Research and Development in the Computer and Information Sciences

Standaras ,-1 n.mu of

NBS MONOGRAPH 113, VOLUME 2

Research and Development

in the Computer and Information Sciences

Volume 2. Processing, Storage, and Output Requirements

in Information-Processing Systems—

A Selective Literature Review 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 Information 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 Mathematics—Electricity—Metrology—Mechanics—Heat—Atomic and Molec- ular Physics—Radio Physics -—Radio Engineering - —Time and Frequency -—Astro- physics -—Cryogenics.-

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, educational institutions, 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 Chemistry—Polymers—Metallurgy—Inorganic Materials—Physical 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 Standards—Weights and Measures — Invention and Innovation — Vehicle Systems Research—Product Evaluation —Building Research—Instrument Shops—Meas- urement Engineering—Electronic Technology—Technical 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 Radiation—Linac Radiation—Nuclear Radiation—Applied 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 Development—Information 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 artd 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 Data—Clearinghouse for Federal Scientific and Technical Information ' —Office of Technical Information and Publications—Library—Office of Public Information—Office 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. 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

2. Processing, Storage, and Output Requirements

in Information Processing Systems: A Selective Literature Review

Mary Elizabeth Stevens

Center for Computer Sciences and Technology-

National Bureau of Standards Washington, D.C. 20234

\J, ' National Bureau of Standards, Monograph 113, Vol. 2 4 • *

Nat. Bur. Stand. (U.S.), Monogr. 113 Vol. 2, 125 pages (May 1970) CODEN: NBSMA

Issued May 1970

For sale by the Superintendent of Documents, U.S. Government Printing Office Washington, D.C. 20402 (Order by SD Catalog No. C13.44:113/Vol. 2), Price gl.2S. MAY 2 2 1970

Library of Congress Catalog Card No. 70-603263 Foreword

The Center for Computer Sciences and Technology of the National Bureau of Standards has responsibility under the authority of Public Law 89-306 (the Brooks Bill) for automatic data processing standards development, for consultation and technical assistance to Federal agencies, and for supporting research in matters relating to the use of computers in the Federal Government. This selective literature review is the second in a series intended to improve interchange of information among those engaged in research and development in the fields of the computer and information sciences. Considered in this volume are the specific areas of information processing, storage, and output. Names and descriptions of specific proprietary devices and equipment have been included for the convenience of the reader, but completeness in this respect is recognized to be impossible. Certain important developments have remained proprietary or have not been reported in the open Uterature; thus major contributors to key developments in the field may 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 material on commercially available instruments, products, programs, or processes constitute endorsement.

Lewis M. Branscomb, Director

III

Contents

Page

Foreword Ill

1. Introduction 1

2. Processor system considerations 3 2.1. General considerations 3 2.2. Management of multiple-access systems 4 2.2.1. General considerations and examples 4 2.2.2. Scheduling, queuing, and priorities in multiple-access systems 5 2.2.3. "Swapping," dynamic memory allocation, and pagination 7 2.2.4. Client and system protection 8 2.3. System control and performance evaluation 9

3. Storage, file organization, and associative memory requirements 10 3.1. Hierarchies of storage, data compression, and data consolidation 11 3.1.1. Hierarchical memory structure 11 3.1.2. Data compression and consolidation 11

3.2. Problems of file organization and structure 12 3.3. Associative memory considerations 14

4. Mass Storage Considerations... , 16 4.1. Digital vs. facsimile storage 16 4.1.1. Facsimile storage 16 4.1.2. Document store compacting by digital storage techniques 18 4.1.3. Combined facsimile-digital storage 19

4.2. Examples of relatively direct access file storage and retrieval systems 19 4.2.1. Retrieval and reproduction of microforms 19 4.2.2. Digital data storage techniques 20

5. Output and post-processing requirements 20 5.1. Direct output and display applications 21 5.1.1. Graphic outputs and displays 22 5.1.2. Machine-aided design 23 5.1.3. Computer-assisted instruction and problem-solving 26 5.2. Multiple output modes 26 5.2.1. Audio outputs, speech synthesis and speech compression 27 5.2.2. Three-dimensional, color, and motion picture outputs 28 5.2.3. Multi-media and special purpose outputs 29

6. System output to microform and hard copy 30 6.1. Microform recording and direct output transmission 30

6.2. Printing, photocomposition and computer-controlled character and symbol generation. . . 31 6.2.1. Line printers 33 6.2.2. Photocomposition and electronic character generation techniques 33 6.2.3. Computer-controlled typographic composition 34 6.3. Hard copy reproduction 35 6.4. Post-processing operations and special forms of output 35

7. System use and evaluation 36

8. Conclusion 37 Appendix A. Background notes on research and development requirements in information processing, storage, and output 39 Appendix B. Bibliography 114 Additional references 114

V Figures

Page Figure 1. Functional diagram of a generalized information processing system 2 Figure 2. Information processing system management functions 3 Figure 3. Aging blemishes on microfilm 17 Figure 4. Microcite equipment 17 Figure 5. Output and post-processing functions 21 Figure 6. The "co-planner" outputs 24 Figure 7. Proposed microform reader 32 Figure 8. Schematic diagram of Bush rapid selector 72

VI Research and Development in the Computer and Information Sciences

2. Processing, Storage, and Output Requirements in Information Processing Systems: A Selective Literature Review

Mary Elizabeth Stevens

Areas of concern with respect to processing, storage, and output requirements of a generalized information processing system are considered. Special emphasis is placed on multiple-access systems. Problems of system management and control are discussed, including hierarchies of storage levels. Facsimile, digital, and mass random access storage media and techniques are considered. A variety of output mode requirements are also considered, including direct recording to microforms; on-line display systems: printing, photocomposition, and automatic character generation; and three- dimensional, color, and other special-purpose display systems. Problems of system use and evaluation are also briefly noted. A bibUography of approximately 480 cited references is included, together with supplemental notes and quotations from the literature.

Key words: Computer-assisted-instruction; information display; information recording; machine- aided design; memory allocation; microforms; multiple-access systems; on-line systems; output modes; photocomposition and typesetting; storage hierarchies; time sharing.

1. Introduction

This is the second in a series of reports concerned operations themselves (especially in terms of with research and development requirements and multiple-access systems), with efficient and eco- areas of continuing concern in the computer and nomical storage, with output considerations, and information sciences and technologies. In the first with post-processing operations on output. report of this series, "Information Acquisition, Many hardware, design, and theoretical consider- Sensing, and Input: A Selective Literature Review," ations necessarily arise with respect to the develop- background considerations and general purposes ment of processing specifications, matching and intended to be served by the series are discussed. processing operations, search and selection, and In addition, the general plan of attack and certain retrieval. We shall defer most of the aspects of caveats are outlined.*'-' information storage, selection and retrieval systems In general, we shall attempt to consider first the for discussion in later reports in this series. In this areas of research and development concern in the report, we will concentrate first on certain aspects computer sciences and technologies with reference of processor system planning and management to a schematic diagram of a generalized information requirements, with particular reference to multiple- processing system, as shown in Figure 1.'-^ In the access systems. It is noted, however, that priority first report of the series, we considered various scheduhng requirements, precedence interrupts, implications of Boxes 1,2, and 3 in Figure 1. In this and the like, are as apt to apply to a batch-job report, we are concerned primarily with some of system as to one that is operated in an on-line the R&D implications involved in the receipt of conversational-mode. (See, for example, Nicholson processing service requests from clients of the and PuUen, 1968).** system and in the management of the processing

Appendix A of this report contains notes and quotations pertinent to the running **See Appendix B for bibliography of references cited in this report, text. For the convenience of the reader. Notes "1.1" and "1.2" in Appendix A reca- pitulate some of the considerations discussed in the first report.

1 1. Information Acquisition

2. Information) 7. Processing p — •« Sensing anc Client Service Input Requests

3. Prep rocessing Preprocessing — -j

4. Processing 9. Processing Operations Specifications 1

5. Direct Outputs

1 2. Retrieval 13. Postprocessing

14. Outputs

15. Use and Evaluation

I .J- J

process flow Legend: feedback flow

Figure 1. Functional diagram of a generalized information processing system.

2 2. Processor System Considerations

Referring to Figure 1, we see that in some cases 2.1. General Considerations the processing operations of Box 4 resuh in direct outputs (Box 5) and/or archival storage (Box 6) Processor system planning and management as in the case of preservation of input records for activities must effectively interlace the requirements subsequent reconstruction if required. In most of multiple inputs of service requests (Box 7), cases, however, processing service requests are the handling of queuing and job scheduling priorities received from cHents of the system, from the system (Box 8), the provision of suitable means for both itself, or by impHcations in the incoming data client and system protection (Box 9), the require- received in Box 2. In parallel with the input data ments of matching the processing service requests transformations shown in Box 3, we may find to the main processing system capabilities by reductions or translations of processing service appropriate supervisory and executive control requests to forms or formats that can be processed routines (Boxes 10 and 4), and the orderly flow of by the system in Box 8, which may also involve both programs and data to and from various levels preprocessing operations that are required by of storage (Box 6). priority and scheduling considerations in multiple In Figure 2 we show those functions of Figure 1 access systems. Processing specifications developed that are most directly involved in the requirements in Box 9 may be exemphfied by controls imposed for efficient planning and management of informa- by measures taken to provide both chent and system tion processing systems as such, together with protection. These specifications, and others, will indications of areas of current and continuing then be matched (Box 10) to the performance capa- R&D concern. We may consider more particularly bihties of the processing system itself and used to here the system design requirements for computer- control the actual processing operations. based multiple access or "time-sharing" systems. "Time-sharing" is the most prevalent and popular and protection facilities. From the system design terminology for the concept of information process- and management point of view, the challenges are ing systems used by many different clients, with to take an overall planning approach, to look toward effectively simultaneous response to their diversified the integration of many processes that are now processing service requests, w^ith at least the distinct, to seek maximum gains from all multi- apparent effect that the processing facilities are purpose processing opportunities that are available, "on-line" to them when they desire to use them and and to undertake a drastic rethinking of possible with certain sharing of both data and programs trade-offs between benefits and costs.--^^ among different users.--^ Sutherland (1965) considers six varieties of on-line In fact, some of this terminology is misleading, information processing systems: (1) processing and the processing system capabilities are not so control systems, where in manufacturing applica- much shared as divided among many users, with tions the potentialities range from feedback control considerable swapping of system resources between to the optimization of profit expectancies; (2) inquiry them, and with only intermittent access between systems, such as those currently used for airline any individual user and the processor-storage reservation purposes; (3) specialized on-line systems system and its procedures. Thus, we generally for specific military applications but also for engi- prefer to use the more comprehensive term neering and design; (4) on-line programming systems; "multiple-access system" to mean "a system in (5) on-line problem-solving systems; and (6) on-line which multiple independent users are provided instrumentation designed to bring a better under- service by the computer via remote consoles which standing of the interplay of the programs and data asynchronously communicate with the computer within the computer. on a demand liasis." (Morenoff and McLean, 1967, As examples of experimental developments and p. 19). In general, the systems so-called do not applications of multiple-access systems whose provide ^ime-sharing capabilities, but they do offer facilities are shared by different users, we may cite system- sharing capabilities. first the case of mechanized documentation operations and secondly that of on-line problem-solving. 2.2. Management of Multiple-Access Systems The "reactive typewriter plan", proposed by Mooers at least as early as 1959,^-5 foresaw not only inter- Considerable current progress can certainly be library cooperation and interchange of records, shown in the development and experimental use but tie-ins to a remotely located library or informa- of computer-based multiple-access or "time- tion center for the individual user as well. Inputs shared" systems.'-''^ Nevertheless, many difficult from remotely located typewriters or consoles via problems remain, including questions of responsive teletype, telephone, and other communication links, scheduling, effective monitoring, fail-safe private with or without secondary mode transmission via file or data protection, languages of access, and voice channels, are the heart of remote-access time- efficient supervisory control.-- A complex inter- sharing systems already in experimental operation. mixture of hardware, software, and human be- Among these experimental systems we note in havioral factors is typically involved. particular those that provide responsive and selec- It is particularly to be noted that major R&D tive access to files of bibliographic data involving efforts may be required to implement the diversified scientific and technical literature. A first example is functions of man-machine reactive procedures with that of the use of the M.I.T. Compatible Time- appropriate degrees of client convenience, system Sharing System (CTSS) in Project MAC.^-^ This has efficiency, and economy.^-^ Licklider suggests been used both by Kessler and in the experimental further that "industry has not devoted as much SMART system developed by Salton and associates effort to development of devices and techniques for at Harvard.'-'' Kessler's application involves

on-line man-computer interaction as it has to machine access, in a variety of selection-retrieval development of other classes of computer hardware modes, to a corpus consisting of bibliographic and software." (Licklider, 1965, p. 66.) Those references to periodical literature in the field of responsible for R&D program planning in the physics. Recent developments include the imple- computer and information sciences should therefore mentation of a "more-like-this" request feature and be urgently concerned with the problems raised by a provision for delayed service.^-* Brown 1966 the potentials for increasing use of multiple-access describes the use of this system for updating a book systems, generally. on the basic data of plasma physics. For the future, Wooster in a 1968 report proposes the Bibliographic 2.2.1. General Considerations and Examples On-Line Organized Knowledge System, or BOOK. At the System Development Corporation, the Phenomenal growth is promised in the use of Q-32 time-sharing system is used both for experi- such multiple-access systems over the next few mental application of question-answering procedures years.^-^ Continuing difficulties are predicted, to reasonably terse, relatively well-structured data however, with respect to such areas as adequate (robbery reports of the Los Angeles Police Depart- programming languages, the appropriate design ment) and for more generalized natural language of small inexpensive but versatile consoles for text processing with respect to the Golden Book personal use, and the provision of effective privacy children's encyclopedia.^-^ 4 The BOLD (bibliographic Organization for systems that provide for "mutual training" of the '^ Z,ibrar\" Display) Project initiated at SDC by Borko. man and the machine.- Sutherland suggests that involves a number of user stations equipped ^\"ith on-hne problem-solving systems "will require CRT display console. Hght pen. and teletypewriter. techniques for pattern recognition, process control, will unite The light pen is used to designate document cate- and heuristic programming, and them gories (all members of which are displayed), to meaningfully."" (1965. p. 9.) specify data to be furnished in hard copy form, In the general case, however, these potentials and to reject data displayed during document are closely related to far-ranging prospects for search. The keyboard is used to enter desired author graphical manipulation and display, especially for or index tag selection criteria. (Borko. 1965: Carter applications to machine-aided design operations. 62-63).2-io et al.. 1965. pp. These prospects, however, will be considered in a Lockheed Missiles and Space Company's on-hne later section of this report (5.1.2). Let us next document reference retrieval system was designed consider here, then, the preprocessing of chent for the in-house report collection of some 100.000 service requests in terms of job scheduhng. queuing, items as covered in the ^L\TICO (.Uachine-.-iided and priorities. Technical /nformation Center Operations! system. Search and retrieval procedures, on-line, are based 2.2.2. Scheduling, Queuing, and Priorities in Multiple- Access Systems upon such access points as personal and corporate author names, keywords in title, subject headings In general, it is noted that "computer time-sharing as assigned, report and contract numbers, and date attempts to phase the simultaneous execution of of pubhcation.- " Search restrictors such as an- two or more programs so that the execution of any nouncement media may be employed and associa- one of the programs occurs during the natural tional displays of index terms used in the system dormancy period of the others " (Brown. 1965. p. 82) may also be provided.-- ^- and that, "typically, a particular user's program Further details are provided by Summit (1967). will be allowed to use a processor for a period of who describes the development of COX^ ERSE into time, will be stopped so that another user's program DIALOG, a system developed to investigate experi- can run. and then at some later time will be mentally the effectiveness of a user-directed continued from the point where it was stopped." language designed to provide flexible means for (Scherr. 1965. p. 1).--'"^ It has been appropriately reference retrieval.- A resource-allocation scheme observed that effective time-sharing thus involves for this Lockheed multiple-access library reference Xime-slicing.'--'^ retrieval system is discussed by Reiter (1967). This is a problem area in which both theoretical Other relatively recent examples include appli- assumptions and pragmatic considerations based cations at the Moore School of the Lniversity of upon relatively limited experience to date require Pennsylvania, such as Project CIDS (Chemical further detailed investigation and evaluation, /nformation Data System), which deals with experimental exploration of alternatives, and study formulas, names and properties of chemical com- of the design requirements for judicious balancing pounds that match a chent's query, and Project of hardware and software contributions to an effec- V ector which provides Hbrary information retrieval tive system solution.- -' services: an experimental system which has The responsive scheduling of both client access been developed by Bunker Ramo for NASA: - '^ the and job-processing requirements will necessarily augmented catalog of project Intrex-'^^ and involve effective queuing procedures, routing SAFARL an on-line text processing system at the service and priority allocations, and on-hne monitor- MITRE Corporation.-^'' It has already been claimed ing of system operations (whether or not cost that "there is a general feeling that on-hne retrieval accounting and billing requirements are also is the next major development in information involved.) Coffman and Kleinrock provide a 1968 retrieval, and represents the retrieval system of the survey of priority policies and scheduling methods future." (Drew et al.. 1966. p. 341). in use and. in addition, they discuss some of the It will be noted that many of the proposed appli- means by which various users may try to beat the cations of such systems imply on-Hne problem- system. Lampson (1968) notes further problems of solving capabilities on the part of the cHent in the management-control enforcement.- -'^ special case of retrospective literature search.-"'^ Some number of different user-control modes More generally, on-hne problem-solving possibihties should be typically available. For example, in terms are claimed to provide one of the most challenging of commerical service-center operations. Adams areas for the further improvement and increased (1965, p. 487) suggests that: "The scheduhng method

use of multiple-access systems.-"'*' used in the KEYDATA system . . . allows con- Such general systems will need to meet system versational users to reserve a level of service for design requirements especially suited to the creative minutes or hours at a time and guarantees that user. They require consideration of the convenience they will receive at least the agreed level of services to such clients in learning to use, in using, and in on a minute-by-minute basis." In other situations, modifying their use of the capabilities offered.--^' chents may wish to control whether their jobs are In particular, such systems may be designed to to be processed by remote batch operations or in some extent as 'teaching' machines-'* and as a conversational mode.- --

5 In the EDUNET planning conferences (sponsored its characteristics (e.g., program size), users seem by the Inter-University Communications Council), to try to eliminate these 'objectionable' features it was pointed out that both priority allocations from their programs and interaction usage." (including pre-emptive priorities) and reservation On-line access scheduling may be arranged on mechanisms would be needed in order to assure either an a priori or a dynamic basis.- Typical different classes of users their desired levels of system design and operation questions also arise service.- In particular. Caffrey stressed the as to scheduling and queuing with respect to multiple following: "Priority is not a simple matter; there are system components and as to the number of multiple several kinds and issues: components necessary to maintain effective service for the clientele.- -^ entry into the system (being able to plug in); # Areas of specific needs for further R&D investi- preferential treatment in the stack or queue; # gation include the following types of questions: # quality of response; (1) Whether the specific time interval (or "quan- efficiency of operation; # tum") allowed to the active client in the # user-priority vs. job-type priority; waiting line allows, for most clients, an # levels of hardware, including storage capac- acceptable delay-tolerance and, at the same ity." (Brown et al., 1967, p. 214). time, good system utilization?-^'' (2) Whether mean-response-time is a good Early evidence of R & D concern in many of these criterion of either probable client-acceptance areas was presented at the 1963 Fall Joint Com- or of system utilization? '- -^ puter Conference (Aoki et al., and Critchlow, for Whether overhead time-and-cost requirements example). Of the early JOSS (Johnniac Open-Shop (3) have been effectively assessed? '^--^ System) developments, beginning in 1963-1964, (4) Whether comparison of program sequences Bryan reports that considerable effort was devoted and data units is sufficiently systematic to to instrumentation measures to record not only the decide which priority choices will best con- use of the system as a whole but also the charac- tribute to effective utiUzation? teristics of usage by different users.'^-^''^ Kleinrock

(1966) discusses the theory of queues in relation to In general, it has been claimed that "the major experience of usage in time-shared computer sys- issues involved in time management are those of tems. Some SDC developments include those of selecting a queue discipline, determining optimum Coffman and Kirshnamoorthi (1964), Fine et al. quantum size, and dealing with system overhead." (1966), Krishnamoorthi and Wood (1965) and (Schwartz and Weissman, 1967, p. 264.) On-hne Totschek (1965). monitoring, it should be noted, will often be required Where there are many system chents for whom in order to adjust for different input and service effectively simultaneous access should be provided, requirements "mixes" and to guard against such appropriate scheduhng and queuing provisions phenomena as a number of different clients "getting must be made. To date, such provisions typically into phase with each other." '^-^^ range from first-come-first-served and round-robin Then there is the difficult question of scheduling arrangements (after a given processing interval, in multiprocessor, multiprogrammed '^^"^ networks if the first client's problem has not been completed, and, more generally, in systems incorporating the his programs, data, results to that point, and status concept of the "information processing utility". ^-^"^ and re-start information are transferred out of the Stanga (1967) discusses the performance advantages active part of the system. Then the processing- developed for the Univac 1108 multiprocessor sys- service requests and data of the next client in line tem in terms of multiple processing units and are swapped in, the system returning to the first multiple access paths to a variety of input-output client only after all waiting clients have had an equal and peripheral storage devices. Manacher (1967) opportunity for processing service) to procedures considers the special problems of multiprocessor based upon considerations of length of running control in the case of a 'hard' real-time environ- time required (e.g., short jobs first), with or without ment—that is, one where there are rigid timing priority-interrupt considerations.'-^'* bounds (both start-times and deadlines) upon In the latter case, overall system efficiency may system performance. have a built-in self-improvement facility through Dennis and Van Horn (1965, pp. 3-4) discuss the training or discipUning of the typical chent. various major characteristics of multiprogrammed, Scherr (1965, p. 105) suggests that: "it seems that multiprocessor systems in terms of the following scheduhng, if properly executed, could be used to factors: (1) computational processes are run con- 'mold' the users to some extent by assigning priority currently for more than one user, (2) many processes on the basis of job type, program size, program must share resources in a flexible way, (3) individual running time, user think time, etc., etc. Then the processes post widely varying demands for comput- user, in trying to 'beat the system', will tend to ing resources in the same time period, (4) separate conform to the image of what the writers of the processes make frequent references to common scheduhng program considered to be the ideal information, and (5) the system itself must be user." Further, "if the scheduhng procedure gives capable of flexible adjustments to changing require- low priority to a user's program because of one of ments.

6 Multiprocessor networks may be required with blocks in terms of processing efficiency. While respect to a large, time-shared, computer-utility paging typically relieves the programmer of the system. In particular, such networks may often need to know and manage the actual physical involve relatively small or relatively inexpensive storage of his programs and data,'- '^^^ the size of the processors to intercede for and to modify processing available page may be critical.--'^**'' Other problems requests addressed to the control system.^ •''^ involve questions of periodic "garbage collec- Multi-level processors may also be involved in tion" -•^^'^ and of projected time-inactive con- routing, screening, analyzing, and consohdating siderations.--^^'* query and response traffic to and from the various A pioneering approach to these problems in areas and levels of storage.--^'* terms of both pagination and of dynamic interchange The assurance of such access, effectively on an with respect to both program and data-set accession-line or '"realtime" basis, however, imphes a bility was provided in the system design of Atlas.--^^ sophisticated supervisory control and monitoring It is noted, in addition, that "pioneering work on system plan, a flexible system language, and a the concepts of segmentation and the use of predic- versatile system execution procedure.-^^^ Tech- tive information to control storage allocation was niques of dynamic swapping, memory allocation done in connection with project ACSI-MATIC." and relocation and of pagination are directed, often (Randell and Kuehner, 1968, p. 299). Such problems, with a high degree of interdependence, toward in general, continue to occupy the attention of solutions of this type of problem. system designers.'--^" "Since the effectiveness of a system increases as the service facilities are shared, a major goal of future research is a system with 2.2.3. "Swapping," Dynamic Memory Allocation, and Pagination multiple access to a vast common structure of data and program procedures; the achievement of With respect to Box 4 of Figure 2, there are shown multiple access to the computer processors is but questions of dynamic memory allocation, pagination, a necessary subgoal of this broader objective."

and "swapping". These questions obviously involve (Corbato, "System Requirements . . .", n.d., p. 3). highly interdependent relationships with the sys- In addition, hierarchies of storage as physical tem design problems that are involved in input- facts, but of virtually infinite file-memory access output for multiple user systems, in the servicing from the user's point-of-view, are implied.-^' of client-submitted processing requests, and in the Problems of dynamic memory allocation and relo- effective management of hierarchies of storage.--''-^ cation, of "placement" and "fetching",-^'^ are thus We have seen in the case of queuing and involved at a number of levels of storage.--^- How- scheduling of client interchanges with a multiple- ever, many current difficulties are to be noted. For access system that, in accordance with various example, Wagner and Granholm state that as of processing specifications, one client's program is 1965 no executive or software system has satis- typically run for a certain interval and then replaced factorily solved the problems of treating the storage by that of the next client to be served. When this hierarchy as though there were only a single store.'-^^ occurs, it is necessary to preserve the status of the In general, dynamic storage allocation techniques first client's program, to make way for the require- are intended to provide flexible means for antici- ments of the next client, and, at a later time, pating demand on the part of either the client or to restore the displaced client's program to the executive control mechanisms of the system active operation. This is the process known as itself,-^^ to optimize the utilization of primary access "swapping." '--^^ memory,'-"*^ and to provide for the calling up of It is noted that the size of the programs and data programs, subroutines, and the data to be processed sets being interchanged between active and inactive only when, and to the extent, that these are opera- status will have considerable effect upon the tionally required.-"*^ efficiency of the swapping process.--^^ Improved An obvious problem arises in terms of variable efficiency may result if both programs and data data structures with respect to the flow of programs sets are effectively modularized so that only those and data to and from the various levels of access.--^^ parts actually needed for the new operation are Such exchanges need also to be accomplished with loaded into main memory.--^^ For another example, minimum delay not only for overall system efficiency at Wayne State University, Abramowich (1967) but also in terms of chent patience with respect to considers special problems of central core storage the time-scale of response to his requirements.-^^ allocation in the case of iterative processes where Specific system design suggestions include the input data items once chosen are never again development of special processors operable in needed. At Project MAC, an "onion skin algorithm" parallel with the main system-'*^ and the use of a is used so that only those parts of memory are unified approach to the interlocking problems of dumped as are actually necessary to make room allocation, relocation, address indexing, and pro- for the incoming program. (Scherr, 1965, p. 36). tection devices.- Ramamoorthy (1966) discusses The problems of pagination relate in particular "look ahead" considerations with specific reference to the size of the blocks (or pages) of information to multiprogramming situations. Walter and that may be swapped between various levels of Wallace (1967) discuss problems of program lengths, storage and to the management of transfers of such execution time requirements, average time to

7 process, and loading times in terms of operations programming and processing penalties), but, in at academic computer centers. other cases, nine of such pages would be required, There will be, in addition, need for the most with even more severe penalties. Beyond this, there careful analyses of typical requirements for pagina- is the question of the programming effort and com- tion or segmentation both of data blocks and files puter processing time necessary to achieve, over and of programs, including the use of grouped or pagination boundaries, bit manipulation tests and clustered pages.^-^^ Again, we must consider and improvements for the total input image area. evaluate the most effective lead-times and critical Another area of current difficulty is discussed by path schedules for foresighted transfers of data Head as follows: "Often such read-in and relocation and programs between the various levels of schemes divide core memory into fixed blocks which storage.--^'^ form a repository for programs of a standard size. For such reasons, Opler concludes: "the achieve- This, of course, forces the programmer to make ment of dynamic flow through hierarchical storage a rigid segmentation of his program into one or will not be easy. Interdevice channels and control more chunks which will fit into the arbitrarily- mechanisms must be developed. Requirements for defined core blocks at read-in time. In a higher-level development of control programs will be heavy. language, each statement written by a programmer The devising of new system analysis and description will produce a fairly large yet unknown number of tools is also required." (Opler, 1965, p. 276). actual machine instructions. This is almost certain Evans and Leclerc (1967) claim that interactive to (1) aggravate the read-in problem by producing a response systems are typically mere adaptations of greater total number of instructions than would one- conventional systems and techniques and are thus for-one machine coding, and (2) prevent the pre- "far from ideal in many respects". They are there- planned division of programs into relocatable seg- fore concerned with developing improved mech- ments of standard length." (Head, 1963, p. 40). anisms for protection, address mapping and sub- Finally we note that continuing R&D concern routine linkages. should be directed to the problems of providing Problems of efficient storage allocations in terms adequate protective devices in multiple-access of hierarchies of access and of dynamic flow systems, to be considered next. between levels of storage are interrelated in many ways.-^- Nevertheless, while techniques of dynamic 2.2.4. Client and System Protection storage allocation and pagination as currently available solve certain problems of processor The R&D problems involved in the general area system management and control, they also raise of development of processing specifications (Box 9 continuing problems of more effective overall of Figure 2) will be considered in more detail in a system design.--^^ For example, studies at System later report in this series with respect to the com- Development Corporation have shown that a variety plicated requirements of truly flexible information of programs may require "considerable reorganiza- search, selection and retrieval procedures. Similarly, tion to operate efficiently in a demand-paging en- background considerations for system design in- vironment." (Fine et al., 1966, p. ll).--^'^^ It may be volving problems of privacy, confidentiality, or noted further that "an examination of the 'space- security will be covered in a separate report. time' product for a program illustrates the dangers However, we may consider here as specialized of demand paging in unsuitable environments." processing specification requirements the provision (RandeU and Kuehner, 1968, p. 303). of means to protect the system from the client, So, too, there are problems of "pagination con- the client from the system, the system from its straints". With many of the systems so far designed own malfunctioning, one client from another, and the for dynamic memory relocation applications, the client from himself. "pages" of either program-control or data constitut- Scheduling and priority-interrupt features come ing fixed-length blocks used for rapid automatic first to mind, but at a more advanced stage of system transfer between different levels of storage and planning and design we need to look at very real processing-access are arbitrarily fixed, perhaps requirements of protection from unauthorized generously. Thus, for many applications, a "page" access, as well as those of system efficiency.--^^'' of say 2,000 machine words is generally adequate for In particular, multiple-user systems require the subroutine callup or file subsection processing. provision of adequate protective devices for in- However, requirements for certain types of experi- dividually owned files and data banks and programs mental investigations and for many practical and "custom" subroutines for shared data, programs, graphical data processing operations typically and program segments, and for the executive, involve the processing of two-dimensional arrays of supervisory control, and accounting programs of the data for immediate and simultaneous access of system itself. First is the need to shield the system areas exceeding fixed pagination limitations. from inadvertent errors, clumsiness, and un- For example, in the case of a 30,000+ quantized authorized access on the part of a particular client. bit scan of a 2 x 2" photograph, the input may well Thus, regardless of the degrees of privacy required exceed the limitations of say, a 2,000-computer-word for different clients, the system itself needs pro- "page". Perhaps the input may be judiciously tection. This type of situation is especially acute juggled to fit within a 4-page reserved area (with both in situations involving user debugging of new pro-

