication within from the Bureau's work in physical measurement, the Federal Government for information on standards data processing, and development of test methods. adopted and promulgated under the Public Law It will also cover some of the work in acoustics, 89-306, and Bureau of the Budget Circular A-86 applied mechanics, building research, and cryogenic entitled, Standardization of Data Elements and Codes engineering. Issued quarterly. Annual subscription : in Data Systems. Domestic, $5.00; foreign, $6.25*.

TECHNICAL NEWS BULLETIN CLEARINGHOUSE The best single source of information concerning the The Clearinghouse for FederalScientific and Bureau's research, developmental, cooperative and Technical Information, operated by NBS, supplies publicationactivities,thismonthly publicationis unclassified information related to Goverment-gen- designed for the industry-oriented individual whose eratedscience and technology indefense, space, daily work involves intimate contact with science and atomic energy, and other national programs.For technologyfor engineers, chemists, physicists,re- further information on Clearinghouse services, write: search managers, product-development managers, and Clearinghouse company executives. Annual subscription: Domestic, U.S. Department of Commerce $3.00; foreign, $4.00*. Springfield, Virginia22151 *Difference in price is due to extra cost of foreign mailing. Order NBS publications from: Superintendent of Documents Government Printing Office Washington, D.C. 20402

draM01.5.10,4^Ar1M91