8 grams or. even more critically, the insertion of new absent or confined to one level of object processes routines into existing programs. that run under the egis of a master control pro- Then there are the questions not only of system gram." (Dennis and Glaser, 1965, p. 9). protection from the user, but also of client pro- from the system. We note that in the detection 2.3. System Control and Performance particular,--^^*^ programs still bugging situation, in Evaluation under development by some users should not, by run, simul- error, interfere with procedures being Closely related to the problems of scheduling, taneously, for other clients of the system.--^^ dynamic memory management, and protection For the protection of the user from himself. on-Hne mechanisms are the questions of effective super- debugging programs require efficient supervisory visory control and performance evaluation of sys- and monitoring control programs (see, for example, tems under various conditions of use.--^"^^ In Evans and Darley, 1965). In other cases, a combina- particular, "the total system should be designed tion of hardware and software techniques may be to do its own bookkeeping and accounting func- used. Thus, in a project at the University of Pennsyl- tions—keeping track of usage, hours, volumes, vania, "a small computer [PDP— 5] has been used times, etc. — both for evaluation of the system and between the questioner and the central processor management of costs." (Brown et al., 1967, p. 219). as filter, that to act a so before a question gets Concurrently, however, the client "needs control through to the big computers, it is stated correctly." over the generation, revision, naming and execution (Electronics No. 18. 36 (1965).) 38, of his job-oriented processing routines, as well as the various clients assur- Next, need reasonable the storage and retrieval of such processing routines ance that the material they have stored in the as part of his data base. He needs to be able to system, both programs and data, will be retained manipulate logical and arithmetical operators, without loss, or that it can be recaptured or restored. planning factors and parameters, models and pro- However, with respect to this aspect of the projections of relevant job-oriented operations." tection of the client from the system, it is noted (Bennett et al, 1965, p. 437.) that, for example at Project MAC, while system Special problems of system design and control, malfunctions have caused some losses of client including processor and program developments, programs (and even of data), this experience has on-line debugging and instrumentation, and system not apparently discouraged continued use.--^'^ In simulation will be reviewed in another report in Kessler's apphcations of the M.I.T. system, monitor- this series (on overall system design considerations). ing programs have been developed to check the We note here, however, that "the burden of estab- continuing integrity of the data on file.--^^ Typically, lishing proper control procedures falls on the the client of a multiple-access system will also system designer who must in so doing use tech- require protection from various types of on-line niques of information theory, quality control, instrumentation techniques.'-''' information redundancy control, error detection Where several users' programs and/or data are and correction, numerical analysis, queuing theory, simultaneously held in main memory, adequate sampling theory, statistical reliability, data reduc- measures of mutual memory protection must be tion, cross correlation, display theory, decision employed.--^* It is necessary that such measures processes, etc." (Davis, 1964, p. 468). Further, should be extremely flexible because of the rapidity "suitable provisions for performance monitoring, with which swappings and relocations can occur trouble detection, and quality control should be and multi-level spheres of protection may be included as part of the system design." (Israel, required.-^" 1967, p. 203). Beyond these questions are the problems of Examples of various diagnostic and monitoring access to shared processor-systems and to multiple- systems include QUIK TRAN as used in the inter- use, consolidated program facilities, common-use pretive mode,^-^"*^ the General Electric GECOS II and centralized data banks. "The most delicate [General Comprehensive Operating System] pro- aspect of the operation of a multiple-access system grams,2 «o^' the PILOT system at M.I.T.,2 «o" and a of the MAC type is the responsibility assumed by computer "Time Monitor" available from Applied the system managers with respect to the users' Logic Corporation of Princeton, New Jersey.^-^*^*^ programs and data that are permanently stored in Hornbuckle (1967) describes a combined hardware/ the disc files. Elaborate precautions must be taken software monitoring system for graphic display to protect the contents of the disc files against applications in the Project Genie time-sharing malfunctioning of the system, as well as against system of the University of California. actions of the individual users." (Fano, 1964, p. 18). Among the types of system performance factors Notwithstanding the urgency of these requirements, of interest to system managers or users (or both) however, much remains to be done. For example, are measures of memory, paging, and subrouting "in present-day on-line systems, protection among activity loop factors and trap activity queue processes performing different tasks is either "^^ length and work profiles. ^ Coggan (1967) suggests other measures as well.^^^ Randell and Kuehner *In this case, obviously, ihe "client" may well be a programmer on the staff of the processing system facility. (1968) stress the significance of the "space-time

9 product". It is noted that "users very often want as well as his 1966 and 1967 theoretical studies of to know whether and how many channels or storage systems. areas are available, how many other users there are, For his Ph. D. dissertation at UCLA, Coffman whether they can schedule use of the system at (1966) has investigated various stochastic models predetermined times, how much of their allotted of multiple and time-shared computing operations. time has been expended, etc." (Caffrey, in Brown Elsewhere, Coffman and Wood (1966) have reported et al., 1967, p. 219). further on inter-arrival statistics for the SDC An M.I.T. example is the TTPEEK command system and Coffman and Varian (1968) consider "which allows a user to inspect both the allotments problems of page relocation algorithms. ^-^^^ Wilkes and usage of his central processor time, as well (1967) reviews a variety of page turning and related as his disc, drum, and tape records." (Corbat('), 1967, problems. Chang (1966) provides a queuing p. 95). A monitor program described by Fiala (1966) model for non-priority time-sharing such that displays for all users the size of programs, priority queue length and response time and their distribu- information, and information as to memory space tions can be readily estimated. Another theoretical activity, among other factors.-*''' However, "it is investigation of both round-robin and priority generally agreed that, constrained by cost, the aver- scheduling systems is provided by Shemer (1967). age response time is the single most objective per- Then there are studies by Martin and Estrin formance measure to the user at present." (Estrin (1967) showing that a priori estimates of computa- and Kleinrock, 1967, p. 86). tion times for given problems on given systems can A simulation program developed by Blunt (1965) be generated by modelling these computations with dealt primarily with first-come, first-served queue transitive directed graphs. Other theoretical models unloading strategy, but consideration was also are exempHfied by Kleinrock's (1967) treatment given to selecting data units with the shortest of time-shared computing facihties as stochastic servicing time, and other factors. ^ ''^ Other examples queuing system under priority service discipline, in the literature include the studies of Krish- and with the performance measure based upon the namoorthi and Wood (1965) with respect to a hmiting average time spent in the system. Nevertheless, distribution of user congestion and of Kleinrock many questions of comparative efficiency remain to (1966) involving the use of a queuing theory model be explored. '"^'^^ for the analysis of sequential processing machines.

3. Storage, File Organization, and Associative Memory Requirements

Returning to Figure 2, it is to be noted that the As of 1968, magnetic core techniques are still the development of efficient and economical storage principal media for main -computer, rapid-access, techniques is the area of continuing R&D concern information stores. Because of multiple-user with respect to Box 6. Many of the specific problems demands, as we have seen in the previous section of organization and input of material for storage, of this report, such "memories" (even although they questions of file management, digital yersus facsimile may have typical capabilities of a million bits storage, and availability of improved storage media storage and cycle times of a microsecond or less), (including microforms) will be covered in other require considerable "swapping" and "pagination" reports in this series (e.g., with respect to the requirements for time-shared, time-slicing usage. domain of information storage, selection, and However, it is noted that "memory paging, or pro- retrieval research, or, in the case of advanced devel- gram segmentation, reduces the size of the main opments in storage media, to overall system design memory but increases that of the bulk storage", considerations). Here, we shall emphasize some of (Riley, 1965, p. 74) and that therefore increasing those aspects of information storage functions and concern should also be directed to the management techniques that are most closely related to the prob- of such secondary storage. It is further to be noted lems of management and use of multiple-access that the effective management of secondary or information processing systems. More particularly, auxiliary storage has been a recurrent problem for we are concerned here with hierarchies of storage both system designers and programmers over the systems, with some of the factors in efficient file years.'*' organization, and with questions of associative Efficient and economical storage will involve the (or "content-addressable") memory requirements. interdependence of a number of different factors. In general, the processing of large volumes of Among them are the following: data to provide varied forms of efficient and eco- (1) The development and utiHzation of storage nomical storage is becoming more and more feasible media at many levels of access, density of technically, operationally, and economically. As a of information packing, and storage capacity.^ - consequence, more attention needs to be paid to (2) The effective organization of multi-level multilevel representations of the information con- storage in terms of hierarchical access tents of stored items as well as to suitable hierarchies provisions.^-^ of access, search, matching, selection, and retrieval (3) The effective organization of multi-level procedures. storage in terms of parallel processing.

10 3.1.1. Hierarchical (4 ) The integrated systems design of programming Memory Structure control, processor capabilities, and storage Hoagland'*** notes that hierarchical memory accessing for dynamic internal memory allo- structures, with various levels set cation and relocation.^"* by access-time characteristics are needed to balance costs, (5) The organization of the file or files at any given capacities, and access requirements, that the level of storage, with due consideration to the "main" memory size is tied to processor rate in most harmonious balancing of maintenance such way that the faster the latter, the larger and up-dating,^"*^ indexing and cross-referenc- and faster the memory must be to keep pace, and that ing, multiple copy deposits, statistical data on economical mass-memory systems represent the usage, usage-expectancy factors requirements, key to many new appHcations. and the extent to which typical users require We envisage, first, hierarchies of storage with responses at such-and-such thresholds (im- appropriate balancing of capacities, access time mediacy of response, comprehensiveness considerations, and usage-demand probabiHties. versus pertinency, volume of response). The total storage facilities would be compart- (7) The provision of useful guides, indexes, mentalized as appropriate by the various types of dehmitors (or restrictors) and identifiers, data to be stored, by usage considerations, and by content-indicating clues, and access-priority other criteria conductive to effective maintenance scheduling, to various sections, compart- and manipulation. They would be organized in ments, or stratifications of the file or files.'* -^ such way as to enable adaptive change to new or (7) The provision of adequate memory protection modified processing requirements. An important devices, including write-protect, read-protect, consideration here is that of the provision of and overwrite protection. * ^^'^ "streaming", or equivalent techniques.'^ ' Hierarchies of storage, especially those involving 3.1. Hierarchies of Storage, Data Compression, and Data Consolidation large bulk memory devices, also suggest the use of redundant recording techniques for improved

Just as dynamic memory allocation is increasingly random access. For example, if fully adequate demanded in processor systems involving multiple- techniques for machine translation were available, access usage, so dynamic space allocations between it is not necessarily the case that significant improve- levels of storage require a new concentration of ments in either mass storage capacity or speed of system design efforts.''-^ As of today, conflicting access would be required before production use demands of fast access and retrievability on the one could become practical. First, advantage might be hand and of compact and economical storage of taken of Zipfs law respecting the distribution of very large masses of data and recorded information, word frequency possiblities, to organize a hierarchy on the other hand, dictate compromises based upon of dictionary storage and access. Bowers (1966) multiple levels of storage.^ '' discusses as one example a dictionary stored for machine translation applications perhaps The trend toward more versatile and convenient where 35 percent of the accesses might be to less than 20 of means for man-machine interaction in time-sharing 100,000 entries. and remotely accessed systems where more than He suggests that "in most cases, some combination of different redundancy fractions mere text-message processing is desired can be for several subsets of the store will achieve the expected to continue its current momentum. New minimum access time." 44) emphases in multiple-access systems planning are (p. Again, hierarchies of storage therefore being directed toward the problems of may be used to improve the efficiency of list-processing operations providing large capacity data storage banks or files as described, for example, by Cohen of the Uni- organized in hierarchies of accessibility, especially versity of Grenoble (1967). This investigator con- where different users have differing need-to-know cludes that the best strategy for selecting privileges. the least active pages is that of time of inactivity. (See also It is likely in many potential applications that file Bobrow and Murphy (1967) for discussion of adapta- organization strategies should be geared to either tions of the LISP programming system to a two-level on-line and multiple-access or to batch operations store.) and/or to suitable admixtures of client-controlled Then there are questions of efficient eco- and job-shop-controlled priorities of access and and nomical storage with respect to the items them- processing. In such circumstances, Benner (1967) selves, including compression, truncation, use of considers certain minimal file design considerations the information involving security, recovery, optimal location/ contained in a record to determine automatically its proper storage address, addressing, flexibility for meeting changing require- and use ments, and compatibility with the programming of storage addresses as the surrogates for search ''^ and retrieval of the stored itself. system and language(s).'' item 3.1.2. Data Compression and Consolidation

System organization of central processors and of *See also the separate report in this series on some of the background considerations storage systems can be expected to become increas- affecting problems of privacy, confidentiality, or security in information processing systems and networks. ingly modular, and with maximum provisions for

366-108 O - 70 - 2 compatibility (or at least convertibility) as between ments, including those of an archival nature), may large and small system configurations available at well dictate microform storage, yet the effectiveness different locations. However, we may also consider, and efficiency of the system may be lowered because with respect to system requirements for efficient of client low-usage or dissatisfaction with microform and economical storage, not only the problems of outputs generally.'^- The economics of storage for hierarchies of access but also those of compression subsequent selection and retrieval of specified of the data actually to be filed. items relate then, first, to the costs and problems A typical instance is that of pictorial data repre- of storage as such and, secondly but not least as sentation. Beyond the photographic (or equivalent) critically, to the problems of selective recall and means for full, conservative, input data storage effective utilization. The obvious requirements here (a copy, microfilm or otherwise, for the facsimile are therefore those of effective file organization representation of the input item) lie possibilities of both for purposes of economical storage and for reductive transformations.* Here, the continuing those of efficient search, selection, and retrieval. R&D requirements with respect to information Beyond these are possibiUties for adaptive reor- processing system design involve fact-finding and ganizations of the file based upon various types of technical analyses. system feedbacks. In particular, we need to learn whether such Prywes comments that "file organization, which reductive storage can be so developed as to provide in the past has been given a secondary position in the appropriate type of reproducibility. A given overall system performance, should instead be the system may require either reduced facsimile of the kernel of future systems", and, further, that "the original, giving only pre-defined significant features, sharing of the information in the common store is or an enhanced facsimile eliminating noise or re- technically one of the most demanding, and intel- dundancy and emphasizing features such as lectually the most intricate, functions of the system. boundaries and edges, or a complete replica of the Continuous enhancement of this capability must be initial input image. In other situations, processes provided through reindexing and reclassifying the of re-computation and re-creation to produce a changing organization of the total information." reasonable reconstruction of the original input (Prywes, 1966, p. 460). may suffice. We would also look forward to increased use of 3.2. Problems of File Organization and redundancy eliminating fact-correlations, duplicate Structure checking, and validation techniques applied to items available as input for potential retrieval in Problems of file organization again involve the order to keep storage requirements within manage- need for dynamic memory allocation facilities for able bounds. In addition, as Mooers (1959) has the hierarchy of storage systems so that files may pointed out,^ '" the achievement of a minimal be set up in layers of segmentation and be rapidly redundancy store is of considerable significance re-organized according to frequencies of usage. from the client's standpoint and from that of overall Thus, "the system must have dynamic memory system effectiveness. In particular: "The . . . allocation, alternate forms of data structure, and [more compelling] reason to avoid redundancy and a data management and transfer mechanism so filler matter is that putting such matter into the that the same data can be used in all aspects of the store is not the end of it, if the retrieval system is problem solution." (Roos, 1965, p. 423). any good. The text, with all its filler and redundancies For whatever purpose machine-searchable files will keep coming out of the system, time after time, are used (record-keeping, inventory control, docu- with a consequent vast waste of human effort that mentary-item-surrogate storage, or data-, record-, will be continued indefinitely into the future." item-, and fact-retrieval), there is a commonality of (pp. 27-28) three principal kinds of operations upon the files. The basic procedures for storage in the overall These three types of file processing are those of system should obviously be simultaneously effective, file input and organization, file search and selection, efficient, and economical. Effectiveness and econ- and file maintenance and up-dating. '^""^ An example omy combined may relate, for example, to storage of multilevel file organization for a dynamic system of textual messages in facsimile-reproducible form is provided by ver Hoef (1966) with respect to the but with brief, independently searchable, content- INTIPS (/n^egrated /nformation Processing System) indicia and with appropriate selection-retrieval of the Rome Air Development Center.^ '**'' addresses as typical query-output. Yet many con- The design of effective file organizations will siderations of user-acceptance, user-convenience, involve first the consideration of the levels of availability of suitable microform readers, queuing storage required in terms of items of different size considerations, and other factors may well affect and of different type. Similarly, consideration should the economics of the situation. be given to frequencies of estimated usage. This For example, cost considerations (specifically may take the form of either activity analysis ^ " or including considerations of total storage require- of estimates based upon the age, currency or likely usefulness of items stored in the file.^-'- Possible *See also Section 3.4 of the first report in this series. Information Acquisition, Sensing, and Input. self-adaptive features should also be considered.

12 In particular, file organization must be capable and utilization. Then there are problems of whether of changing in response to changes in needs for new material may be substituted for or used to re- information and also in response to both qualitative place other material in whole or in part, updating and quantitative changes in the contents of the file. problems generally, and questions of whether incom- There exists therefore a requirement to explore ing items require transcriptions, re-recordings, en- concepts and techniques that will make feasible codings, reductive transformations of various types, self-adaptive features in both file organization and or reproduction as microforms, and the Hke. in search strategy. (J. Blum and J. Guy, private A special problem with respect to efficient and communication). economical storage of items to be searched and There is a wide variety of alternatives with retrieved is that of the encoding of pictorial and respect to different methods of file organization, graphic information for compact storage, but with whether these are planned with reference to most full-scale facsimile reproduction capabilities avail- efficient storage-retrieval operations or whether they able upon demand. In this area of pictorial data are designed with reference to presumed search- coding, a special case is that of two- or three-dimen- selective effectiveness. We may find relatively ran- sional representations of chemical structure informa- dom file organizations or arbitrary orderings of the tion. As has been noted in the first report in this records to be stored in the file; for example, in series, (on information acquisition, sensing, and libraries or document collections, by physical di- input) a number of coding, ciphering, and notation mensions of the stored items, by date of accession, schemes have been under development for linear by age class, by journal volume identification, and representations of such structural data in machine- the like. useful form. Partially arbitrary orderings of the files may be Holm (1965) notes that: "Much work is under way imposed in the form of alphabetical sequences of to store pictorial representation, such as the chemi- stored items by their source or author, or by whether cal structure, in packed coded or binary form, with or not they fall within certain prescribed chrono- the reproduction of the original pictorial form upon logical time periods, as in many typical correspond- request either as a display on film, or printed." ence files. Then, in the case of uneven file distribu- However, other pictorial data, such as photographs, tions as determined by continuing activity analysis, may probably be stored most efficiently in their there are possible random orderings based upon original form, in reduced facsimile such as micro- such assumptions as the foDowing: 'The generally form, or as TV recordings.^ accepted solution is to pick a storage location Other design requirements relate to problems of capacity (bucket), which tends to make overflows access to physical storage and to withdrawals and hkely and empty buckets rare. The storage method replacements of items to and from the store. There then goes on to dispose of the overflows by a second are maintenance problems including questions of rule, e.g., chaining, storing overflow addresses in whether or not the integrity of the files must be the bucket, or by assignment to another area." maintained (i.e., a master copy of each original item (Dumey. 196.5, pp. 258-259). accessible at all times), and whether provisions Hierarchically ordered files typically involve should be made for the periodic purging of obsolete classification schemes and structures of various items and revisions of the file organization in ac- types and, where it is not possible to determine an cordance with changing patterns, of usage or re- exclusive classification for a stored item, the use sponse requirements.^-'^^ Other design questions of cross-reference techniques including the place- relate to requirements for display of all or part of an ment of multiple copies of a particular item under item and/or indications of its characteristics prior several different classification categories, thus to physical retrieval. providing multiple parallel access to different sec- With respect to storage media and equipment tions of the file. This is the practice, for example, considerations, the information processing system not only in the hard copy files of the U.S. Patent designer must not only be concerned with the char- Office, but in microform storage and retrieval sys- acteristics of materials suitable for storing informa- tems such as the Eastman Kodak Minicard tion (to be discussed in another report in this series, developments.^^^^ on overall system design requirements), but also In the case of a partially ordered file organization, with those characteristics affecting the selected there are frequently to be found broad groupings of methodology of file organization and the total sys- stored items with perhaps random or arbitrary tem design. For example, he must carefully consider orderings within each group (e.g., a "bin" approach) the interaction between such variables as the size or orderings by frequency of usage and the like. of the file, its organization, the search strategy or The latter type of organization is designed to be strategies to be used, and equipment speed. (See particularly responsive to a specific clientele or Blunt, 1965, p. 14). usage environment.'^ '^ We shall return to these and other R&D con- Next to be considered in generahzed information siderations in other reports in this series when we processing system design and use are problems of discuss the more specialized problems of informa- input-output with respect to the files involving con- tion storage, selection, and retrieval systems as siderations of volume and processing time require- such, but we note here that many levels of file ments and questions of efficient space allocation organization and file compartmentalization may be 13 used to speed search processes, to conduct multiple tion of the item in the memory should be per- searches in parallel, and the like. In particular, fectly flexible (except for the total memory data bank management design requirements create size). file organization problems of increasing severity.^ ''*'' 2. It should be possible to add, delete, insert and In general, where there are increasing opportunities rearrange items of information in a list at any

for the establishment and multiple-access use of time and in any way . . . large-scale data banks, there are increasingly diffi- 3. The nature of the items in a list should not be cult R&D problems in file organization, file mainte- restricted. An item may be a symbol, a number, nance, and file protection. a combination of both in any length or an As Orchard-Hays has emphasized: "Probably arbitrary list. the knottiest problem facing system designers today 4. It should be possible for the same item to is how to set up, maintain, control and protect huge appear on any number of lists simultaneously." libraries of heterogeneous data, uH changing at (Hormann, 1960, p. 4).

different rates and in different ways . . . It is clear that the systems designs of the past are entirely The area of "associative" or "content-addressable" ." memories will thus require considerable further inadequate . . (Orchard-Hays, 1965, p. 240). R&D effort in both hardware and software, includ- Moreover, it is claimed that "unfortunately little research has been done on methods of organizing ing new approaches to file organization. Starting on the software side, note the emerg- and structuring large files. As a result, the available we ence of list-processing languages intended to facili- concepts are primitive. One principle is clear: our needs for information are changing; therefore, our tate symbol manipulation directly and thereby problem-solving activities more generally.^ Limita- file organization must be capable of changing." tions with respect to multiply-associated data have (Borko, 1965, p. 24). led to variations involving threaded-lists, inverted 3.3. Associative Memory Considerations lists, and multilist program structures,^-^^ and special systems such as Rover.^ A deliberate at- concept of associative or The content-addressable tempt to compromise between fixed file organization ^-^^ memories has been hailed for well over a dec- and list-processing techniques, providing for the ^-^^ as the potential, for inter- ade panacea many building of associative sublists if and only if needed, locking, multiple-aspect processing problems, was indicated in the relatively early NBS model of specifically including those of information storage, "selective recall". (Stevens, 1960). selection and retrieval systems.^^^ As of 1967-1968, For multiply-related, multiply-associated data in however, little practical realization has been a large file, the problems of efficient storage, selec- on a very small scale ^ and some achieved except tion, and access may involve considerable emphasis observers predict that this condition wiU continue upon formal modellings of the possible system for some time to come.'^ '^'* parameters and configurations. Here, considerations Large-scale associative memories have, on the of efficient machine manipulations of graph-the- other hand, been simulated on computers, notably oretic techniques, input-output economics in the in Fuller's 1963 dissertation investigations,^-^'' at most general sense, control system theory, and the Moore associates,^-^" School by Prywes and by studies of the problems of aggregation and partial Landauer and by Feldman.^ In particular, aggregation may have considerable pertinence. Prywes and Gray claim that the use of an (1963) New technologically-feasible approaches to truly addressable memory to carry out an associative massive direct access file media and to file struc- memory provides a flexibility that would scheme be turings of the associative memory type point to difficult to achieve with built-in associative hardware. significant alleviation of some of these problems In a 1967 state-of-the-art review, Minker and in the not too distant future. On the other hand, it Sable comment: "It was refreshing to see the is not yet clear that enough is now known about Government support studies leading to quantitative multiply-related and associated data to establish results in the study of vs. software imple- hardware organizational schemas that would take best ad- mentation of associative memories. that We note vantage of these promised technological advantages. these studies did not the associative show hardware Other questions as yet largely unresolved include to significantly memories be advantageous. Addi- those of the development of performance measures tional quantitative studies are to define the needed adequate to depict the appropriate trade-offs be- types of problems for which hardware associative tween storage economy and selection and retrieval memories of various sizes could useful." be (Minker effectiveness for a particular application. A familiar and Sable, 1967, p. 151). question in the literature of the information storage, following desiderata, however, are indicative The selection and retrieval field relates to the relative of continuing R&D concern with respect to memory efficiency of "linear" or "unit record" or "term- system design: "Features of the memory structure on-term" as versus "inverted" or "item-on-term" desirable for a complex processing are ^^^^ system files, and combinations of these two approaches. listed below: Some investigators who are concerned with 1. The number of different fists of items of in- problems of efficient file organization and file formation, length of lists, and length of informa- structuring from the points of view of effective

14 compartmentalization and efficient search strategies advantages lessened by the use of a relatively small have never-the-less tended to neglect the problems associative memory coupled to a large capacity and prospects of screening or sieving devices as random access store. an important contribution to search tactics.^'*^'' There may be other, as yet inadequately explored, An example of compartmentalization and screen- alternatives, however. At Sylvania, for example, ing techniques has been suggested in the case of a reference-pattern-p/ane organization for the fingerprint identification as follows: "If each finger- processing of unknown pattern inputs against many print in a set is simply classified according to wheth- stored reference pattern property-criteria, has been experimentally in er it conforms to a particular type of fingerprint realized a combination of auto- pattern, e.g.. 'whorl', the file can immediately be matically processed plastic sheets affecting the divided into 1,024 separate file sections representing behavior of a solenoid-transformer array for char- the different possible combinations of the 10 fingers. acter recognition purposes. There are possibilities More detailed analysis permits further refinement of that the same addressed, transformer-sheet arrays, the groups. With over a thousand file sections and addressed in parallel searching mode, can be the potential for easy subdivision within each sec- utilized as a practical associative memory of some- tion, the searching of even a multimillion-print file what larger capacity, but slower speed, than those available for internal ^ '^^ is not too forbidding." (Cuadra. 1966, p. 7). techniques scratchpad auxiliaries. On the hardware side, we are faced with severe In the light of present limitations of associative problems of economic and practical feasibility in memory developments, Giuliano suggests achieving large-scale, relationally associated data (1965) this alternative: files to date. Small, very fast (e.g., tens of nano- "In opinion it would pay to look further seconds performance), memories of the associative my ... into the area of large capacity, inexpensive per- or content-addressable type are beginning to appear manent memory devices which would handle asso- in operational systems, primarily as "scratchpad" ciative processing in a special-purpose manner", memories, which are defined as "small uniform Wilkes suggests still another possi- access memories with access and cycle times (p. 260) (1965) bility: "So far the slave principle has been applied matched to the clock of the logic" of the main to very small super-speed memories associated processor. (Gluck, 1965, p. 662). with the control of a computer. There would, how- These scratchpad memories are" typically used ever, appear to be possibilities in the use of a normal for such purposes as reducing time of access to sized core memory as a slave to a large core instructions, microprogramming, buffering of ." memory . . (p. 270) instructions or of data that is transferable in small Climenson comments: "A basic organization con- blocks (as in the "four-fetch" design of the Bur- cept receiving little attention recently is King's roughs B 8500 system),^-^^ storage of intermediate photostore, where a large read-only disc memory is processing results, table operations, as lookup use considered a logical extension of the stored program index registers,'^ '-^^ and, to a limited extent, content concept. The store is content-addressed, using a addressing.'' '''* Gunderson et al.,of Honeywell (1966), longest-match principle; the function found can be discuss associative memory techniques as used for data or instructions, or both. The photostore is men- control functions in a multiprocessor system — more tioned here because of the ultimate influence such specifically, to provide dynamic control over proc- devices could have on file organization. There may essor assignments, to mechanize automatic page be renewed interest with the announcement of turning schemes, and to provide other functions ITEK's version of the photostore: the Memory Cen- relating to I/O executions and to parallel tered Processor (MCP). ITEK's view is that the MCP processing.^-^" is eminently suitable for sorting, compiling, file con- Other than such scratchpad memory usage, version, typesetting, and a host of other applica- several special-purpose and experimental associations beyond the usual table lookup and file search." tive memory developments of interest. may be One (Climenson, 1966, p. 112). is Librascope's APP (/issociative Parallel Proces- Present use of large-capacity associative memory sor), as described by Fuller and Bird (1965), intended techniques, however, has been limited not only by for use in such tasks as pattern-property-extraction technological constraints but also by the even more ^-^^^ and pattern classification. Another example is difficult problems of deciding, in advance, what as- the Pattern Articulation Unit and related parallel sociations in a given body of data are most likely to processing capabilities of the ILLIAC III be valuable for future processing service requests. computer.^"'' Similarly, there are questions of how to represent In a 1966 survey, Hanlon reports that "memory efficiently the multiple cross-associations and inter- cycle times have been reported as low as 50 nano- dependencies that may be identifiable. For example, seconds" and that "although the majority of re- in a 1966 study, Dugan et al. conclude that consid- search to date has been with small memories (up erations involving the interrelationships between to 1000 cells), projections are indicated in the associative memories and general-purpose com-

10^-10* bit range, (p. 519)" Hobbs in one of a series puters in various system configurations are the of state-of-the-art reviews (1966) concurs in the most important from the standpoint of system opinion that advantages can be attained and dis- effectiveness. 15 Licklider has remarked that: "Associative, or relevant." (1965, p. 64). He adds a caveat, however, content-addressable, memories are beginning to as follows: "Only when the relative merits of var- make their appearance in the computer technology. ious associative-memory organizations are under- The first generation is, of course, too small and too stood in relation to various information-handling expensive for applications of the kind we are in- problems, we believe, should actual hardware mem- terested in here, but the basic schema seems highly ories be constructed." (p. 65).

4. Mass Storagi Considerations

Turning next to very-large-capacity storage re- With respect to problems of document miniatur- quirements, especially for permanent or archival ization for more compact facsimile-reproducible information storage (such as record files, document storage, the solutions of preference, to date, are those collections, and libraries), we note the questions of involving the use of microforms.^-^ The actual digital as against facsimile storage media and tech- microform used may be roll microform; ^-^ cassettes, niques, the problems of document miniaturization cartridges, or strips; microfilm aperture cards,^-^ or compacting of the document store, and the avail- or microfiche. "^-^ ability of so-called random access devices and sys- Beyond the conventional microform having typical tems. (The development of general purpose file reduction ratios of 35-50 : 1 or less, a few systems management procedures and programs will be con- involve 100 : 1 or greater reduction ratios, but typi- sidered in a later report in this series). cally only for very large, and very expensive, experimental systems. ""-^ Other factors are technological 4.1. Digital vs. Facsimile Storage and involve, for example, the effects of size, contrast, media, resolution and other factors on the usefulness In the general area of efficient and economical of microcopy, especially at high reduction ratios.^-* storage, there is considerable continuing controversy Another approach to document miniaturization with respect to digital as against facsimile storage, with facsimile image reproduction capabifities in- especially for documentary items to be maintained volves the use of video tape, but areas of continuing in files, with various levels of possible compromise R&D concern are generally similar to those of the also to be considered. There are, indeed, many con- photographic-media microform techniques. In addi- siderations indicating that paper records, docu- tion, it appears that video tape has the disadvantage ments, and books will be always with us (the most of allowing only a hmited number of replays (on the obvious form of facsimile storage being that of stor- order of 100 or less). ing the physical item itself), notwithstanding the Continuing system design problems in the han- possibilities for future development of new forms dling and use of conventional microforms include and media of information recording, publication, and questions of quality control of copy reproduction dissemination."*^ through several generations from a master ^ ^ and of Furthermore, "the storage of the massive amounts aging and preservation, (Fig. 3) especially of rare or *" of lexical material in libraries in digital (machine irreplaceable items.^ In general, system and cli- readable) form such as on the card catalogs, not to entele requirements continue to point to needs for mention citations, abstracts, tables of contents or higher quahty and higher resolution, with greater full text, requires memories of a size and organiza- accessibility and convenience of use, but at lower tion not being met elsewhere in the data-processing overall costs both to the system management requirements and to the individual user. industry." (King, 1965, p. 91). On the other hand, It in however, it has also been suggested that "it will not may be noted passing that an earher claim for microform storage retrieval — be long before it will be cheaper to store english text and advantages in the mass memory of a computer system than on namely, the convenience of using integral index techniques where content-indicating and other selec- paper in cabinet files." (Fano, 1967, p. 32). In general, there are both advantages and disadvantages tion criteria codes are recorded physically adjacent in the choice of either digitalized or facsimile- to the item image(s) and where matching of query reproducible storage for documentary items. and item codes automatically triggers selection and reproduction of the indicated images^"— is cur- 4.1.1. Facsimile Storage rently somewhat out of favor. A major reason for this is the developing sophistication of search strategy Let us consider first the more conventional case and multiple index manipulation techniques which of facsimile storage which may consist of the storage may nevertheless be coupled with direct and auto- of one or more copies of the original record or docu- matic retrieval from microform files, but there are ment itself (e.g., carbon copies in correspondence other reasons as well.'* '- files or bound volumes of printed pages and books An unusual example combining microform storage on library shelves). Even in the most conventional of document identifications, citations, and abstracts library case, however, critical system considera- with coordinate indexing and Peek-a-boo type search tions may arise in terms of the questions of whether, and retrieval techniques is provided in the Microcite and when, to bind or to film."*^ developments at the National Bureau of Standards.

16 Figure 3. Aging blemishes on microfilm.

Figure 4. Microcite equipment.

Peek-a-boo systems use term entry, aspect, or index dences of punched-out holes. When such a coinci- term cards to index literature for later search. That dence exists, the document corresponding to the is, coordinate locations on the card corresponding to hole position meets the parameters of the search. document acquisition or identifier numbers are The output of such systems is usually a list of the punched out as holes in just those index cards that document identification numbers, with a separate represent the index terms applicable to that partic- retrieval operation being required to select the cor- ular document. In searching, the cards correspond- responding documents from file, but the Microcite ing to the terms used in a query or search prescription equipment (shown in Fig. 4) provides means for are superimposed and visually inspected for coinci- display and output copy of abstracts of documents

17 or of other data. These abstracts or data are stored developments, two models of portable cameras as microimages on a film matrix. The location of introduced by Data Reproduction Systems in 1965 each indicated hit is used as the reference point allow for placement directly over the pages of an from which to position and project the image of each opened bound volume,*^^ and competitive capa- '-^ abstract for visual inspection by the searcher.'* bilities are claimed for a Houston Fearless micro- One major advantage of microform facsimile fiche camera-processor.*-^^ Page turners remain a storage is the relative ease with which copies of problem. Questions remain as to the relative merits information items of interest can be made available of positive vs. negative images from the point of view to the individual at his desk.^-'^ Another advantage of the user.*--^** For archival storage, however, studies beginning to be claimed for certain of the microform at NBS have led to recommendations for storing media is that of erasability and therefore of at least positive copies together with careful periodic in- limited potentials for updating, correction, or spections of the filmed records (NBS Tech. re-use.*-^^ Among the media with this potential Note 261). capability are some of the photosensitive dyes and 4.1.2. Document Store Compacting by Digital Storage thermoplastics.^- An important feature of photo- Techniques chromic techniques is the capabiUty provided for on-hne inspection and opportunity for limited cor- As opposed to facsimile storage of documentary rection of errors.''- (These and other advanced items and other extensive record collections, docu- techniques for high density storage will be discussed ment compacting may also be achieved by digitaliza- in more detail in the third report in this series). tion of the information contained in the original Questions of standardization and avoidance of item or record. The advantages of digital data duplication of effort are of concern to designers and storage are several fold. First is the advantage of users of new or improved microform storage sys- direct accessibility. Accessibility to digitally stored tems. A significant advance toward systematic data by an on-line combination of search, retrieval, efforts in greater cooperation and compatibility was and display techniques is an important new area of noted in the adoption, first by the U.S. Govern- man-machine interaction, as noted in Section 2.2.1 ment,^-^^ and second by the United States American of this report. Standards Institute, of microfiche standards.''-'^ Significantly, direct accessibility to digitally It is to be noted that technological quality control stored data can be provided both for machines and standards are also of importance (e.g., the NBS for communication links without the need for manual Microcopy Resolution Test Chart, or line density or manual-mechanical handling. This capability standards),^^" that the National Microfilm Associa- leads to a second major advantage: that of manip- tion continues to be interested in a variety of ulatability. Direct machine manipulation of standards and standard specifications,''-'^^ and that digitally stored data can provide for reformatting of "two of the three most important objectives of the information stored, for re-orderings of the store standardization are already being realized, i.e., itself, for automatic transliterations, and for (1) recipients of microfiche from the four principal interconvertibility with respect to multiple modes of report-producing agencies are able to interfile input-output and data transmission. Machine microfiche, and (2) users are able to use the same analyses of a wide variety of types and automatic equipment for viewing or otherwise processing such report generation techniques are directly avail- microfiche." (Schwelling, 1966, p. 36^'^^). able.*-29 Avoidance of duplication of effort in the prepara- A third advantage might be termed "re-computation and use of items stored in microform media bility", which could provide for encoding and involves current awareness of both existing stand- decoding of records and messages received (Schatz, ards and of availability of microcopies of various 1967, p. 3), stored, retrieved, and re-transmitted. items in various forms.^-'-'^ The National Register This capability is also available for direct application of Microform Masters, issued by the Library of of error detection and error correction techniques, Congress beginning in September 1965, provides a for machine reconstruction and display of digitized bibliographic record listing titles for which master graphic images, and for image enhancement and microcopies exist, thus serving to minimize duplica- information enhancement operations such as the tions of microcopying operations.^ -- National and cleaning-up of noise in a pattern recognition system, international guides to the availability of microform the automatic correlation of synonyms occurring in equipment provide additional sources of informa- text, or the resolution of homographic ambiguity by tion.*-^ machine examination of contextual clues. (These In terms of the conversion of hard copy to fac- topics are discussed in other reports in this series). simile storage media, one particular problem area Automatic machine control of the redundancy of is that of the range of quality of source items for digitally stored information is also in prospect, input to the microform store.*-* Another is the especially with respect to duplicate checking, question of satisfactory reproduction of multilevel validation, and data correlation and consolidation. gray scale and of color.*-^*^ Handling problems are Such operations may be directed first toward the particularly severe in the case of microcopying from reduction of duphcation as between items stored and bound books.* However, as examples of recent with respect to the information contents of various

18 items. Secondly, such consolidation or correlation Jan. 1968) will be discussed in the next report in 1 with overall information i may be designed to provide for the validation and this series, concerned verification of information contained in several processing system design requirements. messages or items. A third purpose of data consolidation is to provide, through aggregation or 4.2. Examples of Relatively Direct Access partial aggregation, protection of the identity of File Storage and Retrieval Systems individual reporting units from unauthorized terminology "random access" as applied to disclosure.*-^^^ The mass data and file storage and retrieval techniques 4.1.3. Combined Facsimile-Digital Storage is generally misleading. What is usually meant is that the access time to items or records wherever Since there are both advantages and disadvan- located in the storage medium or device is approxi-

' tages to either facsimile or digital storage, some mately the same for all items and "very much

. ' information processing system design considerations shorter than the initial access time of serial

' involve various possible combinations of storage memories". (Lickhder, 1965, p. 16). Further, as media and retrieval techniques. The classic prece- Poland emphasizes, "only with direct-access, mass- dent is the Bush Rapid Selector, where the facsimile storage equipment can randomly related references image on microfilm is directly associated with a to data files be made in an economical and expedi- digitally recorded selection-criteria encoding and tious manner." (Poland, 1965, p. 249). where machine matches of selection-criteria codes in a query with those recorded for stored items 4.2.1. Retrieval and Reproduction of Microforms automatically trigger the reproduction of the facsimile images of the selected items."*-^" A wide range of equipment is available for the In relatively small-scale and inexpensive systems, manual, semiautomatic, or automatic (e.g., com-

-^' '^-) a wide variety of digital codes (in edge-notching,'' puter-manipulated retrieval, re-enlargement, color coding,^-'*'^ and so forth) are used for "homing" display, and copy reproduction of conventional on subsets of items recorded for storage and re- microforms such as roll microfilm, microfilm strips trieval in facsimile form or for direct selection of and microfilm aperture cards. In addition, systems "^-^^ the individual item itself. Similar techniques are such as LODESTAR and other current develop- employed in some of the more reoent large-scale ments involve the use of cartridges, cassettes, and microform retrieval systems, such as Magnavue magazines to improve access. Relatively recent j (to be discussed below). examples include the VSMF (Visual Search Micro- film Files) System of Information Handling Services, Then it is to be noted that . . the photographic Inc.,4-34 Filmac 400,*35 i emulsion can also be used to store digital infor- 3M's and Recordak's ^-^^ [ all Miracode Microstrip Systems. Other relatively j mation. If this is done, a reel of film acquires of the functional virtues of digital tape while avoid- small-scale systems include Remstar,^-^^ the I f ' ing its information storage limitations. Optical Randomatic Data System's equipment for key- f means, such as flying spot scanner techniques, can boarded indexing and retrieval of edge-notched J;

E paper, plastic or film cards, the Randtriever j be used to read the digital information at rates

P System, and Mosler Safe Company's 1 which compare favorably with the electronic reading of digital tape." (Condon, 1963, p. 137). An advanced Selectriever.''^^ technique involves laser recording of digitally Examples of commercially available retrieval I devices for microfiche in particular include the I [ encoded identification, content-indication, and other Fearless Film (Compact Automatic i ' selection information on high resolution micro- Houston CARD provides ' desktop reader, which ;. forms.''-^"' Retrieval-Display) Another approach to multi-bit storage is that of 4-second (or less) random access to approximately in of a of inter- ^ Lamberts and Higgins of Eastman Kodak who use 70,000 pages any one number changeable magazines; the Itek 18.24 Reader- , the recording of diffraction grating patterns on high- Printer which handles aperture cards, roll film, ! resolution film involving composites of the grating j 5" 8" i'. patterns of seven spatial-frequency components of microfilm jackets, and microfiches up to x a character. These investigators point out that this with selective masking, variable size print and high- contrast opaque or translucent copies (Systems 6, 1. approach is closely' similar to that of holography: I ' "In fact, it turns out that a composite grating is No. 6, 39 (1965).), DuKane's desktop viewer for micro- essentially a Fraunhofer-type hologram of a series fiche in various sizes also up to 5" x 8" at 15 X i readers; a r of bright points." (Lamberts and Higgins, 1966, magnification; Recordak FILMCARD p. 730).''-3'^- desktop reader from Data Reproduction Systems; In general, however, advanced techniques for the Microcard EL-4 Automatic Enlarger-Printer, truly high density storage, promising such possi- and others. Ibilities as 13 million bits in a photochromic film Advanced or large-scale systems are also begin- memory plane two inches square (Reich and Dorion, ning to come into operation, at least on an experi- 196.5), or Honeywell developments indicating mental basis. Among these are the DARE (Docu- potential storage of "two million bits ... on a mentation Automated Retrieval Equipment), the surface the size of a dime" (Commun. ACM 11, 66, Ampex Videofile system, the high-resolution Micro-

19 Vue, and SDC's Satire system. Thus, at the U.S. materials and techniques. For high density digital Army Missile Command, the DARE system utilizes data storage, magnetic technique developments Magnavox Magnavue equipment, under computer include larger and higher-speed bulk core storage, control, for a large file of engineering drawings especially with new types of tiny cores;* *^ discs and recorded on film chips.^-^' Also under U.S. Army disc pack units with capacities ranging to billions of sponsorship, at Huntsville, developments include bits of data and access times of 100 or less milli- combined usage of Alden/Miracode techniques seconds;*-^" drums,*-^^ and magnetic cards and with the semi-automatic generation of Miracode data cells.*-^^ film. Pyke summarizes the current situation with The Videofilm techniques involve the storage, respect to magnetic media as follows: "A large retrieval, and reproduction of microforms in the effort is now being expended on the development of special sense of recordings on video tape. The initial mass storage devices. Larger and faster bulk cores videofile approaches to these techniques at Radio are being designed. Drums operating with parallel Corporation of America have apparently been at transfer and with logic for queuing access requests, least temporarily abandoned, because of probable thus optimizing drum operation, are forthcoming. high costs, in favor of RACE, techniques involving Disk units with an individual head per track are the use of relatively large-sized magnetic cards with becoming available and devices such as data cells apertures for optical images.^^^ However, Ampex promise many billions of bits of storage at a reason- has embarked on a program of development and able cost." (Pyke, 1967, p. 162). marketing of a somewhat more modest Videofile From an applications standpoint, Bonn emphasizes system, including remote query capabilities.'*^^ that "the currently used devices are drums, disk The Micro-Vue system developed by the Republic files, magnetic card devices, and tape loop units. Aviation Division of Fairchild Hiller Corp., involves These machines all store large amounts of data, reduction ratios up to 260:1, so that as many as from 6.5 million to 4.8 billion bits, on-line in one 10,000 page images can be recorded on a single device at any one time. Any of the stored informa- 4" X 5" microfiche. These "ultrafiches" can be tion can be retrieved without reading the interven- loaded into 20-chip holders with random access to ing data between two desired records. Any record any one of the 200,000 stored images in, at most, can be written or modified without operation on any 30 seconds. (Systems 8, No. 1, 6 (Jan. 1967).)»« other record. Records in mass storage devices can Photochromic materials (as discussed elsewhere be updated in place." (Bonn, 1966, p. 1861). in this series of reports) also promise high reduc- In the area of advanced developments for digital tion ratios. For example, developments at NCR data storage involving electron or laser beam record- include the recording of up to 3,000 microimages ing onto photographic media, semiconductor films, on a single microform convenient for multiple distri- or thin metallic foils, examples include the Dove ''^^ bution, i.e., as a microform publication medium.^ Data Storage and Retrieval System,*^^ IBM's Other experimental developments in microform "trillion-bit" memory recently deUvered to the storage include work at Aeroflex Laboratories,*-*^ Atomic Energy Commission,*-^* and the UNICON and further improvements in both photoplastic technique developed by the Precision Instrument recording and the Photocharge recording process Company.*^^ In addition, Kump and Chang (1966) at CE.**"^ Then there are the successive Walnut, provide a discussion of a thermostrictive recording Cypress, and 1350 systems developed by IBM, technique for Permalloy films, achieved by the although there is some doubt that any of these will application of local stresses induced by laser or continue to be available for graphic storage.**^ electron beam, which is claimed to promise future The IBM 1360, a film chip system discontinued after mass memories with storage efficiencies of a million the delivery of two models to the Atomic Energy bits per square inch. Commission, was described by Kuehler and Kerby, Then we note, on the one hand, that portable 1966.* However, the high density photographic digital data storage modules are under development storage technique has considerable promise for that will enable interchangability on standard input/ very large capacity digital data storage, to be output equipment, much as magnetic tapes reels considered next. are now handled,*-^^ and that, on the other hand, "breakthroughs we seek in mass memory may come 4.2.2. Digital Data Storage Techniques from the molecular engineers, who can develop The media and methods available for digital data improved materials with which we can fully exploit storage include high-resolution photographic the inertialess properties of light and electron beams materials, magnetic media, advanced developments to write and read data bits closely approaching using laser recording techniques, and other special molecular size." (Hoagland, 1965, p. 57).

5. Output and Post-Processing Requirements

Referring back to Figure 1, we see that system In Figure 5, we show some of the areas of continuing outputs may occur directly (Box 5) or, in Box 14, R&D concern with respect to the output and post- as the result of file searches or as subjected to a processing functions. It has been claimed that, variety of post-processing operations (Box 13). in general, too little attention has been given to 20 Reactive outputs and display- 5. Direct Machine- aided Outputs design techniques

13. Post Processing Operations Formatting Transliteration Translation Consolidation

Advanced Printing developm ents 14. Output Photocomposition and Character generation Hard copy and facsimile reproduction Multiple output modes

Figure 5. Output and post-processing functions.

output requirements in information processing For each of these purposes, if the computer out- '^ system design.^ Further, it is noted that "the puts are direct, they should in the first place be completeness, accuracy, and accessibility of system responsive to the processing service requests, they outputs are each individually dependent upon should be presentable in terms of legibility and system design and can be improved through convenience, and they should be produced with improvement in system design." (Davis, 1967, p. 8). capabilities for on-line modification within a reason- First, advanced developments of direct output able time scale. Bauer presents a concise statement modes with interactive man-machine response may of a minimum of four capabilities that a responsive be considered, especially with respect to graphic display system should have in order to provide displays, to applications such as computer-assisted assistance to the human partner in many man- instruction and machine-aided design, and to machine interactive situations. They are: prospects of color, three-dimensional and motion picture outputs. "1. Correct an input 2. Accept a query or the specification of an 5.1. Direct Output and Display Applications operation, then perform the required or desired operation which should provide the Direct output and display requirements for the proper information to the user necessary to foreseeable future encompass a wide variety of accomplish the next step. user-oriented capabilities, some of which are com- 3. Accept and appropriately handle information mercially available on at least a limited basis today to enable the computer to interpret correctly and some of which will require further research, future information which it may receive from development, and production engineering. Appli- the user, or cation areas of current R&D concern include those 4. Direct, on a step by step basis, a procedure of computer-aided instruction, machine-aided for information input and output." (Bauer, design, and machine-aided problem-solving systems. 1965, pp. 17-18).

21 In addition, console capabilities for "soft" (or graphical communication system itself,^-^ and an transient) displays should provide flexible and con- extension to three-dimensional drawing applica- venient editing and control facilities including tions. Sketchpad III, was outlined by Johnson. The insertions, deletions, corrections, rearrangements, latter investigator remarked: "General three- scale-changing, size and rotational transformations, dimensional graphical communication, which deals automatic multiple copying, and the like. Still with arbitrary surfaces and space curve intersec- other uses of on-line reactive displays are for pur- tions, presents many difficult problems; the begin- poses of text editing, sorting, and printing as for ning has been modest and much work remains example in programs under development at Bell before the complete graphical communication Laboratories (Mathews and Miller, 1965) and at problem is solved." (1963, p. 347). the University of Pittsburgh. (Roudabush et al., Organizations neighboring M.I.T. have also been 1965).'^^'' When such techniques are applied to a engaged in the development and utilization of complete publication cycle, by feeding the corrected graphic input-output equipment for man-machine tape products either off-line to automatic composing reactive display in problem-solving applications, machines or on-line to a computer typesetting pro- notably. Bolt, Beranek and Newman and the the loop is closed from original data to finished gram, MITRE Corporation.''** More recently, it is to be output product. noted that: "Computer-made maps presenting physical, economic and social data in easily understood 5.1.1. Graphic Output and Display graphic form will be produced, along with similar diagrams, charts and graphs, in a new Laboratory At least as early as 1953, a CRT display system for Computer Graphics at Harvard's Graduate was available on the lUiac.'' - Similarly, "Whirlwind School of Design." (Commun. ACM 9, 310 (1966).) (at MIT) had a cathode ray tube and light pen in Chasen and Seitz (1967) consider not only the early the early 1950's. A prototype of the APT system sketchpad developments, but also those of (computer controlled machine tool) was programmed Lockheed-Georgia, IBM, Control Data Corporation, on Whirlwind in 1955." (Wigington, 1966, p. 86). and General Motors.'^ Reactive display for purposes of air traffic control applications was investigated in the early 1950's Other experimental on-line data analysis and at the National Bureau of Standards. A combination display systems include APEX at M.I.T.'s Lincoln of the DYSEAC computer, special display equip- Laboratory (Forgie, 1965), and the similar capa- ment, control devices including a joystick serving a bilities of the SDC Variable Display system light pen funtion, and a radar link to the Washington (Schwartz et al., 1965) and its Graphic Tablet National Airport provided the basis for on-line, Display.''^'' A West German example involves a real-time experimentation. With actual radar data Telefunken combination of a display console and being displayed on one scope, the operator was able its TR-4 computer for air traffic control applica- to indicate targets of special interest which should be tions. (Stevens, 1968, p. 9) In this system, a manually marked or brightened for him, and a second scope controlled moving ball permits the translation of displayed the results of computer predictions of displayed images and the multidirectional scanning expected positions of target aircraft within specified of different areas shown on the display scope. At time intervals. The same equipment could also be the National Bureau of Standards, developments in used for problem-solving simulations, such as war graphic input-output, such as ACCESS and MAGIC, gaming.'^ are designed for differing types of requirements in By 1962, an advertisement asked, with appropriate U.S. Government organizations.^-^ illustrations, "Would you believe a Calcomp plotter Then there is the time-sharing facility for experi- and any computer can draw pictures like these in ments in man-machine interaction at the Computer seconds? ... A perspective sketch of your new Center of the University of California, Berkeley, which is described Lichtenberger Pirtle plant. . . . Statistical trend charts . . . molecular by and In effect, the user available to a structure diagrams . . . apparel patterns, graded (1965). has him Scientific Data Systems (SDS) 930 computer with for sizes . . . and even the Mona Lisa". (Commun. ACM 10, A9 (1962).) 16,000 words of fast memory, modified to exclude The development of full-scale graphical input- his direct access to input-output instructions (which, output communication and processing systems, with instead, are carried out for him by the system execu- particular emphasis on man-machine problem- tive) and by additional software-interpreted in- solving is of course represented first by M.I.T.'s structions. Remote access facilities include tele- Sketchpad. This was publicly introduced at the 1963 types, CRT display keyboards and a small DPP-5 Spring Joint Computer Conference where Coons processor equipped both with a CRT display unit outlined the requirements for a computer-aided and a RAND Tablet. design system; Ross and Rodriguez discussed In addition to the Digital Equipment Corporation's the theoretical foundations of such systems, with graphic input/output equipment used in many of emphasis upon appropriate languages and data the pioneering and continuing responsive-output structures; Stotz described the man-machine applications,^ '" a relatively wide variety of graphic console facilities, specifically including the display output and display devices is also offered by various requirements; Sutherland outlined the Sketchpad other suppliers. Examples include, but are obviously

22 not limited to Bell Laboratories' "Glance";^" computer graphic output device despite its hardware Benson-Lehner developments;^^- Bunker-Ramo limitations? This developmental research problem equipment: ^-'^ Datatronics Engineers, Inc.;^^^ involves discriminatory analysis, differential Gerber Scientific Instrument Co.;^^^ Information equations, and simulation." (p. 107). Displays, Inc.;^'^ Informatics' DOCUS (Display Then we note that: "Davis has stated that Oriented Compiler L sage System) for the Rome Air because of the universality of pictures the 'improve- Development Center; ITT Federal Labora- ment of display techniques has gone hand-in-hand tories: ^ Minnesota Mining and Manufacturing with man's progress in every field of human en-

5.19 '^ (^Q . Philco; Sanders Associates: -' Strom- deavor.' Unfortunately, the road to good displays has berg-Carlson (especially the SC-1100 Display not been smooth. In her paper, which surveys the Inquiry Station and the SC-1200 Digital to Video history of displays, she points out how disorganized, Display for computer applications involving multiple spasmodic and even serendipitous has been the access to data at remote locations), and the Tasker improvement in displays." (Davis, 1966, p. 236). Instruments Corp.^-^- In general, as of 1967, the Mills reports that "a very thorough treatment of majority of computer manufacturers also offer computer-driven displays and their use in man- graphic input-output capabiUties via peripheral machine interaction is given by Van Dam. This equipment that can be connected on-line to the paper — really a mongraph — includes a brief history, central processing facihties. Obvious examples a thorough treatment of display technology at a very include UNIVAC, IBM, Honeywell, Control Data, satisfying level of technical thoroughness and depth, -^ and others.^ and a survey of some of the applications of man- Some of the new applications of user-controUed machine interactive systems involving display graphic input and output devices include the on-line terminal devices. Sixty bibliographic citations are alteration of a PERT (Program £'valuation and included. This is one of the few papers to recognize /?eview Technique) network display to determine that imaging techniques other than those based on -^ the effects of setting different target dates; ^ cathode ray tubes may have an apphcation in graphic Control Data Corporation's computer-directed terminal devices; some half-dozen non-CRT construction drawing system involving long distance techniques are mentioned. The communication telephone lines for transfer of problem and solution problems raised by moving the display terminal data (Business Automation 12, Nq. 7, 55, July, to a remote location are not considered. Sutherland 1965). and the Comex (Command Executor) system treats the state of the art in computer graphics and which allows "several simulated commanders to indicates some further requirements. The paper is use a subset of English to control the flight of simu- not reaUy limited to the issues of computer graphics lated objects on the system's display console. Head- but considers them within the context of man- ing, course, range, altitude, velocity, and destination machine coupling." (Mills, 1967, p. 231). can each be controlled by English statements typed into Comex on-line." (Schwartz et al., 1965, p. 29). 5.1.2. Machine-Aided Design Walter (1967) claims that "immediate application of graphic data processing to solve a variety of One of the most challenging areas of information problems is both physically possible and eco- processing system output is unquestionably that nomically desirable. Graphic data processing can be of reactive display and response between man, effectively applied in mathematics, engineering, machine, and data banks in machine-aided design banking, education, communications, medicine, applications.^"--^ Graphic input-output capabilities, management information systems, programming, together with man-machine-interactive on-line and many other fields." (Walter, 1967, p. 107). modifications, are essential to the machine-aided Similarly, Prince reports that "a number of indus- design procedures. A pioneering type of application trial, government, and university laboratories are to problems of hospital and architectural design, exploring various applications of graphical com- discussed by Licklider and Clark (1962) several puter-aided design. Applications receiving primary years before the General Motor's DAC (Design attention include the preparation of digital tapes for Augmented by Computers) system's claim for a numerically controlled cutting tools, trajectory "first", (as of Fall, 1964, see Hargraves et al.) has studies, structural analysis, aircraft and automotive had continuing development and expansion by Bolt, shape design, shipbuilding, flight test data reduc- Beranek and Newman personnel. (Fig. 6) The DAC tion, circuit analysis, and printed circuit board system, on the other hand, includes rapid-response layout. Most of these must be considered as experi- microfilming and microfilm re-display, also ev- mental programs; only a very few are in a produc- idenced in such developments as an IBM graphic ^-^^ tion status." (Prince, 1966, p. 1701). data processing system and the D-200 equip- Yet, there are many areas of continuing R&D ment offered by Strand Division of Datatronics concern with respect to character sets and remote Engineers, Inc.^-^^ terminal design (to be considered in the next report There are a number of isolated, but intriguing, in this series), hardware limitations, and human examples of machine-aided design applications engineering considerations.^ -^^ For example, Walter ranging from the design of ships ^-^^ to the CADET of Honeywell, Inc. (1967) asks: "Can mathematical (Computer-^ ided /design fi'xperiment Translator) techniques be used to improve the accuracy of a system described by Lang et al. (1965), which has 23 WHAT TYPE GRAPH ? WHAT MODE ?

1 . TIME SERIES 1 . FREQUENCIES 2. HISTOGRAM 2. PERCENT OF TOTAL 3. EAR GRAPH 3. CUMULATIUE FREQ. ^. LINE GRAPH 4. CUMULATIUE PERCENT

T Y P E I N N U N E R A y TYPE IN NUMERAL

Figure 6. The "co-planner" outputs.

24 been designed for such processing tasks as three- (CIRCuit AnaLysis) programs available on the dimensional shape description and the development Project MAC faciUties (Dertouzos, 1967). Ninke of non-Hnear electronic circuit designs.^-^ Other (1965, p. 846) describes an IBM 7094 application in areas of applicability include aerospace research,^ '''' which "a system block diagram composed at the design of aircraft and automobiles,^-^' other archi- console can be used as input to a special compiler, tectural developments ^-^'^ and civil engineering the compiled program can be run and the results and highway design, including the design of a viewed at the console." However, "we are only super-highway interchange in a brief ten minutes by just beginning to explore systems where the com- an engineer using a remote communication Unk to puter asks questions of the programmer to resolve MIT's MULTICS (Multiplexed /nformation and ambiguities in what it is told." (Sutherland, 1965, Computing Service) system.'^-^'^ Machine-aided p. 11). design has also been extended to the fields of the Hardware, software, and systems planning con- arts and crafts, notably in the area of textile siderations involving human engineering considera- design.^-^- tions and human behavioral factors, come especially In addition, at the University of California at to the fore in experimental applications, such as the Santa Barbara there is the Culler-Fried system for following: "computer displaying and of transforming mathematical functions in their pathologic intervals" (1) "A few users have experienced the delight of (Cheydleur, 1965, p. 175) and Engelman states that sketching and printing to the computer and

"W. A. Martin . . . [is] working on input from scopes having it 'clean up the input' and then generate signifying a light an interesting ." achieved by with pen labeled graphs and drawings in response . . subexpression of a previously 'printed' expression, (Licklider, 1965, p. 182). as well as anticipating using the scopes for hand- (2) "In the study of computer-aided design . . . written input".^^-^ Ruyle et al., provide a 1967 graphical communication should be a two-way status report on systems developed for on-line math- process, for the designer wishes to enter ematical problem-solving including AMTRAN,^-^^'' drawings of objects into the computer for Culler-Fried,^^-'^ the Lincoln Reckoner,^''-'' analysis and likewise he wishes to see com- IViAP,5 32e MATHLAB.-32f and The work at Hudson puter-generated or modified drawings." Laboratories, Columbia University, is also to be (Langet al.,1965,p. 1). noted.-^-^-'* (3) "The engineer must do more than use the A general survey of computer applications to computer. He must actively participate in the design of computers is provided by Breuer (1966), computer solution. To do this he needs a who cites some 287 references. Also in the area of language to communicate with the computer, "shoes for the shoemaker's children", reactive physical accessibility to the computer, and a display and input devices are already being applied mechanism for obtaining engineering-oriented to machine-aided programming, flowcharting and ^•'^'^ results from the computer." (Roos, 1965, system design operations. It is noted also that 423). Anderson has used an IBM 1620 together with a p. Calcomp digital incremental plotter to produce (4) "We are therefore witnessing the first inroad annotated flow charts on line ^-^^ and commercial of machine assistance into the human decision packages for this purpose, such as AUTOFLOW,^^^ areas of printing design. It is ironical that •'^^^ or Autodigrammer ^ are now available. Then it is only a few years after printers and designers claimed that Bell Lab's FLOWTRACE can produce threw up their hands in horror at the idea of "flowcharts written in 'almost any' programming machines robbing them of their traditional language." (Sherman, 1966, p. 845). Other examples role in printing practice, we should see them include the IBM System/360 Flowchart ,^•35'' and being actively taken back into a partnership the Usercode Documentor and Flowlister,^-^^'^ with the machine in which their specialised while Abrams (1968) discusses a number of sys- talents are being very much enhanced by tems.-^-'^'^'' machine assistance." (Duncan, 1967, pp. Developments have been reported in machine- X— xi). aided circuit analysis, schematic circuit design, and (5) "An engineer may, for example, substitute system block diagram construction.^-''^ LeVier different values of components in a circuit and (1965) describes a program for the circuit design of observe the effect in the form of a picture of a transistorized flip flop. Wall and Falk (1966) the output wave that would be produced by describe an IBM program for electronic circuit the circuit in question under any combination analysis, and Shalla (1966) reports a mapping from of conditions. He may also record the various block diagrams of electronic digital circuits into list changes on microfilm and observe them later structures and the use of a list processing program via a projection device at the same console. for the Control Data 3600 to carry out circuit He may magnify sections of a diagram for analysis. In particular, some examples of actual detailed study and manipulation, retaining the hardware for an Argonne National Laboratory ability to return the original picture to the computer were tested. Then there are the CIRCAL screen at any time." (Silveira, 1965, p. 37).

25 Next, therefore, we may ask how conveniently the ijig)5A3a g^^^ Socratic System as described by potential system client may thus interact with the Muerzeig in 1965. IBM has developed the 1500 machine. Instructional System '^^ and Engvold and Hughes What processing operations are automatically of IBM describe (1967) adaptation of the 7044 available to the user as he watches and wishes to Graphic System to a teaching system which can be modify the data pertinent to his problem displayed interleaved with programming problems to be han- for him? In systems such as Sketchpad he may apply dled in a variety of languages at the display console. scaling and rotational transformations of the image Still other examples include the simulation by displayed,^ 38 in CALCULAID^ or MAP* he may students of experiments in chemical analysis ^-^^ call directly for regression or Fourier transform and a Bolt, Beranek and Newman development operations,^-39 and, in general, he needs to be able where the computer deliberately introduces am- to request as directly as possible the performance of biguities into an interactive system designed to a specified operation, a display of the consequences teach medical diagnosis. (Science, Oct. 29, 1965, of his proposed modifications, and the re-input of pp. 572-576). Byproduct advantages of CAI tech- specifications of the next processing operations to niques include automatic performance measure- be performed which he has determined on the basis ments of student progress.^'*^ Then it is pointed of the previous results.^"*" out that "remote terminals used for CAI can also Finally we note that utilization of direct outputs, be used to provide instructions for leading the whether by man or machine or by both, involves student to appropriate supplementary material also the possibility of reactive control and of in- stored on microform." (Burchinal, 1967, p. 4). terruptibility and re-direction of source data Crowell and Traedge, 1967, provide a review and automation, intermediate processing, intrapolation bibliography of developments in computer-assisted and extrapolation of results. Feedback of various instruction. In addition, they suggest areas for types is an important concern.^"*"^ The problems of continuing R&D efforts. They conclude, in par- man-machine interactive situations at the levels of ticular, that "before assigning system functions to terminal design, user-oriented languages, human either man or computer . . . much research is engineering, and behavioral or attitudinal factors needed in analyzing the decision-making process. wiU thus require continuing R&D attention.^^"'' When such information is added to our present store of knowledge concerning man-machine factors, greater efficiency in CAI systems will be realized." 5.1.3. Computer-Assisted Instruction and Problem- Solving (p. 423). Other commentators similarly suggest that a The problems of computer-aided instruction great deal of work remains to be done (Adams, 1966), present, at least in the general literature surveyed, that, "without discredit to the good work now under somewhat less in the way of continuing R&D way, . . . the total level of the effort is unacceptably implications is for than the case machine-aided small" (Mills, 1967, p. 237), and that "akhough design or machine-aided inference. Nevertheless, man-machine interaction may eventually revolu- the field of CAI represents an important area of tionize problem-solving, even the leaders in the development and application.^"*' In its computer field are just taking their first hesitant steps into the directory issue of June 1968, Computers and Auto- deep unknown of joint human-computer problem- mation provides a roster of 40 CAI centers and solving techniques." (Davis, 1966, p. 243). laboratories. Silvern and Silvern emphasize that: "CAI, computer-assisted instruction, developed from dissatisfaction with simple text formats and 5.2. Multiple Output Modes teaching machines which did Httle more than automatically turn pages for the student. Early For information processing systems generally, investigators saw, in the digital computer's ca- output, like input, should be available in a variety of pability, the seeds of two-way, learner-machine modalities — visual, as in printing and in graphic dialogues which would lead a student along a path displays; voice and other audio signal outputs; " which more primitive methods failed to travel. control signals, mechanical displacements of ^-^'^ (Silvern and Silvern, 1966, p. 57). pointers and plotting devices, and the Hke. Output Examples of CAI systems include the PLATO results may be presented immediately to the oper- (Programmed Logic for /Automatic Teaching ator or user, recorded photographically, or fed to Operations) system developed by the Coordinated various types of communication links, including Science Laboratory of the University of Illinois,^-^^ telephone, coaxial cable, and facsimile transmission the CLASS (Computer-Based Laboratory for^^uto- lines. Before discussing the more conventional forms mated School Systems) facilities developed at of output, such as the use of printers and plotters, System Development Corporation,-^ ''^ PLANIT and hard copy requirements generally, let us con- (Programming Language for /nteraction and Teach- sider in this section audio outputs; three-dimensional, color, and motion picture outputs, and some

examples of systems using various combinations of I *See note 5.32a. output techniques. !

26 5.2.1. Audio Outputs, Speech Synthesis and Speech literation of connected speech as related to reading Compression machines for the blind. Problems of formant amplitudes in synthesis of Unlike the problem of voice inputs, audio output natural-sounding speech have been studied by Fant devices are already available, at least for somewhat and his associates.5-^^^- while Woodward has been limited sets of response messages.^-^^^ Examples concerned with formant persistence and with are the IBM 7770 device with a fixed, pre-recorded intonation.^-5^^^ AUen (1968) points to a number of vocabulary of up to 128 response words and the R&D problems remaining in the area of the syn- IBM 7772 audio unit which by vocoder tech- thesis and output of connected speech. These niques can generate several thousand distinct ^-^^^ include problems of stress, punctuation and phrase words (in one or more languages) as auditory boundaries, intonation, and the hke.^-^"'' This outputs, the Cognitronic's Speechmaker ^-^^ an investigator is exploring a syntactic bracketing ap- RCA unit designed for use with Spectra systems,^-^*' proach to the solution of some of these problems. ^^^'^ developments by Burroughs.'-^"'^ equipment de- Becker and Poza (1968) are investigating natural veloped by \^ estinghouse Electric for computer- rather than synthetic speech for computer-controlled student communication at Stanford Lniversity,^-^' output and are concerned with the significant reduc- and an early model developed at the Rome Air tion of storage requirements for natural speech Development Center.^-^'^ w^aveforms. More particularly, they are concerned Application areas of interest range from direct with dyads (that is. the last half of one sound and assistance not only to students ^ but also to prac- the first half of a following sound). These dyads •^'^ to developed for ticing physicians:^ the systems may be manually inspected and marked out from the New York Stock Exchange (Proctor. 1966). for displays of typical speech segments and subse- operations.-^ and for engineering design banking quently recombined to provide at least some of the Aviation.'-^' information recall at North American natural transitions of speech for large vocabularies, that answerback systems Gentle suggests "voice but requiring only a relatively few number of dyads will increase in number as more computer manu- to be stored (e.g.. 800 for monotonic and 8.500 for facturers build the capabiUty into their equipment. intonated speech).^-^^*^ The increasing availabiUty of tone diahng telephones In terms of audio output and speech synthesis, w ill also hasten the use of these systems because the Bhimani et al. (1966) stress the importance of use of the regular telephone for inquiry and voice developing efficient means for automatic conversion answerback permits almost anyone in any location from the character sequences necessary for or- to use the service." (Gentle, 1965, 81)?-^°^ p. thographic representation of a word to the phonetic The closely related areas of speech synthesis and equivalent of that w'ord.^-^^ Other examples in this vocoding techniques are also important w ith respect area include the work of Rabiner ^-^^^ and that of to improved communication and processing systems 5.5Hg -pj^g converse of this problem, that of of the future because of the potentialities offered "spelled speech" is also being investigated. for significant compression of speech signals to be In the area of speech compression, in addition transmitted via various communication links. to examples cited in an earlier report on information at Laboratories R&D objectives the BeU Telephone acquisition, sensing, and input, the following is to and elsewhere are to the typical bandwidth reduce be noted: "Although audio communications is a of 78.000-56.000 bits per second for direct digital slow form of information exchange, advances have encoding of analog speech signals to 2.400. 1.200 been made in the production of audio tapes that or even fewer bps. more closely approximating the permit more information to be recorded per unit of probable information content of digital speech time. The President's Committee on the Employ- signals (i.e.. about 50 bits per second).^-^-^^ ment of the Handicapped, through the Division Examples of current and continuing R&D efforts of the Blind of the Library of Congress, is encourag- in the areas of speech synthesis and speech coming compressed speech research because it promises pression are to be found at the Royal Institute of to provide a method for recording two to three times Technology in Sweden,^-^^ at the University of as much information per linear inch of tape without Manchester. England:^-^''^ at M.I.T.,^-^' and at the distortion or loss of comprehension. Direct access Haskins Laboratories.^-^* In aU of these cases, and to stores of audio information in libraries offers in other organizations as well, there is considerable inexpensive and practical method for updating R&D interest in the design of improved aids to information frequently and distributing it to both the handicapped. For example, at R.I.T.. Risberg the sighted and the blind over estabhshed telephone has demonstrated speech analysis hardware with networks." (Becker. 1967, p. 8). rapid response displays of both spectrum and Two final special cases of audio output should also frequencies so that the deaf child can see the results be mentioned. They are those of the use of the of his attempts at speech, and can therefore make computer to produce musical tones ^-^"^ and of instantaneous adjustments. At M.I.T., a recent computer-assisted music research and composition. example involves developments in reading aids In a recent review, Forte (1967) indicates the fol- for the blind.-^ -^^ Similarly. Lee reports of a computer lowing principal areas of interest: experiments in speech generation technique designed for trans- auditory perception, the study of musical "gram-

27 366-108 O - 70 - 3 in systematic rotation about an indicated axis.^^^ In mars" in order to produce "new" music a i specified styple, music analysis as applied, for the latter case, there may be interesting require- i ^-^^ example, to pattern recognition procedures for the ments for handling problems of 'hidden lines'. determination of the style characteristics of a The hidden line problem has been further attacked particular composer or period, and the computer by Loutrel (1967) and Appel (1966), among others. | ' development of original musical compositions.^ ''"'^ A In addition, Appel (1968) has reported on techniques somewhat less exotic example is provided in ex- for machine-generated shading or toning of draw- j periments designed to use computer audio output ings.^-^*^ Perspective manipulations, shading, and to teach correct pitch to singers. display also provide representations of the variable depths of objects in space. ^-^^ A specific example of multiple input-output modes ji as applied to music research is provided by Teitel- Investigations into three-dimensional data dis- I man as follows: "The standard teletype of the DEC plays include questions of applicability to air

PDP-6 has been augmented by an 88 key piano traffic control operations,^-^^ to the possibility of |i keyboard which is connected directly to the com- creating half-tone, 3-D photographs of conceptual- I puter. Thus the user can play a melody, hear what it ized shapes and objects,''''*'^ to visual simula- i tions,^-^^'' sounds like — as performed by the PDP-6 — and and to the plotting of three-dimensional I ^•^'^ also see the score displayed visually on the screen. contour maps of the moon's surface. Denil He can then edit the score, using the Hght pen, the (1966) reports on the development of a design teletype, or the piano keyboard, and hear it played language (DLI), which involves a prototype com- again. Programs are being written to allow the user puter system to handle the design of three-dimen- ^^^^ to request the computer to fill in harmony to a par- sional objects. Beyond this, there are possi- ' ticular melody, or to construct variations on a bilities for n-dimensional display of hyperobjects. theme and to play them back to the user." (Teitel- In the latter case, the investigator concludes: "The i man, 1966, pp. 16-17). Still other intriguing possi- importance of the techniques presented ... is the bilities lie in the development of three-dimensional use of a digital computer to generate visual displays i data processing and display techniques, color, and of the three-dimensional projections of the hyper- ^ motion picture outputs, to be considered next. objects. Such displays of rotating hyperobjects

could be produced most efficiently by a computer i

5.2.2. Three-Dimensional, Color, and Motion Picture since the projections and drawing would be too | Outputs tedious and impractical to produce by any other method. Although no actual mental visualization of i first that "motion pictures, 3D, and It is noted the fourth dimension resulted from the computer- multicolor displays via color separation represent generated displays, it was at least possible to practical capabilities in addition to the more com- visually display the projections and be puzzled in of plotting and printing." (Peterson, mon forms attempting to imagine the rigid four-dimensional 1966, 131). Three-dimensional data processing, p. ^ hyperobjects." (Noll, 1967, p. 473). manipulation and display is of considerable current several years ago, Rosa reported that ques- ' interest not only for machine-aided design pur- As of tions of color, three-dimensional data display, and ' poses,^*''"^ but also for direct scientific and engi- display were matters of continuing con- neering findings. For example, a paper by Serhn size of troversy and concern. For effective radar and ' (1966) considers the use of three-dimensional com- sonar screen displays, many different colors may be puting and plotting techniques in terms of the partial required. On the other hand, as Hobbs points ' differentail equations involved in the cooling of a out, difficult design compromises may be required thick glass slab. ^-^^^ to achieve good performance, and Baker and .. Levinthal (1966) has developed techniques for indicate that desired capabilities have yet sequential display of two-dimensional output images Rugari ^-^^"^ to be achieved. • in such way as to simulate the on-line presentation of the three-dimensional model of a protein mole- A color display device, developed by the Digital > cule. Again, a computer-display system may be Equipment Corporation for the Air Force Cahibridge used in studies of human perception, such as Research Laboratories, uses a 3-gun tube to plot perceptual masking phenomena in sequential designated points on a 512 x 512 raster. Mazzarese \ perception. (Mayzner and Tresselt, 1966). Tate (1965) reports that: "Color has proven particularly comments: "Understanding of three-dimensional useful at AFCRL in filter design studies, which structure is an essential part of the growth of require overlaying graphic information correspond- science. Greatly increased attention must be focused ing to different ways of manipulating raw data. on developing the needed systems for storing the Different colors can code different parameters, or data available in the literature and manipulating varying intensities of one color can show the changes

' the stereochemical details for study." (Tate, 1967, with time of one value." p. 296). Further, "the National Aeronautics and Space

Output representations of three-dimensional data Administration has awarded the Philco Corp., two i may be either mechanical, as in a spatial data contracts to develop an experimental color tele- plotter,^ *^^ or they may be two-dimensional graphic vision display system for possible use in the mission displays involving variable perspective views and control center of the Manned Spacecraft Center in

28 Houston." (Electronics 38, No. 18, 40 (1965).) and Crane at the BeU Telephone Laboratories in Mahan (1968) reports on developments at ITT, 1928. Following the development of vocoder tech- ^-^^ Philco-Ford, General Electric, and Sylvania niques (Dudley, 1936) a tactile output vocoder was while Hartsuch (1968) mentions RCA and Crosfield built by Levine and others at M.I.T., in the period Electronics. ^^^'^ 1949-1951. Then there are possibilities for hybrid displays, A tactile vocoder at the Royal Institute of Tech- such as the following: "One intriguing prospect is a nology is designed along lines that are similar to hybrid display — a wall type panel made of solid these earlier lines. First, it uses an amplifier to state materials that emit light when activated by enhance the speech signal as it is received and then a light beam that is deflected electrically. Such a passes it through a differentiator to provide high- system would combine the advantages of the two frequency emphasis. The signal is next divided by principal approaches to optoelectronic display: the means of overlapping filters into ten channels of higher resolution and more modest switching require- different center frequencies. The output signals ments of light-beam 'writing', plus the multicolor from each channel are rectified and smoothed to capability, higher reliability, brighter light and provide a control voltage. Next, in turn, each of the " lower power consumption of panel displays. ten control voltages modulates the amplitude of a (Soref and McMahon, 1965, pp. 56-57). 300-cycle per second sinusoidal signal. When these Knowlton (1964) describes an ingenious technique signals have been amplified and adjusted for for computer generation of animated movies, such channel sensititivity, they are fed to ten bone- as those that might be desired for psychological transducers serving as vibrators to stimulate the experiments or for expository educational films or tips of the user's fingers from left to right across for visual display of computer processing results both hands in the order of the lowest to the highest including the dynamic drawing of flow charts or channel. (See Pickett, 1963). "" wiring diagrams.^ Then we note that: "Researchers at Stanford Another example of motion picture output is cited Research Institute under Dr. James C. Bliss have by Gomolak in connection with "electronic drafts- devised a 12 x 8 array of air jets which, under man" applications.^ More recently, at BeU Labo- control of a CDC 160-A, is capable of transmitting ratories, animated movies have been machine- a perceptible message." (Datamation 9, No. 11, generated with respect to the effects of different 61 (1963).) gyroscope constraints on satellites.' (NoU, 1966, The potential multipHcity of different input- p. 143).^ Then there is the VISTA system (Visual output modalities may be illustrated by two ex- Information for Satellite Telemetry Analysis).^-^^'' amples. First, the future potentialities of man- It is to be noted with respect to three-dimensional machine interchanges between verbal and graphic and dynamic displays that: "General three-dimen- representations of chemical structure and other sional graphical communication, which deals with diagrammatic information are just beginning to be arbitrary surfaces and space curve intersections, explored. Such possibilities are of considerable presents many difficult problems; the beginning has interest in such areas of application interest as been modest and much work remains before the studies of interactions of chemical structures with complete graphical communication problem is biological effects. Some of the possibilities may be solved" (Johnson, 1963, p. 347), and that: "The exemplified by the following case: Without having matter of three-dimensional displays is closely allied undertaken a prior literature search, the researcher to the problem of dynamic display and is still a has synthesized a new steroid compound in the matter of concern." (Rosa. 1965, p. 413). laboratory. He prepares a chemical code notation and/or structure-connection-table listing as input. 5.2.3. Multi-media and Special Purpose Direct Outputs As a result, he receives:

In a 1967 lecture, Mayeda discusses as a new a. An indication as to whether or not his presum- objective of system design the development of multi- ably "new" compound has been previously media networks, pointing out that, with the excep- reported in the literature and, if not, whether tion of certain military and intelligence applications, it is possibly patentable. most information processing systems have been b. A two- and preferably three-dimensional dis- "single path", but that new network requirements play of the structure he has described. This wiU necessitate the use of information processing may or may not be over-written in terms of technologies in combination as well as the "mixing coincidences with previously recorded infor- of normally incompatible media into one information mation. By pointing at specific subsets of the ''^ transfer system." (p. 4).-^ diagram, he will direct the system to retrieve In addition to the varied combinations of textual, and report all available data with respect to visual, and audio output presentations stressed by observed physical characteristics and bio- Mayeda (1967), a special purpose output mode may logical effects of known compounds containing be required, involving encoded tactile-pattern this specified subset structure. messages. For example, equipment providing tactile c. He may then add to or delete from listings of means for the reception and perception of speech multiple-substituent possibiUties at variable dates back to the "Teletactor" developed by Gault locations of attachment to the given nucleus.

29 and request display/report of the consequences nition. Between the simplest and most sophisticated in terms of previously known results. modes, there lies a spectrum of intermediate modes." (Quarterly Progress Report No. 80, Research As a second example, a reading-machine system Laboratory for Electronics, M.I.T., 217-218 (1966).) under development at M.I.T., "will include an Then there are developments in what has been opaque scanner for the sensing of printed matter, termed an . . Interaction Screen". This may be real-time data-processing facilities for real-time scan achieved by an electronic system "consisting of control, for operating upon the data, for controlling five major components: two consoles, one for each tactile and auditory displays, and for monitoring subject, through which, by operating dials and and analyzing the performance of human subjects. switches, they interact with one another; an experi- In its most sophisticated mode of operation, the menter's console for monitoring subjects' inter- system will recognize printed characters and gen- action; a control unit for altering the response erate artificial speech. A less sophisticated output possibilities open to the subjects; and a card punch mode will utilize Braille for tactile sensing and that records all interaction by both subjects." spelled speech for auditory reception. In its simplest (Sawyer and Friedell, 1965, p. 447). Beyond such mode of operation, the system will simply reproduce, interaction capabilities, large scale and high reso- as tactile patterns, the black-white shapes sensed lution dynamic displays may be desired for group by the scanner, without automatic character recog- interaction with processor systems. ^-^^

6. System Output to Microform and Hard Copy

In the preceding section, we have considered some of the remaining output mode possibilities. direct output and display primarily as associated Thus, in this section, we shall be concerned with with Box 5 of Figure 5. In fact, this arbitrary computer-based microform recording and direct association is indicative of some of the difficulties output transmission, with printing, photocomposi- of presentation in this and other reports of this tion, and character and symbol generation, and with series, since in many cases the distinction between requirements for hard copy output. these direct outputs and those outputs that follow post-processing operations (Box 13) is entirely 6.1. Microform Recording and Direct artificial. (Certainly the question of a multiplicity Output Transmission of output modalities and of multi-media output

presentations is at least as pertinent to Box 14 as it As we have seen in consideration of file media and is to Box 5.) data storage techniques (Section 4), there are both However, the intention in the preceding section advantages and disadvantages of either digitaUza- was intended to emphasize the possibiHties of tion or of document miniaturization in facsimile direct, on-line interaction between the user and the form for many different purposes. In many situa- system. In particular, as previously noted, new tions a judicious mixture of the two approaches may problem-solving capabilities are offered by advanced be the best solution.^ ' To the extent that document information processing techniques, especially in or data miniaturization processes are adopted, the form of a two-way process of graphical input and however, there are specific questions of direct output and of effectually on-line communication output, post-processing operations, and delayed or between the problem-solver and the machine so off-line output. These questions include output that, at the man's own pace, he is able to study the recording of processed material to microforms, effects of modifications and changes by having the chentele requirments for hard copy, quality of results displayed to him almost instantaneously.-^-^*^ display and reproduction, output via transmission By way of contrast, in discussing the areas of fines in various communication systems, and on-fine R&D concern with respect to Box 14 of Figure 5, alteration of microform images. we wiU consider those output techniques, media, and An inportant recent trend in microform recording products that may be produced either on-line or has been an increased coupling of computer process- off-line and that may often involve the use of various ing to either off-line or on-line outputs into one or intermediate media of storage and/or transfer. It is more types of microforms. ^-^ In a 1965 survey, of course also arbitrary to consider the system out- Kornblum predicted that microfilm, microfiche, and put products of Box 14 prior to discussion of the microfilm aperature card techniques would have special post-processing operations (Box 13) that increasingly more challenging applications, specifi- typically precede them (this is also true with many cally including on-line tie-ins to computer systems if not all of the techniques and products considered and to multiple-access data banks. He cited, in in association with Box 5). However, since post- particular, the evaluations of the Micro-Data processing operations such as transHterations and Division of Bell and Howell and of NCR.*'-^

re-formatting are often closely dependent upon By 1967, it could be specifically suggested that: choices of specific output modes, we have delib- "The genuine advantages of microform computer erately deferred discussion until we have covered output indicate that all output, except possibly that

30 ^

used for debugging purposes, should be converted With respect to direct output transmission via directly to some sort of microform, eliminating the various types of communication links, it is noted intermediate stage of hardcopy. Indeed, IBM's that Long Distance Xerography techniques de- listings of systems programs produced for its veloped by Xerox can be used for transmission of customer- and systems-engineers are stored on graphic materials from computer-based stores,^ ''^ as microfiche. The ease and cost of getting hardcopy can other types of scanning converters transmitting when essential will naturally depend on progress in to remote facsimile recorders. the microform reader/printer technology; the high A more recent Xerox development is the Graphic stakes involved in being able to handle computer Terminal Hardcopy Printer, used at the Bell output should encourage such progress." (Van Telephone Laboratories in conjunction with a Dam and Michener. 1967, p. 192). computer-driven data display system. ^ '''^ A special- Equipment such as the Stromberg-Carlson purpose device, the Microteleviser, providing for SC 4020 microfilm plotter and later "Microma- the remote viewing and continuous focus magnification" equipment (Kalagher. 1968. p. 37).6 « an IBM tion of microfilm images, was introduced by General system ^-^^ and the 3M Electron Beam Recorder ^-^ Precision, Inc., at the 1965 WESCON exhibition. are commercially available for such purposes, Output via closed-circuit TV systems is also offering in the latter case speeds of up to 60,000 presently Hmited primarily because of high costs Unes per minute. At Bell Laboratories, such an for use of coaxial cable transmission. However, output mode has been available since 1961. and it various trials of such systems have been and are now includes computer-generated movie capa- being made — for example, the Council on Library bilities. ^^'^ In addition, a package of FORTRAN Resources has supported studies of the "Tele- programs has been developed that provides for reference" system involving closed-circuit TV and a drawings of orthographic views of combinations of remotely controlled card manipulator for multiple plane and quadric surfaces that are output to a external references to a library card catalog at a microfilm recorder. (Weiss, 1966). central location,^-^ a closed-circuit system has been On-line modification of displayed microform- installed for the Lake County libraries so that the recorded graphic images is specifically included in common collection can be appropriately shared,^"* many machine-aided design and machine-assisted and a laboratory use of such a link for graphical problem-solving applications, including the pre- presentations of computer output has been reported viously mentioned DAC (Design Aligmented by by Neilsen.^-^^ In addition, Brookhaven National Computer) system,^-^ and others (see p. 23). Laboratory operates a network for video storage and A first factor involving the quality of display and display at the local stations. subsequent copy reproduction of materials proc- Examples of commercial developments include essed and stored as microforms involves the quahty Remstar, which couples closed circuit TV trans- of the input material.*'-^^ A second factor involving mission with automatic records retreival equip- quahty of display and reproduction is that of proper ment.^ Then, in Great Britain, a patent (No. maintenance of equipment as well as of the 997,946) has been granted to Pye, Ltd., for a system microforms themselves. that converts normal high-speed TV signals into From a systems design point of view, however, the low speed signals generated with a slow scanning most critical factors are those reflecting client speed and occupying a narrow bandwidth (see willingness to accept and use the products and Electronics Weekly, No. 258, for Aug. 11, 1965). processes of microform techniques. For example, it Dixon Industries, Inc., has developed a video sys- has been quite widely assumed for some years that tem, VIDEO— IR, allowing remote transmission via the use of microforms for documentary and library slow-scan methods over ordinary telephone Hues, reference purposes is generally unacceptable to and offering a unitized tape record the size of a scientific and technical users. With improvements punched card to hold up to 150 pages, as well as in the availability of microforms and in the tech- tape reels. (Sci. Inf. Notes 7, No. 1, 2 (1965).) nology of use however, there is now evidence of growing acceptance and in some cases of certain " advantages to the user.^ 6.2. Printing, Photocomposition and Nevertheless, the systems planner and designer Computer-Controlled Character and must still address himself to the questions of how Symbol Generation convenient, how compact, and how portable available viewers having good quality legibility are? Can The earliest and stiU most common form of output hard copy, enlarged printouts be obtained either of information processing system results is that of directly or indirectly on a selective basis from such typing or printing. In the past few years, increasingly devices or must there be resort to separate processes automated techniques have emerged for use in involving other equipment? Further, it has been pubUcation and printing processes. These range noted that "the general problem of a satisfactory from use of direct computer printouts, using high hand-reader for microform stiU remains to be speed, but low quahty printers, through computer- solved." (Council on Library Resources, 1966, p. generated tapes that drive automatic typesetting 52). Wooster. (1968) however, suggests a "cuddly" operations to batteries of low speed composer- microform reader, as shown in Fig. 7. printer devices, to computer control of high-speed

31 32 composition systems that do provide output of good example is the development of Miniaturized Tech- typographic quaHty. There are obviously continuing niques for Printing, or Miniprint, by NCR.^ '^'^ On problems of advanced system design, in terms of the other hand, higher speeds may result in the loss appropriate comprises between speed and quality.^ of visibility for the user of an on-line terminal.^ '*'' Nisenofif reports that, as of 1966, "several de- 6.2.1. Line Printers velopments were undertaken to produce ultra-high- speed printing mechanisms of a non-mechanical In general, direct, on-line printouts from informa- nature. Radiation, Inc., for example, built an tion processing systems are made at high speed, electrostatic printer which required special electro- but with monocase limitations.^'^ fixed inter- sensitive paper. With the paper moving at nearly character and interword spacing, and poor utiliza- 100 inches per second, a maximum printing rate of tion of paper.^ In the case of some chain printers, 60,000 characters per second (500 lines per second such as the IBM 1403 '^^'^ and the Potter Chain or 30,000 lines per minute) was attained. The cost, Printer.^ '^ somewhat larger character sets than the however, was high for general use of this type of single case set can be accommodated. Then it is subsystem." (Nisenoff, 1966, p. 1824)^'^' to be noted that "'use of 2-case print mechanisms on high-speed computer printers for the display 6.2.2. Photocomposition and Electronic Character Gen- of extended chemical character sets has been under eration Techniques joint development by IBM and Chemical Abstracts." A major type of non-impact printing technique (Burger, 1964. p. 2). is represented by the various types of photocomposi- In addition, it should be noted that "printing speed for these [chain] printers depends both on the tion equipment. By contrast with impact printers, information to be printed and the character arrange- with other types of non-impact printing, and with ment. Since most printing applications use some hot-lead typecasting machines, the photocomposition techniques involve the transfer of optical characters more often than others it may be possible to increase the printing speed by repeating high images of characters from a negative master to a usage characters on the chain more frequently photographic emulsion, with suitable provisions for illumination, than low usage ones." (Eichelberger et al., 1968, image magnification or reduction, and positioning. Many techniques and configurations of p. 130). Further, "given character usage statistics mechanical, optical and electronic devices and de- we can (1) reliably estimate printing speed for a velopments have been employed for such purposes. given character chain arrangement and (2) improve the arrangement so as to increase the average print- Typographic quality output is achieved in photocomposition techniques by the automatic selection ing speed." (Ibid, p. 136). There are continuing trends for higher and higher of the right one of several sets of characters, by speed of printer output, especially in the area of point size adjustments for the individual character non-impact printing devices, since, as Becker points including those necessary for subscripts and out, 1200 lines per minute is "about the maximum superscripts.®-'*'^ by controlled positioning of the speed of a line printer because there is a limit to character as projected, and by insertion of special '^ the firing time available for mechanically operating symbols and characters.® '" generation, a print hammer." (Becker, 1967, p. 2).^ Advanced The accurate placement, and perman- hardware techniques for output printing include a ent storage of hundreds or thousands of high-quality thermal imprinting development by NCR for the symbols per second presently require that optical- U.S. Army Electronics Command.*" photographic techniques be employed, since the In a brief 1965 survey of non-impact printers, speeds are too great to permit mechanical symbol Webster notes first a facsimile process technique selection. Photography on film or paper is the cur- which forms a latent character image on electro- rently preferred recording method, although other conductive paper by pulsing selected wires in a graphic storage techniques, such as electrostatic, matrix of styli. The visible image is formed by liquid thermographic, thermoplastic, photochromic or or power toner that adheres only to the image area video signals on magnetic tape offer alternative -" and is bonded by heat and/ or pressure. Representing recording possibilities.® this approach today are small serial devices such as More than two decades of development in photo- Motorola's TP-4000, and "high-volume printers composition techniques (from the Higgonet and such as the Radiation 'Super-Speed' printer — over Moyroud "Lumitype" and "Photon" patent appli- 31,000 lines per minute". cations of 1945) have culminated as of today in a Planned variations in the TP-4000 teleprinter wide variety of available devices in a broad range equipment also illustrate new developments of of costs and performance characteristics. At one interest in the multiple access and man-machine extreme, we find 6-10 character per second devices. interactive systems. "This new teleprinter will . . . At the other end of the speed spectrum is the Benson provide a single facility for the human operator to Lehner Transdata system with a theoretical text make inquiry of a computer at his maximum but composing speed of 62,500 characters per second, plodding rate of 10 to 15 characters per second, and but with a character set (or "vocabulary") limited receive his answer back at a rate of 300 characters to less than 100 characters or symbols. The inter- per second." (De Paris. 1965. p. 49).^-'^ Another mediate speed range is bracketed by the Photon

33 900 series (300-500 cps) and the specified Linotron ards (Lee and Worral [eds.], 1968), as well as by the rates of 1,000-10,000 cps. Hartsuch reports 1967 Stevens and Little review (1967). developments by Photon, Fairchild Graphic Equip- It is to be noted in particular that "the Govern- ment, American Type Founders, Alphanumeric Inc., ment Printing Office was instrumental in the de- Harris-Intertype, and K. S. Paul and Associates, velopment of computer programs and production among others. techniques which have made it possible to utilize Among the electronic character generation tech- the most recent developments in printing technology niques noted in Webster's 1965 survey w^ere the and to apply automation to the publication of . . . A. B. Dick Videograph,^-^^ the Stromberg-Carlson [the] latest edition of the Library of Congress sub-

Charactron shaped-beam tube,^-^^ and "the . . . ject heading list. This is the largest single pubHca- Httle-known Xeronic High-Speed Computer Output tion which has been produced by the Government Printer developed by Rank Data Systems Division Printing Office using computers and photo composi-

. . . Character formation is by a pair of cathode ray tion." (L.C. Info. Bull. 25, 612, 1966). Further, tubes which form the latent image on the charged there are even computer typesetting procedures paper, which is developed by a continuous xero- designed for publication of computer language graphic system. The 'form' repertoire is held on a specifications, such as that developed for Autocode strip of film which is superimposed over the frame AB, in Sweden, (von Sydow, 1967). of variable information." (Webster, 1965, p. 45). The field of typographic-quality composition and Character generation techniques for high quality printing actually comprises several quite different as well as high speed composition are employed in technologies. However, a number of these tech- the Linotron system,*'-^ in RCA developments,''-^ niques may be made to converge in the design of a and in the Digiset equipment developed by Dr. single large-scale system. At one extreme, a skilled, Rudolf Hell.^-^^ An entry of Bell Laboratories is highly trained operator must perform all of the com- described by Mathews and Miller (1965) as having positor functions (spacing, leading, horizontal and the following graphic arts display characteristics: vertical justification, hyphenation, page makeup) "The unit can produce on a single scope face high required to operate typesetting equipment; at the quality printing with as much as 200 lines of 350 other extreme, text copy with or without font, style, letters each . . . at speeds of 500 to 5,000 characters/ size, or format specification can be processed with a second. . . . Since the image is described in com- high degree of automation through all the operations plete generality digitally, line drawings, mathe- that are necessary to produce type or plates set for matical equations, musical scores, and an unlimited printing. number of type fonts can be produced by the same, Among the pertinent technologies are: special- completely standard, means." (p. 397) More recent purpose devices or computers to provide justifica- developments in high speed character generation for tion, re-spacing and hyphenation information phototypesetting also include the IBM 2680 CRT automatically; ^ photocomposition developments; printer.^-^^^ electronic character generation devices for both on-line and off-line applications, and the program- 6.2.3. Computer-Controlled Typographic Composition ming of general-purpose computers to accomphsh the type compositor functions. Thus, in the past several years, three major, and Improvements in size of character repertoire and often closely interrelated, developments in hard provisions for rapid switching between character copy production technology have gained con- sets under automatic control are under active de- siderable momentum. These are: the development velopment. Some systems incorporate possiblities of techniques for computer-based typesetting, infor program-controlled, light-pen inserted, or cluding automatic justification and hyphenation; reproduced-from-storage interpolations of graphic the further improvement and widespread use of material with text, ranging from formatting grids photocomposition techniques, and the development and lines (in order to separate or to emphasize of both off-line and on-line devices for high speed tabular or columnar matter) to the incorporation of character and symbol generation, including the reproduced photographs with half-tone fidelity.^-^®^ recent breakthrough in "computer calligraphy", Other combinations may permit the intermixture or notably, at the U.S. Naval Weapons Laboratory, superpositioning of pre-recorded graphic informa- Dahlgren.6 tion (including pictorial data) with that of text or The automation of the traditional compositor graphics generated effectively on-line.^''^'' For functions is well on its way toward revolutionizing example, the new Photon 901 allows variable char- the printing industry. Overviews of this field have acter density, together with other improvements of been provided by Barnett (1966), by publication of the 900 (or Zip), and in addition provides for the the Proceedings of the two Computer Typesetting insertion of graphics from projected slides.^-^^'^ Conferences (the Institute of Printing, 1965, 1967), Going beyond the direct composition onto the and of a 1965 Symposium sponsored by the Center page of two-dimensional symbols, many apphcations for Technology and Administration of The American of the future will demand the automatic insertion University (Hattery and Bush [eds.], 1965), and by a of previously stored graphic material such as Symposium held at the National Bureau of Stand- drawings, illustrations and photographs and of

34 similar material received via communication means should be a way to capture any frame of the display such as facsimile. Actually, this type of sophisticated in the form of 'hard copy', and the hard copy should output has been investigated in several quite early be coded automatically for machine fifing, retrieval, projects,''-" and it has been claimed that "there and redisplay." (1965, p. 94). It is, however, im- appears to be no real technical barrier to the simul- portant to determine the extent to which hard copy taneous or practically simultaneous composition- options are actually needed, rather than merely imposition of both text and graphics." (Duncan, desirable.^-" 1964. p. 135). In terms of technological developments related to The Linotrons and related equipment built for hard-copy output options we note various types of the Air Force wiU provide for the generation of electrostatic, magnetic, and smoke photographic graphical material, such as line art illustrations, methods."^" We note further that RADC has continuous tone illustrations, and screened halftone awarded Xerox Corp., a contract to build a selective illustrations in a separate video display system, with photocopier that will separate printed text from optical mixing of the lexical and graphical ma- illustrative and pictorial material.^^^ Another terial.^ Beyond such questions are those of "Com- example is the Bolt, Beranek and Newman puter Calligraphy" (Hershey, 1967), including "Symbiont", a computer program system designed problems of computer-controlled composition of to facihtate the use of technical documents by chemical structure diagrams and mathematical students that includes an automatic note-taking and formulas, to be considered in a more general dis- copy feature for student-selected portions of text cussion of character sets and display terminal (Bobrow et al., 1966). Thus, photocopying techniques design requirements in a later report in this series. continue to be an important means of system As of 1967, then, the computer has demonstrated response. However, such techniques are encounter- capabilities for ehminating the need for complex ing steadily increasing difficulties in terms of coping compositor skiUs on the part of keyboard operators, with the steady growth in volume of demand. There for providing significant increases in the speed with have been complaints, moreover, that continuing which type composition functions may be carried development efforts are required in the conversion out, and for keeping pace with those developments of microfilm to hard copy, particularly in the areas in electronic photocomposers that promise graphic of speeding up the processes and reducing the arts quahty typesetting at speeds of several thou- cost of copies.^-^'^ sand characters per second and the superimposition of line art, screened halftones, and continuous 6.4. Post-Processing Operations and Special tone illustrations. Forms of Output

6.3. Hard Copy Reproduction In the case of computer typesetting, photocomposition, and character generation techniques, the It is obvious that direct printout, delayed printout, post-processing operations of Box 13, Figure 3, are photocomposition, and symbol and character gener- represented first by formatting processes to provide ation techniques can provide the typical system horizontal and vertical justification, hyphenation, client with hard-copy versions of the information and the Hke. processed by a system or contained in or retrieved On-line formatting in terms of output through from its files. In other cases, however, the primary either computer-printer or computer-controlled system output (whether in Box 5 or Box 14 of typesetting devices is typically limited to intra-line Figure 3) is some type of visual display of complete and interlinear spacing control. As we have seen, items or their surrogate representations which the controlled variation of printing weight may also be user scans. Such displays may also be accompanied available through repetitive overprinting of the by address-locators such that the item in its entirety same character in the same spatial position on the may be retrieved or reproduced in facsimile from output media (e.g., special chain-printer arrange- its location in physical storage. In addition, the ments). In certain ingenious applications even so option of hard-copy or facsimile reproduction may Umited a device as the high-speed line printer has often be desired from the system response and been used to simulate gray-scale variation in two- display as such.^ -^^ dimensional graphic representations by choice of In particular, the provision of hard-copy retrieval heavily "inked" character symbols to fill in darker options is important in most information selection areas and those characters involving relatively less and retrieval systems involving storage of docu- weight for the lighter areas. With the introduction ments and data in microforms. While some con- of electronic generation and display techniques, troversy exists with respect to the extent and however, on-line formatting potentialities take on urgency of client requirements, user satisfaction much more versatility. considerations would appear to dictate continuing Other than for formatted printout, output response agreement with such commentators as Mikhailov of may be required at several different levels. Post- VINITI in 1962, "the production of a large number processing operations will thus involve re-arrange- of copies directly from microfilm must be made ment, re-formatting and transliteration or translation ." possible . . (1962, p. 51) and of Licklider with of the direct results of processing, matching, selec- respect to an on-line system of the future: "There tion, and/or retrieval operations into forms and into

35 languages required for subsequent presentation, intended to be used in connection with the develop- transmission, and use to the system clientele.^-'^^'' ment of reading aids for the bUnd.^-^^ For example. Press and Rogers describe TRACE, Post-processing operations will also include, as which "is a program with extensive capability to appropriate, various summarizations and condensa- edit, reconfigure, and display multivariate data." tions of the information that has been processed or (Press and Rogers, 1967, p. 37)^^'' retrieved. For example: "Large-scale data-process- In such post-processing operations, for another ing systems have been planned in which the input example, we note a speech recognition requirement: is in the form of on-the-spot observations in special "The speech-recognition device, after having iden- formats. Retrieval is then required, not just of the tified phonetic stretches on the basis of their observations themselves, but of [inferences] from perceptual characterisitics, must transform them them, through conditional search, summing, averag- into strings of linguistic signs . . (Garvin, 1963, ing, correlating, finding maxima, reporting sudden p. 113). As we have seen in an earfier report, changes of state, etc." (Bohnert, 1963, p. 155). speech compression requirements are directly Yet, as of today "too many systems are designed related to problems of achieving more efficient and to display all the data, and not to display only the more economical use of communication and data data needed for the decisions the system is called transmission links. upon to make." (Fubini, 1964, p. 2). Postprocessing operations may also be applied Finally we note that system outputs based upon in order to provide special forms of output that are post-processing operations will potentially involve dictated for particular potential appfications; feedback information, such as activity analysis, or transliteration into Braille,^-^^ into synthetic or pre- the following: "A programming capability is to be recorded spoken messages, and translation from provided which, when initiated, will tally the number orthographic word records into phonemic form,*'-^'^ of requests for specified combinations of data for example. Nelson (1965) suggests a relatively items thereby enabling analysis by operating person- simple device for coupling blind users to on-line nel, leading to refinement of the output formats and text libraries via a Braille display. output request procedures through experience with Another special form of output for the handi- the system." (Geddes et al., 1963, p. 23). In par- capped is to produce from information processing ticular, output becomes input again in many of the equipment apphed to speech analysis rapid displays advanced techniques that have been investigated of both spectrum and frequency analyses so that and demonstrated. More expecially, feedback in- a totally deaf child can see the results of his attempts formation, whether generated by man or by machine, to speak and compare them with desired per- may control various "threshold" settings as shown in formance in order to make rapid adjustments.'''-^'* Figure 1 and other figures in this and other reports "Spelled speech" is another output variant that is in the series.

7. System Use and Evaluation

With respect to Box 15 of Figure 3, that of use and unique combinations of objectives. It is doubtful evaluation of information processing system that the comparative tests that have been made or products, we will consider many of the current that are planned can decide the relative merits problems and difficulties in more detail in another of systems for general application. They can only report in connection with the severe case of evalua- help to decide on the merits of certain particular tion of information storage, selection, and retrieval systems for certain particular applications. Hence (ISSR) systems. However, we note here a few there is real danger in drawing general conclusions examples of current concern with problems of from experience that is as yet too limited." (Lee and systems evaluation methodology generally. Campbell, 1959, p. 1555). Basic problems of evaluation methodology relate Current techniques for system evaluation pose first to the size, complexity, and open-endedness of many difficult problems and have been severely the systems themselves; the number of inter- criticised as to their present inadequacies. For dependent variables; the difficulties of defining and example, to Bryant's complaints about 'lack of isolating variables, and the amount of data to be well-defined objectives' and 'lack of meaningful collected and processed.'^-' In particular, it is models', (1966, presented in 1964) Davis adds the noted that: "Information collected about a very following difficulties: "Uncertainty concerning large system may become obsolete faster than it measures of effectiveness", those of determining can be collected." (McCarn, 1965, p. 95). system scope, and lack of quantitative evaluation It also still appears obvious, as of 1967, that criteria, as well as those of security or proprietary "comparison and evaluation of systems are still interests and those where disinterest in evaluation very hazardous, because of the many variables that results because there are no obvious penalties for enter into the design of any particular system and poor system design. (Davis, 1965, p. 81). the fact that the design of any system is decided An obviously difficult problem is that of identifying to some extent on specific objectives and somewhat meaningful evaluation criteria.'' - Even with obvious

36 objective system performance criteria, such as substitutes one-for-the-other. Test problems, simu- cost, equipment reliability, or response time, many lating the actual system usage, will show up system practical difficulties are encountered. Both true defects not detectable through the application of difficult costs and true benefits are often or even measures of effectiveness." (Davis, 1965, p. 82). impossible to measure. True costs, for example, Nevertheless, it is to be emphasized that "unfor- certainly include those of the personal time and tunately, no real guidehnes are available which effort required of the potential user, while true could be used in the design of evaluation procedures, benefits may be intangible, for example, in the sense and most of the methods actually proposed are of the negative benefits of not missing pertinent based on ad hoc rules which stress theoretically information. desirable features, and do not concern themselves The questions of cHent-utilization of system prod- with practical questions. As a result, much of the ucts, with special attention to the interrelated prob- proposed methodology cannot, in fact, be imple- lems of system evaluation, are of particular concern. mented reasonably in a test situation." (Salton, We are faced, first of aU, with the lack of quantita- 1965, p. 201). tive measures of effectiveness, with subjectively Even more discouraging, perhaps, is the opinion rather than objectively determined performance of some investigators that the development of factors, and with a variety of uncontrolled variables evaluation methodology itself is even less advanced involving human factors." than that of system analysis and testing. Thus, Research and development requirements in the "one can raise reasonable doubts, however, that the area of system evaluation involve the establishment development of evaluation methodology has kept of research corpora, "core" problems, and "test- pace with the development of the systems." (Morey bed" data banks against which comparative experi- and Yntema, 1965, p. 349). Further, "progress in ments may be run. Better system models and more designing and applying information handhng sys- precisely selected and defined measures of perform- tems has outraced progress in evaluating their ance effectiveness are obviously required. Morey performance. As system complexity has increased, and Yntema emphasize that: "As the economic and the relative adequacy of existing evaluation tools operational implications of man-computer systems has decreased." (Calingaert, 1967, p. 12). loom larger, the need for more powerful evaluation We may also concur with the opinion of Blunt tools and techniques increases. System testing, as that "the system engineer presently lacks sufficient the prime source of data upon which any evaluation tools to efficiently design, modify or evaluate com- is based, shares this need for improved method- plex information systems." (Blunt, 1965, p. 69). ology." (Morey and Yntema, 1965, p. 349). Several of the new tools that are being developed We need both better measures of system effec- include formal modeUing and simulation techniques, tiveness and the development of "core" or "bench- but we shall discuss these in terms of overall system mark" problems for purposes of test. As Davis design questions, to be considered in the next report comments: "Measures of effectiveness and use of in this series. test problems are complementary rather than being

8. Conclusion

In this second report, a number of developments predictions of hardware developments for the 1970's have been adduced, and a number of areas of con- recapitulates many of the areas of hardware R&D tinuing R&D concern have been indicated, with concern that have been indicated in various sections respect to processing operations, processor system of this report. In particular, he points to the following management, storage, postprocessing operations, major problem areas: and on-line and off-line outputs from generalized information systems. As we have seen, for example, "Low-cost, large capacity, alterable on-line output products involving graphic two-dimensional storage. or three-dimensional and voice response represen- Low-cost, reliable input/output equipment. tations of stored data are relatively new but are of Proper selection and organization of storage increasing importance in on-line, reactive, selection hierarchies. and retrieval systems and, in particular, for systems Low-cost inquiry/display consoles for man- oriented toward machine-aided design applications. machine interaction. Two-dimensional data outputs are already widely Dynamic real-time large-screen displays. available both in the form of facsimile and TV System design and application concepts that recordings and transmissions and in the form of minimize input/output operations at the ex- dynamic, on-line displays, but these are all areas pense of more internal logic and memory. requiring extensive further efforts and developments Improved concepts of increased functional aimed, in particular, at lower cost. With respect to modularity to utilize large interconnected hardware requirements alone, Hobbs in his 1966 circuit arrays.

37 Improved automatic fault isolation and main- Other considerations, such as those of program- tainability. ming and other software requirements, are also of obvious concern. In many cases, however, such Economic associative memories or other hard- considerations relate to the many interrelated ware techniques to facilitate file access by functions a generalized and not merely, content. of system for example, to input rather than output or vice Protection of data base in multi-computer, versa. For these reasons, we will consider overall systems with many remote users. multi-user system design requirements (including problems Communications between remote user consoles of system and network design, programming and the processor, between different proc- languages and related software developments, essors, and between remote processors and advanced hardware technologies, and the like), the data base." (Hobbs, 1966, p. 44). in a third report in this series.

38 Appendix A. Background Notes on Research and Development

Requirements in Information Processing, Storage, and Output

In this Appendix we present further discussion formation processing systems. A number of illustra- and background material intended to highhght tive examples, pertinent quotations from the currently identifiable research and development literature, and references to current R and D efforts requirements in the broad field of the computer have been assembled. These background notes have and information sciences, with emphasis upon been referenced, as appropriate, in the summary processing, storage, and output functions of in- text.

1. Introduction

1.1 There are certain obvious difficulties with of this report, its necessary selectivity, and the respect to the organization of material for a series problems of organization and emphasis are to be of reports on research and development require- noted. Obviously, the reviewer's interests and ments in the computer and information sciences and limitations will emerge at least indirectly in terms technologies. These problems stem from the over- of the selectivity that has been applied. laps between functional areas in which man- In general, controversial opinions expressed or machine interactions of both communication and implied in any of the reports in this series are the control are sought; the techniques, tools, and sole responsibility of the author(s) of that report and instrumentation available to achieve such inter- are not intended in any way to represent the official actions, and the wide variety of application areas policies of the Center for Computer Sciences and involved. Technology, the National Bureau of Standards, The material that has been collected and re- or the Department of Commerce. However, every viewed to date is so multifaceted and so extensive effort has been made to buttress potentially contro- as to require organization into reasonably tractable versial statements or imphcations either with direct (but arbitrary) subdivisions. For example, the pres- quotations or with illustrative examples from the ent report is concerned with problems of manage- pertinent literature in the field. ment of multiple access systems in terms of processing system service requests, among other The author of the present report must apologize topics, but questions of programming languages and for a necessary selectivity and for many inadvertent other overall system design considerations will be omissions in coverage. It will be appreciated if discussed in a separate report. specific omissions are called to our attention. It is to Other topics to be covered in separate reports in be noted that neither inclusion nor citation is in- this series will include specific problems of informa- tended in any way to represent an endorsement of tion storage, selection and retrieval systems and any specific commercially available device or sys- the questions of maintaining the integrity of priv- tem, of any particular investigator's results with respect to those of others, or of project ileged files (i.e., some of the background considera- named tions with respect to the issues of privacy, objectives. Conversely, omissions are in no sense confidentiality and/or security in the case of intended to imply adverse evaluations of products, media, equipment, systems, project goals and multiply-accessed, machine-based files, data banks, and computer-communication networks). project results, or of bibliographic references not included. In general, the plan of attack in each individual report in the series will be to outline in relatively There will be quite obvious objections to this short discursive text the topics of concern, sup- necessary selectivity from readers who are also plemented by background notes and quotations and R&D workers in the fields involved as to the repre- by an appendix giving the bibliographic citations of sentativeness of cited contributions from their own quoted references. It is planned, however, that there work or that of others. Such criticisms are almost will be a comprehensive summary, bibliography, and inevitable. Nevertheless, these reports are not index for the series as a whole. intended to be state-of-the-art reviews as such, but Since problems of organization, terminology, and rather they are intended to provide provocative coverage have aU been difficult in the preparation of suggestions for further R&D efforts. Selectivity this series of reports, certain disclaimers and must also relate to a necessarily arbitrary cut-off observations with respect to the purpose and scope date in terms of the literature covered.

39 These reports, subject to the foregoing caveats, access from and to many users at many locations. are offered as possible contributions to the under- This implies multiple inputs in parallel, system standing of the general state of the art, especially interruptibility. and interlacings of computer pro- with respect to long-range research possibilities in grams. It is assumed, further, that the overall a variety of disciplines that are potentially applicable scheme involves hierarchies of systems, devices, to information processing problems. The reports and procedures; that processing involves muhi-step are therefore directed to a varied audience among operations, and that multi-mode operation is whom are those plan, who conduct, and support possible, depending on job requirements, prior or research in these varied disciphnes. They are also tentative results, accessibihty, costs, and the like. addressed to appHcations specialists who may hope It should be noted, next, that techniques suggested eventually to profit from the results of current for a specific system may apply to more than one research efforts. Inevitably, there must be some operational box or function shown in the generaHzed repetitions of the obvious or over-simplifications of diagram of Figure 1. Similarly, in a specific system, certain topics for some readers, and there must also the various operations or processes may occur in be some too brief or inadequately explained discus- different sequences (including iterations) and sions of other topics for these and other readers. several different ones may be combined in various What is at best tutorial for one may be difficult for ways. Thus, for example, questions of remote another to follow. It is hoped, however, that the console design may affect original item input, input notes and bibliographic citations will provide of processing service requests, output, and entry of sufficient clues for further follow-up as desired. feedback information from the user or the system 1.2 Certain features of the information flow and chent. The specific solutions adopted may be process schema of Figure 1 are to be noted. It is implemented in each of these operational areas, or assumed, first, that the generalized information combined into one, e.g., by requiring all inputs and processing system should provide for automatic outputs to flow through the same hardware.

2. Processor System Considerations

2.1 "Besides being useful to individual users "One could say that time-sharing is the use of a who wish to borrow each other's routines, a sharing multiprogrammed processor for the purpose of mechanism is also useful to a group of users who permitting a number of users to operate the proc- wish to reference certain segments in common. essor in such a way that each is unaware of the use Such segments might be a set of library routines, of the processor by the others." (CoUila, 1966, p. 51). or a set of procedures making up a programming "We may, therefore, expect to see in the next few language system." (Dennis and Van Horn, 1965, years a meld of time sharing and multiprocessing, p. 31). i.e., simultaneous employment of two or more "With such a mechanism [common program- time-shared processors in association with a com- segment use], it is fruitful if the scheduHng logic mon memory, which in a sense wiU be both time- clusters 'like' programs in the schedule or queue shared and space-shared." (Licklider, 1965, p. 26.) and thereby amortizes program-code swap time "By the spring of 1967 . . . there were some 37 across many users." (Schwartz et al., 1965, p. 20). systems in operation, 25 research-oriented and 12 "There are various means of achieving a time- commercial systems. These vary from small to shared system, and the distinctions among them are large but all are characterized by a central computer sometimes quite fuzzy. Clouding the picture even and main memory that are time-shared among the more is the tendency evident in the literature to users, and an auxiliary storage that is space-shared couple the goal of 'intimate' man-machine com- by the programs of aU the users and the systems munication to one method for achieving time- programs." (Harder, 1968, pp. 235-236). sharing and then implicitly to associate time- Weil (1966) provides general coverage of the his- sharing solely with this one method. For example, tory and probable impact of time-sharing concepts 'time-slicing', or the assignment of particular and systems on the management of data processing segments of time to a given user or problem on a operations. repetitive basis, is a form of time-sharing. 'Dedicated 2.1a See, for example, the bibliography compiled time' modes of operation, where a user is guaranteed by BeU and Pirtle, 1966. a certain percentage of the computer's time in a 2.2 Opler points out that such applications "may given period, could be construed as another form. require considerable queuing of elements and the A number of users have indicated that they feel the development of complex priority assignment and strongest 'interaction' with the computer when the recognition schemes. This too produces new system operates in the dedicated time mode." problems for both language and processor design- (Davis, 1966, p. 224.) ers." (1966, p. 197). "Time-sharing typically means expense-sharing 2.3 A provocative comment with respect to among a large number of subscribers, with reduced flexible system design to meet varying client re- computing costs for many kinds of applications." quirements is that of Greenberger: "The average (Sackman, 1968, p. 1). user should not have to carry water to the faucet 40 in his own buckets, one at a time. The system should "Another use of the reactive typewriter will be provide a continuous supply for him, with the hot, for preparing documents, reports and books. Not cold, and lemonade taps all within easy reach." only will it be of assistance in making editorial (Greenberger, 1965, p. 34.) changes and revision, but, when the text is com- See also Orchard-Hays, who comments: "The pleted, the reactive typewriter can be used to specify final value of an efficient and flexible system is not that it the page make-up for printing, the type styles saves a little machine time. Rather, its value is that and sizes, and location of illustrations. From this specification, the reckoning will it allows us to make mistakes cheaply and thus to machine prepare gain more experience in shorter time and with less a tape record to be used to control one of the new expenditure." (Orchard-Hays, 1965. p. 242.) high-speed photo-composing machines for producing a high quality, typographically perfect final 2.4 In particular, it should be recognized as product. On the other hand, if only an editorial predicted by Bauer that if, as of 1965, "on-line proof copy is desired, and quality of printing is not computing probably represents 1 per cent of the important, a high speed printer at the site of the total computer activity in the country today. It wiU reckoning machine will be used. These can print probably represent 50 per cent in five years." (p. 14). out some ten to twenty pages a minute. Mail or 2.4a "We must realize always that a penalty or a messenger service wiU ordinarily be fast enough to price is extracted for each increment of increased bring the results back to the user." (Mooers, 1965, human efficiency; more computer time, more com- p. 34.) plex computers, greater input/output equipment, 2.6 "The CTSS environment consists of approxi- and increased console costs. The tradeoffs undoubt- mately 250 users whose individual load on the edly favor ever increasing on-line capability. system varies from nearly zero up to the equivalent However, no system design should occur without of several hours of IBM 7094 time per month." recognition of the prices demanded." (Bauer, 1965, (Scherr, 1965, p. 5). p. 18). ". 2.7 . . The possibility of retrieval via an 2.5 "In a cooperative input transcription on-line dialogue. This approach has been under project, the basic transcription into machinable study by Kessler of the Massachusetts Institute form for each report or document will include: of Technology, implemented in a small experimental authors' names, title, all identifying numbers and design by Salton of Harvard,, and is realized in the dates, name of issuing agency (at- all divisional current Lockheed working system." (Drew et al., levels), any other identifying text, and an abstract 1966, p. 3). of the contents of the document. Additional material With respect to the application at Harvard, which may optionally be added is pagination, text Rocchio reports that in order "to establish the value excerpts, descriptors, classification signs, or other of real time interaction in mechanized document index terms added at the point of transcription. . . . retrieval, and to investigate the effects of the various "The transcription — instead of being done at types of feedback which might be provided to the each every retrieval collection over the country- and user, three basic facilities are required: is done once and for all either at the source of the a time-sharing system to effect on-line man- document or at one of a group of cooperating (1) machine communication at a reasonable cost, retrieval systems, with the machinable transcription such as the Compatible Time-Sharing System in the form of cards, or perforated or magnetic (CTSS) available at M.I.T. for the IBM 7094. tapes, being prepared and then circulated to all of a flexible set of document retrieval programs the cooperators." (Mooers, 1959, pp. 20, 21). (2) which are capable of incorporating informa- "This plan offers several advantages to the tion received from the user and of providing retrieval center and cooperating institutions in alternative retrieval operations; addition to the speed of manuscript transmission via (3) a communication-oriented executive routine paper tape. First, this machine-readable record of designed to provide the man-program inter- the bibliography or portions thereof, will be made action and to control the scheduling of the available to cooperating documentation units, and various processing options in a manner the possibility of reciprocal exchange exists. Also compatible with the time-sharing system." the preparation of edited machine-readable biblio- (Rocchio, 1964, pp. XII-1, XII-2). graphic entries is done once, never having to be

repeated . . . 2.8 "In its present configuration, a user may "The first large scale implementation on an in- sit at an electric typewriter, scan a stated range ternational basis of the 'tape-typewriter plan' for of literature and perform a search based on key cooperation between documentation centers, has words, keyword in context, citation index, biblio- been initiated between Italy and the United States. graphic coupling, author, location and various com- This particular application of the plan entails the binations of these. The response is printed back transmittal of entries in an international serial cur- on the same typewriter within seconds of the rent bibliography from the . . . FAO . . . head- request. The interaction between the user and the quarters, Rome to the Aquatic Sciences Information system is free of intermediaries and is accomplished Retrieval Center (ASIRC) at the University of Rhode by means of a language very close to natural ." Island . . (Scient. Info. Notes 5,3, 1 (1963).) English." (Kessler, 1964, p. 1). 41 In addition, provision is made for maintenance of (who) can check this list and determine whether personal files for individual users: any new topic tags, equivalence words, or little "Instead of specifying what to find the program words need to be created. The topic tags in the is directed to read a file called SMITH. . . . [which] Document Word File are ordered alphabetically, may contain a list of authors, citations, and/or key duphcate topic tags are deleted, one is added to the words that Mr. Smith is particularly interested document frequency tally in the Word Screen File in . . . He may change it from time to time . . . associated with each topic tag . . ., the request This provides the user with a combined author frequency tally and the document frequency tally index, citation index, KWIC and share search are copied from the Word Screen File and entered tailored to his individual needs." (Kessler, 1964, along with each topic tag in the Document Word p. 21). File . . . "Night Letter. In this program a stated literature "The Document Word File now consists of the range is to be searched. When a reference is found equivalence words and the topic tags in the docu- to a prepared list of papers, a letter is printed, ment, along with the request frequency and docu- telling addressed to the author of the cited paper ment frequency tallies . . . (and, the) assigned him the journal, volume, page, title, author, and document number. This version of the Document location of the paper that cited him." (Kessler, 1964, Word File is used to update the Topic Tag Index p. 19). File and also to create the automatic abstract of the ." "A set of programs has been developed under the document . . (Geddes et al., 1963, pp. 179-180). general name of SHARE. In this process, we name 2.10 Borko gives the following account: "Now an article and ask that other articles be found that you see the Usted subcategories, and again you use share some element with it. One may ask for author, the light pen to make your selection. In this manner word, location, or citation share." Further, "it is you narrow your field and reduce the number of possible to create 'standing orders' that will be documents to be searched. You may then ask to fulfilled at some slack time when the computer is see the documents in the subcategory that you have free from immediate demands, usually late at night." selected. On the scope you will see the author and (p. 17) title of the books which have been classified into 2.9 "TEXTIR (TEXT /ndexing and /?etrieval) is that category. You may now browse through the file a set of programs which were developed as part of a and make your selection. On request, you will be joint experimental project undertaken by the Los given the appropriate call number and you may ask Angeles Police Department (LAPD) and System for a microfilm copy. The microfilm image could be

Development Corporation (SDC) in the computer transmitted to you on the display console or . . . retrieval of crime data in natural English lan- mailed to you."

guage . . . The basic capabilities of TEXTIR Some further details of the SDC system are as include automatic indexing of every substantive follows: word in machine-readable reports or documents, "The BOLD data base generator builds tables of and retrieving preselected portions of these indexed structured information from a Hollerith prestored reports in response to natural English language magnetic tape. The tables are designed with queries." (FarreU, 1965, p. 7). This system based extensive linking between entries referenced by upon the AN/FSQ—32 computer, conventional identical key words or phrases to permit rapid Model 33 teletypes, CRT display consoles, and light retrieval. The data base that is presently being used pen input facilities is also discussed by Schwartz was obtained from the Defense Documentation et al. (1965). It is in operation nine hours a day for Center and consists of abstracts of approximately more than 30 users simultaneously. (Commun. ." 6000 documents. . . ACM 8, 253 (April 1965).) "The function of the data base generator sub- "The entire document ... is read into the com- system is to process prestored data tapes and to puter . . . All the words in the document are establish the linking between the index terms and ordered alphabetically and duplications are elimi- other descriptive attributes such as authors, titles, nated to produce a Document Word File . . . The contract numbers, etc. The data tables produced by Document Word File is compared against the Word the data base generator are used by the retrieval Screen File; all the little words and obsolete words program. ..." are deleted ... all the topic tags are retained and "The display and retrieval subsystem is designed the relative address of each topic tag is entered . . . to search the data base that has been structured by all the equivalence words are retained along with the data base generator program and to retrieve and the associated topic tag and its relative address . . . display the requested information. . . . "At this point in the processing, the Document Word File consists of topic tags, their associated "The retrieval program has the following char- relative addresses in the Word Screen File, equiv- acterostocs: alence words, and the words of the document which (a) Allows, and therefore monitors, many users are not in the Word Screen File. This latter set of who are interrogating the same data base but unidentified words is set aside and pointed out as who are doing simultaneous, independent,

an output to the personnel of the central facility . . . real-time retrieval.

42 (b) Permits the user to interrogate the dictionary offline printer, and a 2260 display/1053 printer and find out whether a word has been used input/output terminal. The reference file consists as an index term — and how often — and of some 300,000 NASA announced citations." whether there are synonyms or related terms. (Summit. 1967. p. 52.) (c) Allows the user to formulate his requests in 2.14 "In another application, called Project natural language with very few computer- CIDS — for chemical information data system — the oriented restrictions. computer will hst all chemical compounds that fit (d) Displays retrieved information on a teletype or the specifications in a user's inquiry, giving formulas, CRT console. physical and chemical properties, and all known (e) Copies the retrieved information on a magnetic chemical and commercial names. If the inquiry is ." tape for off-line listing. . . too general, the satellite computer will indicate this

"The BOLD inquiry station consists of a tele- before the search is started . . . typewriter and cathode ray-tube display unit. The "An information-retrieval system called Project user interacts with the system and requests infor- Vector is being developed, with which an unin- mation by typing on the teletypewriter. He may also formed person can use an inexpensive terminal to make certain requests by using the light pen and gain access to a vast hbrary of information that the display scope . . . would otherwise not be available to him . . . ". . . [A] display appears that indicates the 32 "Other applications include research in the divisions or main subject categories into which the locating of legal precedents and experiments in data are divided. If the user wishes a further man-machine interaction in game-playing." (Carr breakdown, he may use his light pen to flash a and Prywes, 1965, p. 89). division. By doing so, he is requesting more informa- 2.15 "Under a [NASA] contract with the Bunker- tion about that category and receives a display of Ramo Corporation, consoles wiU be installed in the subdivisions. The display also suppHes the users' immediate working areas, hooked-up by number of entries in the category. If he chooses to telephone lines with a central computer-search browse through the items in this category, he may system . . . The user 'types' topical terms, authors' do so . . . The display he would receive is the first names, or similar identifying data which he believes abstract in that category. are pertinent to his query. Within seconds, corres- '"Should the user decide to keep a permanent copy ponding citations begin to appear on the viewing

of the display for future reference . . . the informa- screen ... If the user hkes what he sees, the tion is transferred from the scope to the teletype." pressing of other buttons wiU record his interests (Borko and Barnaugh. 1966, pp. 6-7, 8-10. 12, 21). so that the cited documents may be sent to him. If 2.11 "'The system which is now working at not, he can modify his search until relevant citations LMSC depends upon two other operating Lockheed do appear." (Sci. Info. Notes 8, 12, 13 (1966).) systems. The first, MATICO (Machine-Aided "One of the concerns of Project Intrex is to con- Technical Information Center Operations) supplies duct a series of experiments to determine how the the machine-readable data base. The second, ADA traditional library catalog can be effectively aug- (Automatic Data Acquisition), supplies the on-line mented and combined with on-Une computer

computer and its time-sharing monitor . . . operation to provide users with a more powerful, "The types of information used as descriptors are: comprehensive, and useful guide to library re- (a) personal author, (b) all significant title words, sources. Present plans call for augmenting the (c) corporate author, (d) all subject headings (an traditional catalog in scope, depth, and search average of three per report), (e) contract number means. For example, the augmented catalog will con- (f) original report number, (g) secondary report tain entries for reports and individual journal articles number, and (h) date of publication." (Drew et al., as well as the traditional entries for books. Further- 1966, p. 4). more, in addition to the title and author of each item, 2.12 "It is possible to further restrict a retrieved such things as the bibhography, an abstract, key set of citations by year of announcement, an- words, and key phrases of each item will be included nouncement media ... or announcement series as part of its catalog entry. A user will be able to

number. . . . conduct searches on nearly any combination of the "By displaying the alphabetically near terms, the data contained in a catalog entry. Present plans also user is able to see not only if and to what extent call for providing alphanumeric communication the term he entered has been used as an index between user and computer by means of a high- term, but also any spelling or ending variations on speed, flexible display console." (Haring, 1968, the term which have been used." (Summit, 1967, p. 35). pp. 53-54). 2.16 "SAFARI is an on-hne system for text 2.13 "The current operating environment of the processing designed to be used by a group of in- DIALOG system is an IBM 360/30 (32 thousand formation analysts for building their own personal bytes of core) together with two 2311 disk packs files. The system allows an analyst to scan through (7.5 million bytes each) for programs and inter- documents displayed on a CRT and to select mediate storage, a 2321 Data Cell (415 miUion bytes) relevent facts which then are processed hnguis- mass storage device for the reference corpus, a 1443 ticaUy and stored in the computer in the form of

43 366-108 O - 70 - 4 logical content representations. To locate informa- systems. He covers such relatively early develop- tion, the analyst can initiate a search through the ments as Strachey's 1959 paper, the M.I. T. -Lincoln stored representations to identify those with relevant Laboratory-SAGE developments, the Speed Tally content. When an item of interest is found, the system, airlines reservation systems, stock exchange original text from which it was derived can be re- quotation services, the CTSS (Compatible Time covered. On-line editing capabilities allow the Sharing System) at M.I.T., the Bolt, Beranek and analyst to take all or any parts of the recovered Newman system, and others. We also note the texts, add commentary and interpretation, and following: generate reports about their content." (Walker, "The most complex system to date is the INTIPS 1967, p. 144). system at the United States Air Force Rome Air 2.16a "It is felt by many that an on-line computer Development Center. It combines multiprogram- system, which allows a user to converse directly ming, multicomputing, remote personal consoles, with the computer in his quest for relevant docu- remote graphical input-output devices, and remote ment citations, can provide a more effective en- conventional I/O devices. It is a lineal descendant vironment for information retrieval than is possible of the original RW 400 Polymorphic Computer with off-line systems. The on-line systems permits developed by the Ramo-Woodridge Corporation." information retrieval to be a highly individualized (Wagner and Granholm, 1965, p. 286). ". process with respect to time of occurrence, question 2.20 . . An algorithm is used to determine at hand, and characteristics of user." (Summit, dynamically how much of the system's resources a 1967, p. 51). task ought to be allowed to consume. If this threshold 2.16b "We believe that this approach to data level is exceeded, the task is forced to stop and wait analyses — a computer-mediated interplay between while other tasks have a chance; in other words, the investigator and his data — holds promise of a forced multiprogramming of sorts. (This has been more effective inductive analysis than either man or generally termed as time-slicing).'' (Comfort, 1965, algorithm could produce alone." (Press and Rogers, p. 620.) 1967, p. 39). "A time-sharing system slices time in such a way "One of the more promising areas of computer use that each user gets a small amount of attention on involves the coupling of a man to a computer system some periodic basis." (Ware, 1967, p. 288.) for real-time problem solving where the procedure for 2.21 "The scheduling algorithm is that set of solution of the problem is either unknown or involves rules, or more strictly its embodiment in a program, complex tasks such as pattern recognition, that can which govern the frequency with which each user best be performed by humans." (Brown, 1965, p. 82) is serviced and the length of time each program is

"Use of the graphical input-output console allows permitted to run before it is interrupted . . . The the designer to pause and study his design at all correct form for a scheduling algorithm is one of stages in its process without unnecessarily tying the most hotly discussed aspects of a time-sharing up the computer. Furthermore, he can make modi- system, and a great deal of experimental work will fications at any time and be appraised within still have to be done before the question can be seconds of the effects of each change." (Hamilton settled." (Samuel, 1965, pp. 12-13). and Weiss, 1965, p. 3). "Real-time programs may require considerable 2.17 "It seems vitally important to press on queuing of elements and the development of com- * with the development of multiple-console computer plex priority assignment and recognition schemes. ' systems, particularly in organizations in which This too produces new problems for both language ' creative potential users abound." (Licklider, 1965, and processor designers." (Opler, 1966, p. 197). ^ pp, 68-69.) One obvious implication is that: "Responsive J 2.18 "Another area in which responsive data time-sharing . . . needs more collaboration between processing offers recognized economy and con- software and hardware designers than has been | venience to the scientist or engineer is ease of learn- evinced thus far." (Adams, 1965, p. 488). 'i ing, ease of use, and ease of making corrections . . . 2.21c "Some attention must be paid ... to the ^ A properly designed system can function as a methods by which the policy decision, once arrived ' teaching machine, indicating mistakes or mis- at, is to be enforced. It is desirable that these 5 understandings as they occur." (Adams, 1965, p. methods should allow as much flexibility as possible fi 484.) in the choice of quantum and should not add signifi- i ". 2.19 . . The cooperation of man and machine cantly to the cost of scheduling." (Lampson, 1968, i bilateral takes on the character of a dynamic process p. 355). I of mutual training whereby the man, having at first See also the following: adapted himself to the confined language of the "The Manager program provides control for a gen- » machine, then step by step raises it to his level, eral-purpose time-sharing system to make its hard- | partly by concrete clarifications and explanations ware and software usage more efficient in servicing i of general rules, but equally by the confirmation of many users. In substituting computer operator man- t generalizations successfully made by the machine agement control devices for the built-in algorithms itself." (Yershov, 1965, p. 323.) other systems, the program bridges the gap be- of [f

2.19a Glauthier (1967) provides a review of the tween on-line user demands and programmed system \ relatively brief history of computer time-sharing user control, and provides a human element in 44 system management. As time-sharing systems of service. . . . Reservation mechanisms typically expand and improve, new demands on such systems arise where a facility (or right) cannot be shared can be expected. The incorporation of more pro- (pre-emptive rights cannot) and where the facility grammed solutions into time-sharing systems users wish to plan ahead with a high degree of might seem to be the best approach to allocating the certainty that their plans will materialize." (Brown system resources, but the cost of this approach may et al, 1967, pp. 213-214). be very high in terms of maintenance and flexibihty. 2.23a "Although there are many systems today However, on-hne computer operator management that provide time-shared access to a computer, little tools cannot be expected to augment software needs is known of precisely how such machines are used. for long, as systems with hundreds of users begin to This was especially apparent at the beginning of pose more severe demands on this human element. the JOSS development. Substantial effort was Eventually, all large time-sharing systems will therefore made to provide a measuring or instru- include elaborate management programs, with a menting capabiUty within the system not only to clerical staff that manages the systems. Because record use of the system as a whole but also to the human element will continue for a long time to record characteristics of use for individual users be a part of time-sharing system management, more of the system." (Bryan, 1967, p. 769). elaborate display devices will eventually be re- "The JOHNNIAC Open-Shop System (JOSS) is quired, to assist the operator, and to give him as an experimental, on-line, time-shared computing much timely information as possible. On the other system which has been in daily use by staff members hand, the need for users to be aware of certain of The RAND Corporation since January 1964. It system resources should be less of a requirement. was designed to give the individual scientist or When procedural problems do occur, however, engineer an easy, direct way of solving his small they should be able to communicate by voice from numerical problems without a large investment in their teletype stations with personnel who can give learning to use an operating system, a compiler, assistance to them. The Manager program is merely and debugging tools, or in explaining his problems a beginning in the control and management of time- to a professional computer programmer and in sharing system use." (Linde and Chaney, 1966, checking the latter's results. The ease and directness pp. 157-158). of JOSS is attributable to an interpretative routine ". 2.22 . . We may expect future appUcation in the JOHNNIAC computer which responds programs to be built with a mode control permitting quickly to instructions expressed in a simple the user to operate the program in either a batch language and transmitted over telephone lines from or conversational mode." (O'Sullivan, 1967, p. 172). convenient remote electric-typewriter consoles. "The time-sharing system at the University of An evaluation of the system has shown that in spite California at Santa Barbara uses an interesting of severe constraints on speed and size of program, variation of the above methods. A user is required and the use of an aging machine of the vacuum- to state whether his job is interactive (short, fre- tube era, JOSS provides a valuable service for quent requests) or batch (longer, usually single computational needs which cannot be satisfied by requests). Time is divided into fixed length segments conventional, closed-shop practice." (Shaw, 1964, such that during the first half of each segment, the p. 455). interactive jobs are served in a round robin fashion "JOSS* is a special-purpose computing system untU their half-segment is exhausted (or until their designed to provide users with a substantial and collective requests are satisfied). The remainder of highly interactive computational capability. The the segment is then used to service (to completion, if first JOSS system, developed for the JOHNNIAC possible) as many of the batch jobs as possible. computer machine by J. C. Shaw, was operational During the first half-segment, some interactive jobs in early 1963. Work on an expanded system utihzing wrill drop from the queue after one quantum of a modern PDP-6 computer began in 1964, and the service (e.g., those requiring the acceptance of a system became operational in February 1966." single button-push), etc; thus in this system, there (Bryan, 1967, p. 769). is a benefit in declaring the nature of your job in an 2.24 Examples in current systems include the honest way, since in the case where there is only following: "Priorities . . . can be assigned on a one batch job and many interactive jobs, the batch dynamic basis depending on how long the program job receives better service if that user declares has been in the system, or how much time remains himself as being in the batch mode." (Coffman and before the deadline for results." (Bauer, 1965, p. 23.) Kleinrock, 1968, p. 16). "In defining the scheduling algorithm one must "Time-shared systems are tending to find it con- specify: venient to run short jobs to completion and to inter- (a) the sequence in which active users are to leave stacked production jobs into long pauses in receive service and the variation in this se- online operations as background jobs." (Sackman, quence as load conditions change. 1968, p. 2). 2.23 "By whatever means, it seems clear that

priorities and reservation systems will be required for its computer •JOSS is the trademark and service marlt of the RAND Corporation to assure certain classes of users appropriate quality program and services using that program.

45 (b) the amount of time (under all circumstances) "Each program on the library is to have one or

each user or type of user is to receive." more queuing codes incorporated within it to control (Schwartz et al, 1965, p. 19.) the minimum service level of that program during "Scheduling within SHARER is done on a round- busy running periods. The resulting priority chain robin basis, with the central processor being allo- will thus be a combination of first-come, first-served cated to each program in central memory at least modified by minimum levels of serviceability asso- once every half second." (Harrison and Schwartz, ciated with each program." (Baruch, 1967, p. 148). 1967, p. 662). "Some experience is now available about the "The management of the queue would be accord- behaviour of scheduHng algorithms. It is found ing to the following principles: that the simplest algorithm — a straightforward 1. Processes are entered into the queue by pro- round-robin in which each object program when gram forks and the completion of ilo functions. loaded receives the same quantum of time — is not 2. A process is taken from the queue and placed satisfactory, and that it is necessary to cater separ- in execution whenever a processing unit ately for small brief jobs and for longer jobs. This becomes free. can be done by having a separate queue on to which

3. A processing unit becomes free when it en- jobs are transferred when it appears that they are counters a quit or i/o function procedure taking rather a long time on the regular queue; step, or the process has exceeded its allotted once on the second queue, they receive larger quanta time." (Dennis and Glaser, 1965, p. 9.) of time when they are activated, but have to wait longer between activations. Corbato imple- "If a user remains in the same queue for more mented in CTSS a more elaborate system in which than 60 seconds without being run, he is moved to jobs which do not terminate, or which reach an the end of the next higher priority queue." (Scherr, input or output wait, during a period of activation 1965, pp. 119-120.) receive double the quantum of time during the "A user may be allotted time in any or all shifts. next period, but have to wait twice as long for it. If he consumes his allotment on a given shift, the The quantum of time received when a program is system automatically logs him out, and he must first activated, or when it is reactivated after a wait, wait until the beginning of a shift in which he has is determined according to the amount of core space resources remaining before he is permitted to log in occupied, larger jobs receiving a larger quantum." again." (Mills and Van Vleck, 1966, p. 5.) (Wilkes, 1967, p. 2). "Priorities are assigned user's programs on the "The scheduling algorithm operates roughly as basis of length and previous running time. In order follows: Each user request is assigned an initial to keep efficiency up, longer programs are given priority which depends only on the size of the slightly less priority because they require longer program that must be run. The highest priority is swapping time." (Scherr, 1965, p. 119.) assigned to the smallest programs. The highest "A user is pre-empted, that is, another user will priority programs are allowed to run for a maximum be swapped in, if the current user is no longer the of 4 seconds before being interrupted, whereas lower first user in the queues, and he has run as long as priority to ." programs are allowed run for longer the new first user will run . . (Scherr, 1965, p. intervals which are multiples of 4 seconds. The 120.) lower the priority, the longer is the aUowed interval. "A variation on the first-come, first-served system If a program run is not completed within the allowed is called round-robin priority. When a program's interval, the program is transferred from core time is up, the computer samples all incoming lines memory to drum (the state of the machine being in a prescribed order until it finds a request to automatically preserved), and its priority is reduced run . . . by 1 unit." (Fano, 1964, p. 16.) "A more complex priority system has been pro- "Jobs for the slowest peripherals are given priority posed whereby a computer could whittle away at until a reasonable amount of output for the device large problems in its 'spare time', when it is not busy is waiting in the huffier, the with simpler, high-priority services. Several different output then scheduler on to the next slowest peripheral. have levels of service could be offered, at different speeds moves We found that the optimum choice for this 'reasonable and prices." (Riley, 1965, p. 74.) "The scheduling algorithm currently operational amount of output' keeps each peripheral in maxi- in the Q— 32 system utilizes three priority queues, mum use if the demand exists." (Morris et al., each circHng in simple round-robin fashion, with 1967, p. 69.) queue position dependent upon the amount of CPU "In ATLAS, aU jobs are fed into the system as time required for a request." (Schwartz and Weiss- soon as they appear. A scheduling program selects man, 1967, p. 265.) those jobs having input/output characteristics which At Case Institute of Technology, "a revised selec- will tend to put as many of the computer components tion algorithm is under development which will into motion as is possible." (CoUila, 1966, p. 49). select jobs according to a complicated nonlinear "A queuing structure more sophisticated than function of both estimated running time and time the simple round robin approach should be imple- spent waiting in the input queue." (Lynch, 1966, mented. Distinction among the factors causing tasks p. 122.) to give up control of the processor provides a reason- 46 able basis for the queue organization. Servicing of 2.28 "The response time of the system to a queues to put interactive tasks back in control of line of input from the user is an important parameter the processor as soon as possible distinctly improves of a time-shared system. In fact, it is one of the response times at no cost to processor utilization." few, well-defined, measurable performance param- (Oppenheimer and Weizer, 1968, p. 320). eters available. This response time determines the "The design of the priority and quantum assign- basic rate at which the user can operate." (Scherr, ment algorithms must take the upleasant possibility 1965, p. 19.) [of a ready process never being run] into account. "The apparent response time is a subjective This might be done by restricting the frequency with evaluation of system performance made by a console which a process may enter the ready Ust with high operator in a man-machine system. Extensive priority. Another alternative is to restrict the length human-factors tests are required to ensure that of time it may run at high priority, although this one an acceptable real response time is, in fact, an may be subverted by the fixed overhead imposed acceptable apparent response time." (Aron, 1967, by the need to swap in the memory for the process. p. 53.) ". StiU a third approach is to increase the priority of . . The ratio of response time to processor processes which have been waiting for a long time time per interaction will be substituted for response at low levels." (Lampson, 1968, p. 355). time. It turns out that this ratio is a more stable "Second priority is given to users who have given measure of CTSS performance. The variation of JOSS output-limited tasks which have been set this ratio as a function of the average number of aside until the typewriters have nearly caught up. interacting users is a measure of how well the Third priority is given to users with unfinished tasks, system responds to a change in its load." (Scherr, on which JOSS works for two seconds apiece in 1965, p. 56.) round-robin fashion. A user's priority changes "The most important performance aspect of dynamically according to this discipline, which any man-machine system is the quantity of results successfully exploits the parallel processing of the obtained (in terms of 'significant' problems solved) communication system. Under a typical load, per unit time. At this time, such a quantity is JOSS responds to simple requests in a fraction of a unmeasurable." (Scherr, 1965, p. 96.) second and rarely in as long as three seconds. "While it is recognized that response time and Users who are skilled in typing can maintain im- costs are not sufficient conditions for the evalua- pressive rates of interaction with JOSS." (Shaw, tion of an information system, they are necessary 1964, p. 459). conditions." (Blunt, 1965, p. 10.) 2.25 "When service is effected according to an 2.29 "Even a very small inter-job overhead may a priori scheduHng algorithm, the communication be a tremendous fraction of the total system network is used to gather user requests arising at capacity. Even a substantial overhead during unscheduled times and to distribute results after the job, e.g., multiprogramming can result in a servicing ordered queues of requests. . . . When gain in total system capacity." (Perry, 1965, p. 246.) service is effected according to a dynamic scheduling "If mean response time is of primary interest, algorithm, processing time quanta, which may be simple mathematical models can be constructed variable, are assigned to each request." (Estrin and having good generality and yielding accurate Kleinrock, 1967, p. 85.) resuhs." (Scherr, 1965, p. 109.) 2.26 "To exploit the increase of capacity gained "Response time is a function of equipment speed; by distributing, the system must provide for efficiency of the man-machine interface, operating queuing. The queue depth permissible should be programs and procedures and the communications at least as great as the number of components over capability within the system." (Blunt, 1965, p. 6.) which the function is distributed." (Amdahl, 1965, 2.30 "When two or more data units are waiting p. 38.) at the same event for servicing, it is often possible "An unconditional guarantee of access (given the to expedite processing by selecting a data unit facility is operational) can only be provided by a under some other criterion than first-come-first- priority system if preemptive priorities are issued serve. For example, selection of the data unit with to no more users than can be served simultaneously. the least expected servicing time may both reduce Otherwise, the guarantee wiU fail on occurrence of overall response time and equipment/personnel idle a peak load." (Brown et al., 1967, p. 214). time . . . Another criterion of data unit selection 2.27 It is noted that, in typical scheduling sys- is to process the data units in a sequence such as tems, client-processing use is of the order of two to minimize computer idle time or to maximize utility to four seconds but that during this interval several of the central processor." (Blunt, 1965, p. 64.) hundred thousand processing instructions may "While different queue unloading strategies can be executed; that the client's own response-time produce different effects in processing, not all of requirements (either to receive a textual message these strategies are likely to be easily applied in and/or display or to type-in a response) generally the real world. Casual comparison of two data units require at least three seconds, and that allowing may not effectively reveal which element will proc- a quantum time of ten seconds is almost equivalent ess faster in a given event, or which processing to allowing any client's problem to run to comple- sequence will better contribute to effective utilization. (See Scherr, 1965, p. 11, 68.) tion of the central processor." (Blunt, 1965, p. 64.) 47 2.31 "The effects of users getting 'into phase' "Another conception of time-sharing is that it with each other . . . This phenomenon occurs when is akin to a community utiHty capable of providing many users require service at the same time and is computer power to each customer. This notion was self-perpetuating. That is, as more and more users introduced in a lecture in 1961 by Professor John begin waiting, service for the individual user is McCarthy. Very little concerning this concept reduced, and there is time for more users to reach appeared in the 1965 Hterature; an exception the point where they require service. Thus, users being Fano's paper, which describes the MAC tend to fall into step with each other; many working system as 'an experimental computer utility which, at the same time, many thinking at the same time." since November 1963, has been serving a small but (Scherr, 1965, p. 25.) varied segment of the MIT community.' Another 2.31a "Multi-programming adds a good deal to exception is the paper of Dennis & Glaser, which the complexity of the design of a scheduling outlines some concepts of computer system organization algorithm. When the running program is halted, it that have arisen from the evolution of an is no longer a simple question of what program is information processing utihty as envisioned by ." to be loaded to take its place. There is first a deci- Project MAC . . (Davis, 1966, p. 225). sion to be taken as to whether a new program should "The use of a separate processor for all inter- be loaded into core at all, and if so which program active data I/O for TSS has given SDC the abihty to interface a large is to be dumped in order to make room for it. If no variety of devices with TSS. These include teletypes, new program is to be loaded it must be decided special keyboards and type- which program already in core shall be activated. writers, hghtpens for displays, special pushbuttons, high-speed data Hues for computer-to- The taking of these various decisions is compli- computer communications, cated by the fact that programs vary very much in automatic dial-up units, as well as the graphic tablet display console. size, and that it is of prime importance to make the To provide quick response time to the user utilization of available memory space as nearly of the GTD console, the PDP-1 performs a significant optimum as possible." (Wilkes, 1967, p. 2). amount of preprocessing on the raw data. A project "Because short response time is vital for an effec- is currently underway to study the trade-offs of tive multi-console system, an effort must be made allowing the slower PDP-1 or the faster Q-32 to to predict for any system design how queueing will perform the required user functions. Although the increase the obtainable response time. One way to user cannot directly modify the PDP-1 executive do this is to assume the 'worst case' situation, in interactively, a system programmer can modify it which a command encounters maximum possible to meet new user requirements." (Gallenson, 1967, queueing delays in being processed, and then de- p. 690). sign for an acceptable worst case response time. 2.33 "Computer-directed procedures are an This approach is satisfactory perhaps while such important aspect of on-hne computing and are little systems are largely experimental. But the obvious understood or appreciated. For example, they drawback in relying solely on 'worst case' predic- allow a complex query to be asked of a machine tions is that the system is over-designed, and the under the control of another machine and with the true capacity of the system is underestimated. This assistance of a third machine.',' (Bauer, 1965, p. 18). can be serious economically when people seek more "Remote batch processing involves a computer at widespread access to use of these systems. So a the user terminal, rather than simply a typewriter ? study of response time which takes into account alone. This low-cost computer is used to prepare a its statistical aspects is desirable both for a pre- local batch of programs for direct transmission to = true capacity, and to point out diction of system the large-scale central computer." (Aron, 1967. which system components limit the capability so p. 63). ! can taken if needed." (Fife that remedial action be O'SuUivan (1967) describes an interface computer | and Rosenberg, 1964, p. Hl-1). used in support of terminal network, to provide ;

2.32 "The first large system to be used like a code translations, bit formatting, error checking 5 public utihty is the IBM Quiktran System. It shares procedures. its consoles not ". among individuals within the same 2.34 . . Special processors should be de- 't organization, but on a service bureau basis to anyone signed to allocate and relocate users' areas in i in the geographical area." (Wagner and Granholm, memories possibly in parallel with the main com- f 1965, p. 286). putation." (Lock, 1965, p. 471). i

"An information processing utility consists of "The foregoing operations are usually delegated. [ processing units and directly addressable main That is, provision is usually made for a control memory in which procedure information is inter- unit, not a part of the central processor, which « preted as sequences of operations on data, a system has the function of finding the desired item or

of terminal devices through which users may com- location in storage, and of monitoring reading I'

municate with procedures operating for them, and or writing. Because of this, the program is not ( mass memory where information is held when not held up by resort to the large storage, at least to required for immediate reference." (Dennis and the extent that the data is not forthwith." needed , Glaser, 1965, p. 5). (Dumey, 1965, p. 257). \

48 "Future time-sharing supervisors will run on status and disk file status of the outgoing and in- computer systems having many processors, and coming programs. This transmission consists of many active memory modules, flexibly inter- approximately 500 words in each direction." (Scherr, connected . . . They will provide for the dynamic 1965, p. 36). ." allocation of space . . (Greenberger, 1965, "It is now becoming possible to overlap the execu- p. 36). tion of one program with the swapping of two or 2.34a For example: "Cooperative communi- more programs so that little or no time is lost." cation programs in the control and central com- (Samuel, 1965, p. 9). puters supervise any interchanges between the 2.37 "With each computation there is associated console and the central facility." (Ninke, 1965, a set of information to which it requires a high p. 842). density (in time) of eff^ective reference. The member- It is to be noted further that "these [operating] ship of this working set of information varies dy- systems encompass all the language processors namically during the course of the computation." and commonly used utiUty routines in addition (Dennis and Van Horn, 1965, p. 11). to controlling aU components of the equipment, "A variable sized data is — structure handled by servicing all traps and interrupts and accounting making it a distinct segment and allocating more or in general running the show . . . Such a system fewer blocks to it as the structure changes in size." is designed to bring to bear a number of powerful (Dennis and Glaser, 1965, p. 8). and standard programs on relatively simple com- "Two eff"ects occur: first, each operating burst is puting tasks, and to keep them moving through lengthened by the amount of additional program the queue". (Orchard-Hays, 1965, p. 238). execution time required to select and reactivate the

2.35 "Today's fastest machine cannot be loaded program and to preserve it upon termination. down and will be idle most of the time unless it Secondly, the idUng delays are lengthened by the is coupled to a large number of high speed channels amount necessary to complete the execution of one and peripheral units ... In order to distribute of the other time-shared programs." (Brown, 1965, input-output requests in a balanced flow, it must p. 82). also be controUed by a complex monitor that chooses 2.38 "Reduction in swap time would result if wisely among the jobs in its job queue." (Clip- only those parts of an object program and its environ- pinger, 1965, p. 207). ment actually involved in the current operation were "Dynamic memory allocation is "a necessary loaded into main memory. Thus, many large pro- requirement for an engineering computer system grams could be modularized operationally to im- capable of solving diff^erent problems with dif- prove their performance under time-sharing." ferent data size requirements. . . . (Schwartz et al, 1965, p. 19). "Dynamic memory allocation does involve 2.38a "The purpose of paging is to relieve the additional machine cycles, but it also eliminates programmer of the need to manage physical store. many bookkeeping requirements and the packing To the programmer a paged store is indistinguish- of information that are necessary without its use." able from an unpaged store of the same size.''' (Laski, (Roos, 1965, p. 426, 432). 1968, p. 35, underlining supplied). "Memory mapping and dynamic storage reloca- 2.38 "The choice of page size is clearly critical. tion under control of the supervisor will make it Both the decreasing cost of core and the importance possible to call library subroutines by name and of minimizing supervisor overheads point to a fairly to eff^ect linkages to subroutines at the time they large page. At the present time, I would regard 4K are called." (Licklider, 1965, p. 182). as the minimum to be considered and would, per- "Storage allocation strategies must be fuUy haps, choose 8K. I am aware that smaller pages integrated with the overall strategies for allocating have often been advocated. The situation is com- and scheduling the use of computer system re- plicated by the fact that the supervisor requires to sources." (Randell and Kuehner, 1968, p. 303). use small blocks of memory (say 64 words) for input Further, and output buffering, and there is a case for making

"The choice of a suitable storage allocation some special hardware provision to enable it to do system is strongly dependent on the characteristics this efficiently." (Wilkes, 1967, p. 5). of the various storage levels, and their intercon- 2.38c "One of the most important problems in nections, provided by the computer system on which the construction of a dynamic storage allocation it is implemented." (Randell and Kuehner, 1968, scheme is that of recovering storage space which is p. 303). no longer being used by the program. The mechanics 2.36 "At the point in time when one user's of detecting the existence of such 'garbage' have program is stopped and another's resumed, the been adequately described (Newell et al., 1964; status of the former must be saved and that of McCarthy, 1960; Collins, 1960; Schorr and Waite, the latter restored. This process is called 'swap- 1965). In the case where the elements of storage ping'." (Scherr, 1965, p. 1). required by the program are of variable size (Com-

"In addition to the transmission of the actual fort, 1964; Ross, 1961; Knowlton, 1965) it may programs, each swap is accompanied by the dump- happen that an element of a certain size is needed ing and loading (with the drum) of the processor and, although sufficient free space is available, no

49 free element which is large enough can be found. established. The selection of the page to be trans- For example, suppose that 20,000 words of memory ferred to the drum is made with respect to this were allotted as a storage area to contain one- and pattern of use." (Morris et al., 1967, p. 69).

two-word elements. If the first, third, . . ., 19999th "In several recent computer systems, mecha- words were all occupied by one-word elements, it nisms (usually part-hardware, part-software) would not be possible to obtain a two-word element have been provided to give name contiguity with- even though 10,000 words were free. What should be out the necessity for a complete address contiguity. done in this case is to 'compact' the store by moving This is done by providing a mapping function all one-word elements to one end, leaving a single in the path between the specification of a name by 10,000 word block of free space." (Haddon and a program and the accession by absolute address Waite, 1967, p. 162.) of the corresponding location. The mapping is "A 'free storage list' of linked cells is the basis of usually based on the use of a group of the most most dynamic storage allocation schemes. Cells significant bits of the name. A set of separate are generally removed from this list as they are blocks of locations, whose absolute addresses

needed by the processes of the language . . . are contiguous, can then be made to correspond

Cells may be returned under program control, to a single set of contiguous names . . . The automatically as they are abandoned, or periodically first example of such a system was the Ferranti by 'garbage collection'." (Raphael, 1966, p. 70). ATLAS computer." (Randell and Kuehner, 1968, 2.38d "Even if an accurate ordering of core pp. 299-300). pages based upon frequency and recency of use "This was also the first use of demand paging could be inexpensively maintained, the value of as a fetch strategy, storage being allocated in units that ordering for determining what core pages to of 512 words. The replacement strategy, which free is questionable. The problem with the ordering is used to ensure that one page frame is kept is that a page chosen to be freed may be assigned vacant, ready for the next page demand, is based to a task which has been between time slices (and on a 'learning program'. The learning program therefore not referencing the page), but is just begin- makes use of information which records the length

ning a new time slice in which it will attempt to of time since the page in each page frame has reference the page. One would like to order the core been accessed and the previous duration of in- pages based upon the time at which they will next activity for that page. It attempts to find a page be referenced and release those with the longest which appears to be no longer in use. If all the projected inactive time (PIT) first. Although the pages are in current use it tries to choose the one system does not have such a projection on each core which, if the recent pattern of use is maintained, page, there is information in the system concerning will be the last to be required." (Randell and the PIT of each task. Since a non-sharable virtual Kuehner, 1968, p. 303). memory page belonging to a task can be referenced 2.40 Thus, for example, in the design of the only when the task is in a time slice, a chain of IBM 360/67, "there are four main characteristics tasks ordered with respect to the PIT of each task of a system that has dynamic address translation. can serve as a partial ordering of the core pages The total program need not be in memory when which are allocated to non-sharable virtual memory any part of it is being executed; those parts that pages. Then when a request is made to release core are in memory may be in noncontiguous, non- pages to the system, core assigned to the task at sequential pages; the parts may be left in memory the top of this core release chain can be freed. after the quantum of time is used up, unless it We now consider the problem of defining a task is necessary to swap pages of another program ordering to serve as a basis for such a core release into their places; and if all this happens, when the procedure." (Fikes et al., 1968, p. 13). original program is restored for another quantum ". it 2.39 . . The importance of some of the well- may go into a different area of memory the }*

known European contributions to the computer second time around." (Iverson and Yee, 1965, v arts, for example, the page-turning procedures p. 81). proposed with the original design of the Atlas, or "Division into pages is performed with both '

the multi-processing algorithm included in the hardware and software. The time-sharing super- ' ." Gamma-60 . . (Sahon, 1967, p. 203). visory program sets up software tables in memory, "In Atlas the instruction format permits the containing entries that define the starting points <

programmer to address directly 2048 blocks each of the pages for every task that the program must i

of 512 words. Of these 2048 block addresses the control." (Iverson and Yee, 1965, p. 82). ' most significant 256 are reserved for the use of the 2.41 "This paging procedure introduces another

supervisor and are sacred, i.e., any attempt by a important concept: virtual memory . . . The 1 user program to access these reserved addresses is programmer is limited only by the addressing '

prevented." (Morris et al., 1967, p. 69). capabilities of the machine; he has a machine I

"For each page of core store there is a- use digit which has as much memory as it can address. !

which is set when the page is accessed. All use In the Model 67, this virtual memory is as large i» digits are scanned and reset at regular intervals as 2^^ — more than 16 million — bytes." (Iverson

by the central executive and a pattern of use is and Yee, 1965, p. 82). ;

50 ". is of whether . . Users appear to have an insatiable appetite such circumstances there the problem for store and we consider large scale 'one level a given item should be fetched to a higher storage store' techniques to be vital." (Morris et al.. 1967, level, since this will be worthwhile only if the item p. 74). is going to be used frequently." (RandeU and Kueh- 302). "Dynamic memory allocation is a necessary ner, 1968. p. requirement for an engineering computer system 2.42 McGee (1965) provides the foUowing defini- capable of solving different problems with different tion: "Given a number of programs (each of one or more segments that have been partitioned into data requirements . . . The result of dynamic memory allocation is that the size of a problem that pages and stored in main memory), the function of a can be solved is virtually unlimited since secondary dynamic relocation technique may be stated as storage becomes a logical extension of primary follows: For each reference (5, i) by program r storage." (Roos. 1965. p. 426). to the iih word of segment 5, determine if the '". reference is legitimate and, if so, translate the . . Provide the user with a large machine- independent virtual memory, thus placing the re- reference to the memory location of the referent. sponsibility for the management of physical storage If the reference is not legitimate, signal a protection with the system software. By this means the user exception." (p. 196). is provided with an address space large enough to 2.43 "The attack on the single store concept ehminate the need for comphcated buffering and [as if there were] must be accounted a failure thus overlay techniques. Users, therefore, are relieved far. No executive or software system to the authors' of the burden of preplanning the transfer of informa- knowledge has solved this problem in a generally tion between storage levels, and user programs acceptable manner." (Wagner and Granholm, become independent of the nature of the various 1965, p. 288). storage devices in the system." (Daley and Dennis, 2.44 "It will be necessary to develop techniques 1968. p. 306). for transferring information on demand, and in "One important aspect in the efficient use of anticipation of demand, from the slow, more hierarchical storage that should be emphasized is voluminous levels of the hierarchy to the faster, the need for development of machine organization more readUy processible levels." (Licklider, 1965, and software techniques that make the entire in- p. 63). ternal and on-hne auxiUary storage appear as a Use of "WILL NEED" transfers and other single uniform storage to the user." (Hobbs, 1966, techniques to manage storage hierarchies have been p. 41). developed for lUiac II in order that this relatively "A very important problem remains to be solved, smaU computer (8,000 words of core) can be run on however, and that is how to provide efficient access time-sharing appUcations. (Fisher and Shepard, to this mass storage. Various schemes for organizing 1967, pp. 77-81). mass memory elements have been proposed, with 2.45 "Dynamic allocation of data achieves the storage arranged into levels by size and speed and foUowing: with hardware for automatic transfer among levels 1. Arrays are allocated space at execution time as required. Efforts have begun to make the entire rather than at compilation time. They are only memory appear as one logical unit to the system aUocated the amount of space needed for the user rather than as a primary and secondary memory problem being solved. The size of the array with separate addressing mechanisms." (Pyke, (i.e., the amount of space used) may be 1967. p. 162.) changed at any time during program execution. 2.41a "The choice of a placement strategy If an array is not used during the execution of should be influenced by several factors. These in- a particular problem, then no space will be clude the relative importance of minimizing storage aUocated. fragmentation, the frequency of storage allocation 2. Arrays are automatically shifted between requests, the average size of allocation unit, and primary and secondary storage to optimize the number of different allocation units." (RandeU the use of primary memory." and Kuehner. 1968. p. 302). "Dynamic memory aUocation would extend to

"When it is necessary to make room in working programs as weU as data. Programs can then be storage for some new information, a replacement brought into primary memory only when they are strategy is used to determine which informational needed." units should be overlayed. The strategy should seek "The aUocation ofvprograms and data must be to avoid the overlaying of information which may properly balanced so that the use of primary be required again in the near future. Program and memory is optimized." (Roos, 1965, pp. 426-427). information structure, conveyed perhaps by seg- 2.46 Thus, "it cannot be stressed too strongly menting, or recent history of usage of information that the strategies of storage aUocation must be may guide the allocator toward this ideal." (Ran- fuUy integrated with the overaU strategies for deU and Kuehner, 1968, p. 302). aUocating and scheduling the computer system "An additional complexity in fetch strategies resources." (RandeU and Kuehner, 1968, p. 301). arises when there are several levels of working We note also the foUowing: "The strategy of storage, all directly accessible to the processor. In storage accessing by the central processor is one

51 a

of planning so that all needed data (and program significant clustering patterns could be estabHshed segments) are available in the highest level storage at a level at which clusters were significantly smaller unit at the time of accessing or with minimum than complete programs. Moreover, clustering pat- delay." (Opler, 1965, p. 275). terns within programs are likely to vary with time, "The case of variable units of allocation is in and is hkely that a look-behind technique for paging general more complex because of the additional of individual processes would be more effective than possibilities of moving storage within working a 'static' clustering technique. . . . space in order to compact vacant spaces." (Randell "The information specifying the 'cluster' of most and Kuehner, 1968, p. 302). recently used pages is an important piece of infor-

2.47 "The assignment of variable size data mation and could be used for page control if it were structures to a conventional memory is a formidable available. It is felt that storage of the set of virtual problem because the space of physical locations page addresses on termination of a process as part for information and the space for names of infor- of an extended stateword, might be a worthwhile mation coincide ... If too few locations are hardware extension. For example, if these virtual assigned to a data structure, difficulty arises when page addresses were available, then the loaded additional locations are needed. Not only must process could check for the presence of these pages physical memory be reassigned, but the addresses and bring in any missing ones in parallel with the used to refer to them must be reassigned in name initial part of its computation." (Wegner, 1967, space." (Dennis and Glaser, 1965, p. 6). pp. 149-150). 2.48 "Multiple memory accessing becomes 2.51a "A basic allocation problem, 'core memory a burden when the total system responds to out- management', is that of deciding just which pages side calls, because then the time scale or patience are to occupy main memory. The fundamental of the caller is the determinant." (Dumey, 1965, strategy advocated here — a compromise against a p. 259). lot of expensive main memory — is to minimize page ". 2.49 . . Special processors should be traffic. There are at least three reasons for this: designed to allocate and relocate users' areas "(1) The more data traffic between the two levels in memories possibly in parallel with the main of memory, the more the computational over- computation." (Lock, 1965, p. 471). head will be deciding just what to move and 2.50 "The various requirements of relocata- where to move it. bility of programmes and data, indexing of ad- "(2) Because the traverse time T is long com- dresses, allocation of storage space and protection pared to a memory cycle, too much data of vital information, are all met at present by movement can result in congestion on the ad hoc provisions in the system design. Clearly channel bridging the two memory levels. they are all interrelated, and a unified approach "(3) Too much data traffic can result in serious to the design problem could perha^^s meet them interference with processor efficiency on more economically." (GiU, 1965, p. 204). account of auxifiary memory devices 'steal- "Dynamic storage allocation is achieved through ing' memory cycles." (Denning, 1968, 324). changing the associations of blocks of physical p. main memory with pages of segments. Common 2.52 "The 'page-turning' or 'look-behind' reference to a data segment by several processes concepts can be employed at three levels in con- is possible simply through the use of the same nection with segmentation. First, page turning segment name by both processes. It is natural may be employed within any segment so that only to implement data lockout on a segment basis." pertinent pages are kept in main memory. Second, (Dennis and Glaser, 1965, p. 8). the look-behind principle can be applied to the "If it is desired to protect the data object from problem of deleting entire segments from core destructive manipulation by an untrustworthy memory so that new information can be brought computation, routines with protected entry in. Finally, a look-behind technique can be used points . . . must be employed." (Dennis and to avoid use of the page index except when reference Van Horn, 1965, p. 15). is made to a new block of a segment." (Dennis 2.51 "It is likely that multiprogramming sys- and Glaser, 1965, p. 8). tems of the future will adopt some form of grouped 2.53 "The concept of a page-turning memory — page storage allocation, where the group of pages scheme by which only pertinent blocks of infor- allocated during a single storage allocation in- mation are kept in main memory while nonpertinent terrupt is determined either by the structure of blocks are deleted through a suitable algorithm — is processes or by the storage requirements during very useful for certain classes of procedures and the previous activation of the process. . . . data structures . . . (but) schemes that have "The problem of efficiently partitioning a problem been discussed to date do not allow for flexibility into hyperpages may be thought of as a clustering of application as they are not tied to a philosophy problem in which individual pages are represented of program structure; that is, the selection of by vertices into clusters having high density of information to be deleted is independent of its traffic within clusters and low density of traffic function in the system." (Dennis and Glaser, between clusters. However, it is not clear that 1965, p. 6).

52 ". . . It is sometimes claimed that a further user, and assurance that adequate secrecy will advantage of a paging system is that it entirely be provided to data when required." (Prince, 1966, eliminates the problem of storage fragmentation. p. 1705). Rather, what is true is that paging just obscures 2.53c "Since the program under test is likely the problem, since the fragmentation occurs within to be faulty, it is desirable to protect both the user's pages. It is only rarely that an allocation request permanent objects, and any objects created by the will correspond exactly to the capacity of an integral PLS (programming language system) on his behalf number of page frames, and many page frames from unintentional use or destruction by the pro- will be only partly used." (RandeU and Kuehner, cedure being debugged." (Dennis and Van Horn, 1968, p. 301). 1965, p. 21). "It is difficult to assess with any certainty the 2.54 "A number of users are testing procedures benefits of a demand-paging strategy in a time- under development that are not free from errors. sharing system. Computer configuration, work Program malfunctions induced by such errors must load environment, and other system characteristics not interfere with correct execution of tasks pro- such as scheduUng and priority schemes all strongly ceeding concurrently for other users." (Dennis and influence system performance; performance itself Glaser, 1965, p. 9). means different things to different people." (Fine "Programs very likely to contain errors must be et al., 1966, p. 11). run but must not be permitted to interfere with With respect to the studies by Fine et al. Smith the execution of other concurrent computations. remarks: "In a recent paper by Fine, Jackson, Moreover, it is an extremely difficult task to deter- and Mclssac, some relevant statistics on the mine when a program is completely free of errors. dynamic behavior of a particular set of programs Thus, in a large operational computer system in were reported . . . The authors concluded that which evolution of function is required, it is unlikely there was considerable doubt about the worth that the large amount of programming involved is of the page on demand strategy. Certainly any at any time completely free from errors, and the suggestion that it would be useful to multi-program possibility of system collapse through software a few pages from programs in the high speed many failure is perpetually present. It is becoming clear memory was negated by these statistics." (Smith, that protection mechanisms are essential to any 1967, p. 636). multiprogrammed computer system to reduce the Smith himself states that: "The techniques of chance of such failure producing catastrophic multiprogramming and paging have been proposed shutdown of a system." (Dennis, 1965, p. 590). as means for efficiently adapting a computer 2.55 "While losses of users' programs and data system to an interactive type of load. The imple- have occurred, their frequency and seriousness mentation of these techniques requires a con- have not discouraged users from entrusting their siderable hardware investment for the handling work to the system." (Fano, 1964, p. 18). of dynamic relocation and the best control policy 2.56 "Daily and weekly monitoring programs remains to be determined." (Smith, 1967, p. 636). have been devised that test the integrity of the 2.53a This conclusion rests in particular upon data. If any fault is detected it can be localized findings that "the initial call rate for pages is and repaired from any of the back up tapes that extremely high; the first ten pages, on the average, are kept for that purpose." (Kessler, 1964, 9). were required within about 5.6 ms; in half of the p. cases where 20 or more pages were required, 2.57 "For n equal to the number of terminals

. . . the time- the first 20 pages were needed within about 7.0 ms." currently connected, we can (make) and that, in terms of the number of instructions sharing operation introspective — that is by having executed between consecutive calls for new pages, it gather statistics about itself. But we cannot use "in nearly 59% of 1737 cases, less than 20 instruc- this device in general, unless we are willing to tions were executed; in about 80% of the cases, change n repeatedly. Even the most callous adminis- less than 200 instructions. In only 2.3% of the trator would hesitate before subjecting his custo- cases, 10,000 or more instructions were executed mers to that series of disruptions." (Greenberger, between calls; these longer sequences occurred 1965, p. 35). usually only after the program had accumulated 2.58 "The technological trend toward large a majority of the pages it required." (Fine et al., random access memory suggests the retention of 1966, p. 6). several users' programs in core simultaneously, 2.53b "Suitable systems programs are just now hence mutual memory protection must be ensured." becoming available which permit time-shared (Lock, 1965, p. 457). mutiple console operation. Much more work re- "Each user, and each user's program, must be mains to be done, however, in the optimization and restricted so that he and it can never access (read, refinement of these systems programs. Much also write, or execute) unauthorized portions of the high- needs to be learned concerinig the optimum use speed store, or of the auxiliary store. This is neces- of core storage for graphics time-sharing, establish- sary (1) for privacy reasons, (2) to prevent a defective ing suitable priorities for the consoles, protection program from damaging the supervisor or another of one user's programs from disturbance by another user's program, and (3) to make the operation of a

53 defective program independent of the state of the mode are provided, and several unusual 'book- rest of the store." (Samuel, 1965, p. 10). keeping' features, similarly based on interpretation, 2.59 "The [memory protection] capability must are available, such as the AUDIT command, which be dynamic and under flexible control of the execu- generates information as to which portions of tive program, since the allowed operating areas of the program were never executed, which variables the memory change on a millisecond basis." (Bauer, were never set, or set but never used, during a 1965, p. 23). given execution of the program." (Evans and 2.60 "Since a segment may be in several spheres Darley, 1966, p. 43). of protection at once, we have a basic framework 2.60c "For accounting purposes a multipro- for allowing independent tasks to share access to gramming operating system usually has to keep procedures and data as appropriate. . . . track of how much processor time is applied to "The attachment registers will only contain seg- each program and how much I/O channel time, ment names to which valid references may be made by channel and device, is used by each program within the established sphere of protection, with during its execution. Thus, the GECOS [General an indication of the class of reference permitted." Comprehensive Operating System] II system (Dennis and Glaser, 1965, p. 10). keeps a running total by program of all processor, "The complete memory protection solution de- channel and device time used. These totals are updated for each period of processor use and for mands that there be the option of read protection . . . Two important cases require it: The first case arises each I/O transaction. When a program terminates, in debugging where a program reading beyond its all of its accumulated times are transmitted to bound has erratic behavior which in practice is an accounting file and the totals are zeroed for indistinguishable from transient hardware failure. the next program." (Cantrell and Ellison, 1968, The second case is simply that of user privacy." p. 214). 2.60d "PILOT is a programming system (Corbato, System Requirements . . ., n.d., p. 6). "Protection of a disk system requires that no constructed in LISP ... an interface between user be able to modify the system, purposely or the user and his program, monitoring both the inadvertently, thus preserving the integrity of the requests of the user and the operation of his software. Also, a user must not be able to gain program. Advising is the basic innovation in the access to, or modify any other user's program or model, and in the PILOT system. Advising con- all data. Protection in tape systems is accomplished: sists of inserting new procedures at any or of the entry or exit points to a particular procedure (1) by making the tape units holding the system (or class of procedures)." (Teitelman, 1966, abstract records inaccessible to the user, (2) by making the input and output streams one-way (e.g., the input p. 1, and p. 26). file cannot be backspaced), and (3) by placing a 2.60e "A time keeping device maintains a mark in the input stream which only the system can continuous log of computer operations assuming cross. In order to accomplish this, rather elaborate two essential record-keeping functions that up schemes have been devised both in hardware and to now ordinarily have been handled by human software to prevent the user from accomplishing beings. Known as the Time Monitor, it records, certain input-output manipulations. For example, for each project, the overall time required to in some hardware unauthorized attempts at I/O prepare, utilize, and unload a computer, the manipulation will interrupt the computer." (Rosin, actual running time, the actual time for preparation operation when 1966, p. 242). and unloading, those periods during 2.60a "Service disciplines may be classified the computer is inactive, the reasons for the In addition to identifying operators, according to the information on which they base stoppage. it can identify programs and programmers." priority decisions. Such a list would be open- 25 (Nov. ended; however, the sources of the information (Computer Design 4, No. 11, 1965).) may be considered to fall in one of three not neces- 2.61 "Memory Activity — This . . . [shows] sarily disjoint environments: (1) the job environ- memory cell activity as a function of time. It pre- ment whereby the information consists of the sents a map of memory allocation and regional intrinsic properties of the jobs (e.g., running time, activity." (Coggan, 1967, p. 77.) input/output requirements, storage requirements, "Paging Activity — This presents a measure and etc.), (2) the (virtual) computer system environment map of page (memory block) use during specified (e.g., dynamic priorities may be based on the state time quanta. Separate displays for activity during of the system as embodied in the number of jobs compile, execute, etc., times may be desirable." or requests waiting, storage space available, (Coggan, 1967, p. 78.) location of jobs waiting, etc.) and (3) the program- "Subroutine Activity — This function . . . aids mers' or users' environment in which management in determining what subroutines should be included may assign priorities according to urgency, impor- as part of the standard program vocabulary." tance of individual programmers, etc." (Coflfman (Coggan, 1967, p. 79.) and Kleinrock, 1968, p. 12). 2.62 "Loop Factors — This display illustrates the 2.60b "[In QUIKTRAN] extensive run-time total number of times around various loops defined diagnostics made possible by the interpretive by the program flowchart." (Coggan, 1967, p. 80.)

54 "Trap Activity— This is a measure of activity as a 2.66a "The logical units of information transfers function of time for different trap types. It presents carried out by recent supervisory systems represent a map of program control interrupts." (Coggan, a change from variable length program and data 1967. p. 79.) structures to the fixed length pages into which these 2.63 "Queue Length — This presents a measure programs and data are fitted. A number of important of the queue lengths for different classes of hard- advantages accrue from this organization of infor- ware, such as processors and memories, as a mation and have been described along with a corre- function of time." (Coggan, 1967, p. 82.) sponding control structure by Dennis and Arden, "Work Profile — This is a display of the time et al. A basic problem in this organizational approach breakdown for equipment usage and can yield infor- has appeared in the form of poor object program mation about particular equipment activity as a operating behavior in a multiprogramming environ- function of time, or serve as a comparison of ment. In particular, the frequency of the page turn- equipment use over a fixed interval." (Coggan. ing (transferring pages in and out of core memory) 1967. p. 76.) necessary to the execution of a program which, in 2.64 "Gibson Mix — This is a measure of fre- general, is never wholly in core tends to degrade quency of instruction usage and serves as an aid system performance by introducing as excessive in determining the types of instructions that should amount of input/output interferences. . . . by hard-wired." (Coggan. 1967, p. 76.) "The two-page replacement algorithms whose "Memory Profile — This gives a measure of the performance is discussed here are as follows: total number of memory cells used as a function of 1. The Least Recently Used (LRU) Replacement time and offers a direct comparison to the total Algorithm. Whenever a page of a given pro- number allocated. (Efficiency of use)." (Coggan, gram must be turned out of core, this algorithm 1967. p. 77.) specifies that the page turned out must be the "Branch Statistics — This is a measure of the one which has been in disuse for the longest branch probabilities for various branch groups. period of time since the beginning of program The display can help iUustrate frequent and time operation. This algorithm was selected be- consuming operations that might best be per- cause it is very similar to existing or proposed formed by more efficient program code." (Coggan, algorithms and admits of a fairly simple 1967. 81.) p. implementation. "Program Overhead — This is a dfsplay of over- 2. The Belady Optimum Replacement (BOR) head time for different programs operating in a Algorithm. This algorithm is based on a prior multiuser system." (Coggan, 1967. p. 83). knowledge of the entire sequence in which 2.64a "The space-time product will be affected pages are used in the execution of a program. by the time taken to fetch pages, which will depend With this information and a specified maximum on the performance of the storage medium on which on the number of pages that can be held in pages that cannot be held in working storage are core at one time, Belady has shown how to kept. If page fetching is a slow process, a large sequence the paging-in and paging-out opera- part of the space-time product for a program tions in an optimum way, i.e., yielding mini- may well be due to space occupied while the pro- mum page turning. Because of its require- gram is inactive awaiting further pages." (Randell ments the algorithm does not appear practical; and Kuehner. 1968, p. 302). however, it was selected in order to provide 2.65 "Managers of a time-sharing operation wiU a means of comparing the performance of the find themselves dealing more and more with the various practical (and less efficient) algorithms problems of how and when to interact with the with the best possible case. . . . system. One of the tools designed and implemented is the manager program. One part of this program "The programs considered for analysis in this displays aU the users, their station number, size of study represent rather different external charac- program, work order number, priority number, and teristics. The programs selected include a SNOBOL information regarding availability of space on disk, compiler, a program for computing Fourier trans- drums and tape drives. . . . The other feature dis- forms, a WATFOR FORTRAN compiler, and a plays all the programs that are on the disk: the pro- differential equation-solver. The last program is gram name, number of tracks, status (active or relatively small compared with the first three. inactive), and date last used. Utilizing these pro- The statistics for the larger programs have proved grams, the monitor can get the information he needs quite similar in nature and will be summarized to make tactical decisions." (Fiala, 1966, pp. 164- by a detailed look at the behavior of the SNOBOL 165). compiler which consisted of 15 instruction and

2.66 "The development of the simulation pro- 22 data pages, with a page size of 1024 words. . . . gram now provides a first-come-first-serve queue "The principal observations from the data we unloading strategy. Continuing effort, however, wiU have obtained are: (1) with the possible exception provide for optional strategies, e.g., selecting the of carefully designed programs, the page turning data unit with the shortest servicing time, con- incurred when programs are made to operate while sideration of what flow will minimize idle time at substantially less than totally core-resident appears the central processor, etc." (Blunt, 1965, p. 15.) excessive in light of current or proposed paging system designs; (2) a least-recently-used page may be constructed. Whatever is done, some super- replacement algorithm yields a performance within visor overheads are incurred. Suggestions have about 30 to 40 percent of that of the optimum page been made for additions to the hardware which replacement sequence; and (3) for page residence will reduce these overheads; perhaps the most confined primarily to small areas within the page useful is that there should be a bit in the hardware size, performance is improved substantially more associated with each physical page which records by increasing the number of pages held in core whether that page has been accessed since the than by increasing the page size." (Coffman and last time of resetting. Suggestions have also been Varian, 1968, pp. 471-472, 474). made for some type of analogue timing circuit 2.66b "In either of the schemes described, associated with each physical page which will or in any of their many variants, there is no diffi- enable the page longest resident to be identified culty about arranging that pages should be loaded very rapidly. It is not clear whether such a circuit only when they are required. There is, however, could be designed so that it would record time more difficulty in arriving at a satisfactory algorithm only when the program owning the page was run- for deciding which page should be dumped when ning." (Wilkes, 1967, p. 4). it is necessary to make room for an incoming page. 2.66c For example: "Some problems have One policy is to dump the page that has been longest apparently been solved in an empirical or experi- resident in core, and another is to dump the page mental fashion. For instance, we know of no that has been longest resident in core without theoretical studies on methods of allocation of being accessed. More elaborate criteria, based processor time to various programs or on the option the past history of all pages in core, and taking mization of information storage in memories with account possibly of the fact that some programs various hierarchy levels." (Pointel and Cohen, are reacting with their consoles while others are not. 1967, p. 46).

3. Storage, File Organization, and Associative Memory Requirements

3.1 "The management of secondary storage decisions as to the allocation of storage space and which has plagued programmers for many years the movement of information between levels of a should be handled automatically by the total soft- storage hierarchy." (Randell and Kuehner, 1968, ware system. The designers of the software system p. 298). for the ATLAS computer were pioneers in attacking "The dynamic nature of multiprogram on-line this problem. We find the designers of most large computation should have a strong influence on modern software systems gingerly probing at the memory organization." (Lock, 1965, p. 471). same problem." (Wagner and Granholm, 1965, p. "If the extra memory which can be added 287). can be overlapped and if there exists a hier- 3.2 "Data base storage will require medium archy of memories of differing speeds then both capacity, random access, solid-state, on-line auxil- the compilers and the operating system will have iary storage; large capacity, low cost, on-line auxil- to concern themselves with making sure these iary storage; and very large capacity, very low cost, features are used." (Clippinger, 1965, p. 211). off-line auxiliary storage." (Hobbs, 1966, p. 41). "Memory management — that is, having the right "Technology offers a broad variety of access and material in the right place at the right time — storage modes, so that less frequently used, or less plays a central role and calls for a carefully thought- ' valuable, records could be stored more cheaply at out strategy if the supervisory program is to handle the cost of slower access." (Bohnert and Kochen, this matter expeditiously and automatically." 1965, p. 157). (David, 1966, p. 44). ". "We need to better understand the flow of data 3.4 . . The newer time-sharing systems con- and develop design guidelines here, recognizing template a hierarchy of file storage, with 'percola- if that tion' algorithms replacing purging algorithms. the access 'gap' from electronic memory through ; electromechanical storage devices will cover a Files will be in constant motion, some moving 10''." relative access speed differential of 10^ to 'down' into higher-volume, slower-speed bulk store, ;

(Hoagland, 1967, p. 258). while others move 'up' into lower-volume, higher- I 3.3 "In large time-sharing systems the means speed memory — all as a function of age and refer- for organizing memory is of prime importance, and ence frequency." (Schwartz and Weissman. 1967, the next several years should bring a variety of p. 267). solutions to this problem." (Pyke, 1967, p. 162). "In a multiple-access multi-programmed system, \ ". . . The problem of assigning data blocks and the use of paging gives a number of advantages. / program segments statically to various types of In the first place, when a fresh program comes to ' storage units is replaced with the problem of moving be loaded, it can go into any physical pages that can such information dynamically through a hierarchi- be made available, whether or not they are con- <

cally organized storage." (Opler, 1965, p. 275). secutive. Secondly, paging allows traffic in the 1 "The information conveyed by the fact of seg- channel between the drum and the core memory to f mentation can be used by the system in making be reduced, since a page of program need not be loaded unless it is actually needed. The procedure of simple adaptive schemes for handling informa- is to load the first page of a program and send tion transfers between storage levels." (1963, p. 176). control to it. If. in due course, an attempt is made 3.7 "Unless an unexpected breakthrough recon- to access a word in another page, a supervisor trap ciles fast random access with very large capacity, occurs and the supervisor loads the required page. there will be a need for memories that effect Since, in multiple-access working, programs are various compromises between those desiderata." often loaded and run for short periods only (until (LickHder, 1965, p. 63). they reach an input or output wait, or have ex- "The concept of storage hierarchies is very impor- hausted their quantum of time), this advantage of tant in considering the use and capabilities of paging is an important one. A further reduction in storage devices. There is no one ideal type of traffic can be obtained if, associated with each storage that fulfills all requirements while pro- page register, there is a bit which is originally viding the maximum speed and capacity for the zero, and becomes set to one the first time the page minimum cost." (Hobbs, 1966, p. 41). is written into. The supervisor can then avoid writing 3.7a "A MIS usually consists of two operational back on to the disc pages which have not been categories of programs; real-time and batch. The altered during their residence in core. Memory file system to support these processes must at least lockout can be associated with pages if extra bits consider: associated with the page registers are provided 1. Security of information for the purpose." (W ilkes. 1967. p. 3). 2. Recovery of system operation after failure, and 3.4a "The reviewers believe that file-main- restoration of data when mutilated. tenance operations have not received adequate 3. Key and Addressing schemes that will result consideration in the design of file organizations. in densely populated storage. e noted that the use of indirect addressing of 4. Need for additions and changes to the data data could result in important savings for mainte- base to meet changing requirements. nance operations in indexed and multilist appli- 5. Performance requirements of all types of cations. It was also noted that an alternative design activity. approach to accommodate maintenance require- 6. Reduction of wasted mass storage, consistent ments was to reserve space in storage for this with performance requirements. purpose. It would be useful if quantitative per- 7. Coordination of the file system with the pro- formance studies were developed so that designers gramming system." (Benner, 1967, p. 291). might know the circumstances under which one is more advantageous than the other."" (Minker 3.8 "Hierarchical memory structures (with the and Sable. 1967. pp. 150-151). levels set by the access-time characteristics) are 3.5 "The system design includes a hierarchical common in mass storage systems to balance cost, directory structure. Each item in a directory asso- capacity, and access time performance against ciates an alphanumeric name with a pointer applications requirements. . . . to a file, an input/output function, an entry point "Economical mass-memory systems are the real giving access to some system service, or a directory key to extending the power of computers into many giving further associations of names with pointers." new applications." (Hoagland, 1965, p. 55). (Quarterly Progress Report Xo. 80. Research "This 'main' memory size is related to the proc- Laboratory for Electronics. M.I.T, 257 (1966).) essing rate; the faster the arithmetic and logic units, 3.5a "Early implementations of memory pro- the faster and larger the memory must be to keep tection were aimed almost exclusively at providing the machine busy, or to enable it to solve problems the write protection function which is essential without waiting for data." (Hoagland, 1965, p. 53). for guaranteeing that one program cannot destroy 3.9 "A stream is a combined allocation of several another. The multiusage environment adds further levels (presently core and disk) of storage, used as dimensions to memory protection requirements. a single-level continuously addressable store." Privacy considerations of privileged information (Connors, 1966, p. 201). (such as payroll data) require that portions of 3.10 "In my opinion, this problem of the minimal memory be protected from unauthorized reading redundancy store must be considered as one of the as well as writing." (Mendelson and England. long-range problems for information systems." 1966. p. 62). (Mooers, 1959, p. 28). 3.6 Barton comments: "Since multi-level An obvious practical requirement for consolida- storage is an economic necessity, its efficient tion or correlation of new input items with items utilization has long been a prime objective, though previously stored is that of checking for duplicate the difficulties appear formidable. In this regard, entries in the file. Then, "in another . . . kind of the machine designers have largely ignored the situation, the flow of input documents will contain problem: extensive programmed control, very many facts or purported facts, and these must be costly in terms of storage, is usually the conse- compared with information in the files for veracity, quence. It now appears that things can be done consistency, novelty, or the hke." (Mooers, 1959, to alleviate this problem by providing adequate p. 17). interrupt signals indicating states of tape, disc, In one illustrative experiment, it is reported that and drum storage components and through use "in the case of a duplicate statement, only the set 57 identifier, R, is stored. The number of R's stored points* becomes possible. Thus a very rich structure serves to enforce the vaHdity of the corresponding can be created and manipulated with the file statement." (Arnovick, 1965, p. 183). system. . . . 3.10a Greenway and Russell point out that: "As mentioned above, any aggregate may be "The final form in which information is presented an element of any other aggregate, indeed it may to the user is a function of economics, required be an element of any number of aggregates. Simi- response time, and human factors. Thus, informa- larly a file may be a member of any number of tion presentation might range from a printed page aggregates. Thus in both cases we have many

to a high-speed sophisticated three-dimensional records pointing to the same record. If such pointers I were to be in the form of an address or anything display device. Tlie importance of human factors | in information presentation is pointed out by the else which was somewhat hardware oriented, each fact that display on a microfilm reader often appears pointer would have to be changed whenever the very desirable from an economic and engineering record to which it pointed were moved. This would Ij standpoint, but meets with considerable resistance require reverse pointers in the record being pointed from users." (1965, p. 3). to so as to know which records require updating. 3.10b "A viable file organization must be One soon becomes enmeshed in an overwhelming adequate from several points of view. maintenance job. Therefore, due to the very dynamic character of the secondary storage in

1. It must permit economical retrieval of items, INTIPS, it is necessary that inter-record references by individual or by class. be symbolic. Aggregates, Files and Data Blocks 2. It must permit economical modification of its can each be assigned a symboHc name by the user items, by individual or by class. but these symbolic names do not readily serve 3. It must be economically representable in one the purpose of inter-record references because or more memory media of a computing system. they are too long and also any given File can have 4. It must be adaptable to a range of computing many symbolic names. Therefore, another set environments, present and future." (Holt, of symbolic names was derived to be used for 1966, p. 465). inter-record references. They are assigned by the file system solely for its own use and are called 3.10c "This paper describes the file handling System Assigned Symbols (SAS's). Since these system developed and being implemented as a symbols are never exposed outside the file system part of INTIPS, (INTegrated Information Proc- they need not be binary representations of any essing the aegis Air System) under of Rome Develop- legitimate alpha-numeric characters. Instead they ment Center. This file system addresses itself to are purely binary and therefore more compact." the physical problem of the storage and retrieval (Ver Hoef, 1966, pp. 75-77, 80). of fixed length blocks and their organization into 3.11 "The activity method of file organization higher-order structures. It does not concern itself has an added appeal, because it provides the data with the contents of these blocks. . . . for systematically purging the file of rarely used "The data base must be grossly expandable items." (Borko, 1965, p. 24). without reprogramming or redesign. Indices must "We make readily available those items which be small and efficient while the data base is small are most likely to be wanted and make relatively but grow automatically with minimal decrease in inaccessible those items which are least likely efficiency as the data base increases to its projected to be wanted." (Hayes, 1963, p. 287). 10-' maximum of blocks. . . . "Similarly, within the record different portions "The file structure is basically that of a tree. of the record have different activity rates. Fre- Blocks are grouped into files where a file consists quently a record of 700 alpha-numeric characters of 1, 8, or 256 blocks. As will be seen in subsequent in length will have a small portion, some 20 to 30 sections, by setting discrete sizes, one gains |

characters, that will get a major share of the activity. 1 advantages in reduction of overhead when accessing Some system designs use a "creamed" file of these related blocks and in simplification of the storage 20 to 30 characters per record in a computer to allocation problem. The choice of these three sizes answer 80 percent of the questions, with the other was somewhat arbitrary but it was desired to have 20 percent handled by manual or microfilm files." file sizes vary by approximately an order of magni- (Tauber, 1966, p. 276). tude. Files as well as blocks, may be symbolically 3.12 "The system will reduce routine search named. file No two blocks in the same may have the time and conserve the more costly storage space but given same name any block name may be used by lessening the rapidity of access to data as in any number of distinct files. Every block is a it decreases in usefulness. Thus data on a particular

member of one and only one file. . . . individual who is incarcerated or has been dead readily that allowing "It can be seen by aggregates for some time will be placed in storage with lower to consist of other aggregates as well as files, a accessibility than data on active criminals." structure can be developed which is many levels (Geddes et al., 1963, p. 24). deep. These levels are called terraces. In addition, by permitting files and aggregates to be members *A convenlinnai tree structure allows any ^iven node to subsume several nodes but it may only be subsumed by one node. If. on the other hand, such a node may of more than one aggregate the creation of lattice be subsumed by two or more nodes, it is defined to be a lattice point.

58 "The optimal organization [of the data file] immediately before it in the file, and relocating a will depend on the primary functions of the system, record only after a specific number of queries using and it may prove advantageous to have different either of the first two methods. Assuming the file parts of the file organized differently so that one must be searched serially for each query, he found part can respond efficiently to one kind of demand that the two basic methods were equally as good in and another part to another kind." (Travis, 1963, improving search efficiency and that using sampled p. 331). queries merely slows the organization process." "The file structure should exhibit the various (Climenson, 1966, p. 112.) levels at which data will be used, yet allow for the 3.14b "The performance region of several integration of all related data. Modified list-type hundred million bytes of data with access in milli- structures designed for the needs of the particular seconds is a key to many on-line 'data bank' type system are worthy of much consideration for such information systems and the appeal of such facilities files in order to provide the desired flexibility, growth, spurs the rapid trend to direct access file-oriented and cross-referencing." (Gurk and Minker, 1961, systems." (Hoagland, 1967, p. 258). p. 270). 3.15 The terms "associative memory" and 3.12a "Multiple entries for a file item expand "content-addressable" memories are generally the size of the file and require more duplicating and used interchangably. Cheydleur, however, makes more infihng. The whole purpose, however, of a the following interesting distinction: "Content-

multiple-entry approach is to achieve an organiza- addressability . . . can be described as the capa- tion of the file so that all duplicates having to do with bility for locating items via their keywords (with a particular code category may be brought together the understanding that at various times any of the in one section of the file, then, when a search must fields of an item suddenly may be subject to treat- be made, only a portion of the total file need be ment as a key word) rather than locating via their presented for the searching operation." (Kuipers actual memory addresses; basically this localization et al.,l957,p. 258). must be eff^ected without benefit of any kind of 3.13 "Economic operation is feasible only if the serial sorting of files or subfiles either at the time of

predetermined pattern of the file organization cor- record or at the time of recall . . . responds to the environment." (Hayes, 1963, p. 268). "The desirability of achieving an associativity 3.13a "More difficult problems arise when paral- capability, distinct from content-addressability, lel use of files is being made and one job is termi- can be based on the need for coping with file require- nated abnormally after only some of its tasks have ments such as (1) multiple response, i.e., the corre- been completed. What is the status of the updated spondence of more than one item to the specification file? The same question arises if a file is updated required; (2) inexact but close matching of 'most early in a job. Should the updated version of the of the specifications'; and (3) rejection of items file be made available to other jobs before the full which have certain specified explicit descriptors." job is completed? Conflicts of this type demand a (Cheydleur, 1963, pp. 57-58). great deal upon the necessity of the file being com- Some other definitions are as follows: "An as- pletely up-to-date and upon the various structures sociative memory is a storage device that permits of the files. In general, the rule is that an up-dated retrieval of a record stored within by referring to file is not available until after its job is completed. the contents or description rather than the relative If it should be available eadier, then that job should address within the memory." (Prywes, 1965, p. 3). be broken into two other jobs so that at the end of "We have described here an iterative cell which the first job the file can be released. This makes the can be used as a content addressable memory rules easier for users to understand and simplifies from which an entry may be retrieved by knowing the restarting problem." (Gosden, 1964, p. E2.2-7). part of its contents." (Gaines and Lee, 1965, p. 74). 3.14 "We will find random-access mass memory "This paper describes a fixed memory consisting handling images as well as digital data by storing of one or more stacks of paper or plastic cards, ." them in video and in digitized form . . (Hoagland, each of which contains an interconnected array of 1965, p. 58). printed or silk-screened film resistors. Each card 3.14a "Mass storage control logic is now begin- is compatible with conventional key punches, and ning to extend to the functional ability to automati- information is inserted by the punching of a pattern cally arrange records within the hierarchy of of holes, each of which breaks an appropriate memory/storage according to their activity, signifi- electrical connection. All punched cards in a stack cantly improving access utilization." (Hoagland, are cheaply and reliably interconnected using a 1967, p. 2.58). new batch interconnection technique which re- "In applications involving the single-item search, sembles an injection molding process, using molten the analysis performed by McCabe can be useful. low-temperature solder. The circuit which results He examines the possibilities of reordering a file is a resistor matrix where the information stored is based on moving the records for which a query has in the form of a connection pattern. The matrix been made. Three methods of reordering the file may be operated as a content-addressable or asso- are suggested: moving the retrieved record to the ciative memory, so that the entire array can be front of the file, exchanging positions with the record searched in parallel, and any word or words stored

59 366-108 O - 70 - 5 answering a given description can be retrieved in at least scratchpad-size associative memories microseconds." (Lewin et al., 1965, p. 428). practical." (Hudson, 1968, p. 42). "The information records in the associative 3.19 "The utility of content-addressable mem- memory are organized into lists. Each list can ories (CAM's) within a general purpose computing be visuahzed as a series of cells, each ceU con- system is investigated ... A simulation package taining an information record, linked along a linear is developed which allows simulation of CAM com- string which connects "associated" records. mands within job programs run on the IBM 7090 and Each of the records may be of variable length and derives tallies of execution times corresponding to may have a number of associations, with a link a particular realization of a CAM system. The simu- to another record for each association. Each string lation package is used to evaluate algorithms for represents an associative notion and has a cor- CAM commands and to evaluate the efficiency of responding name which is referred to in the fol- CAM commands in a job program." (Fuller, 1963, lowing key. The branching lists make up what is abstract). known as a tree. The word tree figuratively describes 3.20 "Memory organization is the kernel of the the branching out of lists from the main chain of Moore School's problem-solving facility . . . With information which provides paths within the the multilist technique information storage and re- associative memory. Thus, the processes of storage trieval are incorporated in the executive program and retrieval are regarded as establishing and and in a man-machine communication language following association links among the information called Muhilang." (Carr and Prywes, 1965, p. 88). records." (Prywes, 1966, pp. 1790-1791). "The functions of storage and retrieval of records "An associative memory has two attributes: are accomplished by the use of an associative content-addressability and parallel-search capa- memory which is simulated on an addressable bility." (Chu, 1965, p. 600). memory by means of the MULTILIST technique.

3.16 "The content-addressable memory (CAM) This technique . . . involves the use of an expand- was initially proposed by Slade as a cryogenic ing tree structure as a dynamic index translating device." (FuUer, 1963, p. 2). See Slade and McMahon keys into addresses, and a multi-association area (1957). where the records themselves are stored." (Prywes, 3.17 "An early example of a parallel search 1965, pp. 2-3). memory was the 'multiple instantaneous response "The Multilist system simulates a storage system file' (MIRF), where the comparison function is that contains an associative memory as a substruc- built directly into the cells storing data items, ture. From such a descriptive memory, a program- permitting a given search specification to be mer can retrieve stored information by specifying reflected against all data cells simultaneously. a key — that is, by giving a description of the infor- Knowledge of the address of the qualifying data mation desired, rather than its address." (Carr and item is not needed. Address independence has Prywes, 1965, p. 88). also been achieved in list-processing systems, 3.20a "An associative memory for large scale where search proceeds through a series of chained data processing can be implemented in a conven- lists rather than in parallel. A short study of both tional addressable memory by simulation tech- approaches has been provided by Reich." (Climen- niques." (Landauer, 1963, p. 863). son, 1966, p. 114.) 3.21 "One of Feldman's main points is that there "Previous work in automatic data extraction, is some evidence that large associative memories automatic indexing, and formatted file processing may be competitively simulated on conventional suggested that the techniques of content addressing computers. His principal evidence is that the execu- and parallel processing afforded by an associative tion times are reasonably competitive, but he further processor would offer significant advantages both argues that large associative memories can be simu- in performance and in ease of programming for lated on immediately available hardware, and that such applications." (Baker and Triest, 1966, the general-purpose machine has the ability to p. 1-1). handle the multiple-hit problem and tree-structured ^ 3.18 "It is very unlikely that associative proces- data more readily. Another point is that "if much sors will be developed to an adequate state for processing must be done outside the associative widespread utilization in the foreseeable future." memory the saving in the time for the first access (Hobbs, 1966, p. 41). may be a negligible part of the total." (Climenson, 3.18a "It is extremely unlikely that large fast 1966, p. 116). associative stores will become practicable in the 3.22 "List organizations of memory are generally useful in problems for which: near future." (Scarrott, 1965, p. 137). "In the past, associative or content addressable 1. Symbols having other than numeric signifi- memories of any significant size have been im- cance are to be manipulated. practical for widespread use. Relatively small 2. Dynamic storage aUocation is desired since associative memories have been built with various storage requirements cannot be determined in technologies, such as multiaperture ferrite cores, advance. cryotrons, and various thin-film techniques. The 3. Orderings exist among symbols or among sets logical flexibility of microelectronics now makes of symbols or both. Sets of symbols may have 60 many subsets. Subsets may belong to more be followed closely in future reviews. The reviewers than one set. believe that additional features, such as indirect 4. Orderings among symbols are restructured in addressing to names of records (as described pre- the course of program execution. viously with the registry number) and data descriptions, will be needed in the operational Problems for which Ust storage has been used in- system." (Minker and Sable, 1967, 141). clude theorem proving in geometry and logical p. calculus, chess playing, assembly line balancing, 3.24 "To achieve the fuU utilization of the translation of natural and artificial languages, and memory space, any abandoned space must be 'recaptured' in order that it information retrieval." (Fuller, 1963, p. 6). may be used subsequently for other purposes. The association memory 3.23 "The choice of file structure within a ran- . . . starts with all available space on a single dom access device is dependent upon the problem. long list, called the 'available space fist'. Whenever Multi-list file structures, chaining, address genera- space is required for building up a new hst, this is tion techniques, and dictionaries have all been obtained by using the words from the front of the found valuable." (Berul, 1964, pp. 10-17). available space list, and whenever information "A record is retained in file for each document. is erased and the words that held it become available The record contains the code for each index term for use elsewhere, these words are added to the attributed to it. Next to the index term code is front of the available space Ust. Thus, the fact the address of another record which contains the that unused space is scattered all through the same term code. A search for all records with a cer- memory creates no difficulty in finding new space, tain term code may be performed by 'threading' for there is a single known word (the name of the through the file from one random access address available space list, ASL) that always contains to another. . . . the address of the next available word." (Hermann, "A further development of the Inverted List 1960, pp. 14-15). method has been developed at the AUERBACH 3.24a "Herner notes the two basic ways of or- Corporation. To reduce. the amount of list manipula- ganizing file material: the 'unit record' concept and tion, a technique is employed which selects the the 'inverted file' concept. The advantages and dis- lists most liable to be critical in the search. Only advantages of each in their pure form these hsts are manipulated, and the number of — — were aptly described by Prentice: unit records ultimately retrieved from random access The record concept has the advantage of having all information in the record is greatly reduced." (Berul, 1964, pp. 8-24). available for search at one time, resulting in high "The work of Lefkovitz et al., in developing an search precision. It has the basic disadvantage that experimental system for a real-time chemical in- all records in the file must be searched for every formation system is of general interest because of request. On the other hand, the inverted file search its file structure features rather than its special need only involve the pertinent part of the file, but interest for chemical retrieval. The work is being it retrieves only record identifications; it deal performed for the Army Chemical Information must with unbalanced record posting densities, and the Data System (CIDS) effort. Except for some special relationship between individual record features, such as chemical substructure searches terms is difficult to express and search." (Climenson, and relational searches between entities, work 1966, p. 109). in the general field of file structures is applicable to the chemical retrieval problem, and the specified "Transience is a strong incentive to conventional chemical problems can be embedded within the term-on-item grouping. If the item has a short life, it is desirable larger framework. The file structure used in the to have terms grouped under items so that CIDS project is a modified multilist file structure." items and their terms can be deleted with ease from the store in one simple operation." (Wall, (Minker and Sable, 1967, p. 140.) "Modifications to the multilist file structure 1958, p. 172).

made by Lefkovitz et al., make it more of a com- "Experience with the program has also shown bined inverted file, multilist structure. An auto- that an inverted file of document numbers posted matic classification scheme is being developed to to indexing terms, the original scheme, is less anticipate the types of questions and to place data suited to the total NASA information program in cells based upon their attributes and near- than a linear file with complete document records related data. The scheme is basically a data-storage placed on tape and searched in sequential order. allocation strategy for disc, which attempts to save With the linear arrangement, it is more feasible accessing data on different disc head positionings. to divide the total file both chronologically and by No experience exists on its usage. As the authors subject area. Thus, to shorten the total time re- note, the efficiency of the system wiU depend quired to make any given search of the document highly upon whether the keys are indexed. For store, only selected portions need be searched. substructure searching, without other keys identi- The linear file is also most suited to a decentralized fying the data, they found that a tape file compares system in that the updating of searching files held by favorably with a disc system. The authors believe subcentral facilities might be done simply by dis- that to be effective, the longest Hsts should encom- tributing the new sections of data which are then pass only a small part of the file. This work should easily added to the ends of existing files in the field.

61 A further result of experience with a decentralized potentially excessive use of memory by simply system is that numerical codes for subject terms are providing a cross-reference capabifity without troublesome, especially when the vocabulary is rephcating the full index record." (Meadow, 1967, changing and evolving as fast as the aerospace p. 240). vocabulary. Conversion of indexing data from codes "Combinations of the inverted file and unit record to straight alphabetics was a natural solution for approach are replacing pure applications of either." the NASA program; together with the linear file (CUmenson, 1966, p. 129). arrangement, it is expected to be effective in mid- "When data are stored on tape for a retrieval 1964." (Wente, 1965, p. 58). system, they are commonly organized by record "Variations in search logic can improve the name and other identifiers, followed by record de- efficiency of a file set with an inverted index. scriptors. To speed retrieval, some designers pro- In the most common mode of usage, when the vide an index that is searched first, followed by search specifies a Boolean combination of terms, retrieval from the main file, which is sequenced on the union and intersection operations can be tape by citation number. However, Curtice shows carried out on the list of record locations retrieved that, for serial media, searching the main file is from an inverted index. Another method that has generally more efficient than searching the index. been proposed by I. A. Warheit is to locate only Thus, an index stored on tape may hinder rather one term of a request set in the inverted file, than enhance retrieval time. Furthermore, the index then search out the index records listed there. must be maintained, thus making maintenance The full logic is applied only to the index records operations more complex. Curtice further notes that retrieved by this term. This method offers an storing both the index and the main file on disc wiU advantage when the expected number of index lead to an advantage of the indexed approach not records to be detail-searched is low, the number achieved for tape-stored data. Thus, we note of comparisons to be made during the record again that the file organization, to be effective, must all search is also low, and the query terms are be matched against the physical storage medium. . . . intersected. The method works best if the least "Experience on an actual system at the NASA probable term can be selected from the request Scientific and Technical Information Facility set and used as the only entry to the inverted file. confirms Curtice's results. Brandhorst & Eckert This technique is called direct access searching .... present a brief historical review and a cuirent up- Only one of the two query terms is looked up in the dating of the NASA efforts as accomplished by inverted index file. All index records containing Documentation Incorporated. They report that the that term are examined for presence of the other straight linear-file search using tape is an improve- query term. This method reduces searches in the ment over the inverted-index search followed by inverted file and eliminates cumbersome logical a linear-file search. However, they intend to use operations on the lists of document numbers the inverted-file approach for discs when their new retrieved from the inverted file. If a third table, computer is obtained." (Minker and Sable, 1967, an index of the inverted file, is used, it would be p. 128). possible to include not only location of term records, "A hnear file search approach was adopted for but the number of index record references in each, use in the system. An inverted file search economic- and this could be used to select the single entry ally demands permanent allocation of random for the inverted file. On the other hand, maintaining access equipment, the expense of which precluded and retrieving relative frequency of use data may its use in the system. However, the Descriptive use up all the gains otherwise provided." (Meadow, Terms Statistical Master File, the Satellite File 1967, p. 239). which is used to produce the term postings report, "It is even less desirable to invert a complete is actually an inverted term file. Therefore, when file in non-document-oriented files. The low search inexpensive random access storage becomes avail- probability of most of the many fields in typical able, the conversion expense to an inverted file data records makes allocation of extra memory to random access search will be minimal. Also, speed the occasional search uneconomical. Consider unless hindsight indicates revision, the basic what happens when the index file is sequenced on question statement method will be retained." some more meaningful basis than accession number. (Ebersole, 1965, pp. 193-194). We must certainly bear in mind that one result is "In an attempt to exploit the benefits of both an a requirement to sort all incoming index records. inverted file and unit-record concept and circumvent Any sequence other than accession number must the disadvantage of each, Warheit has devised a be sufficiently more efficient than access number so-called direct access search system where both order to overcome the penalty of sort time. Sorting kinds of files are used. The inverted file is used as on subject would often be desirable, but many in- an initial search filter on the basis of one or two dexing systems permit and encourage several sub- coordinated search terms. The assumption is that, ject terms per index. Sorting of index records by if there is a reasonably small number of document subject would then require each record to be repli- postings associated with the initial search terms, cated as many times as there are subject terms in only a few records in the unit-record file must be the record. The inverted file concept avoids this searched in detail as the second step. The price

62 paid is storage redundancy. It should be noted that In various forms, these applications include incomplete use of an inverted file is equivalent to microprogramming, journal keeping, and arith- exhaustive search of an inverted file where the terms metic and logic manipulations." (Fleisher et al., have a many-to-one relation to the unit-record 1965, p. 25). terms — that is. in systems where the terms are "Lookahead designs, in which block transfers stored with some degree of imprecision (e.g., using of instructions and data minimize memory accesses one term in the inverted file for a range of record and transfers, have used scratchpad memories. . . . values or a code to represent several record terms). "A major utilization [of scratchpads] has been In both cases, a portion of the unit-record file must the storage of intermediate arithmetic or logical be examined to complete the selection task." results from a computational unit." (Gluck. 1965, (CHmenson. 1966, p. 113). p. 663). 3.24b For example, "Anderson et al.. in work "Many registers, formerly implemented with sponsored by the Army Research Office and con- flip-flops, are now stored as words in scratchpad ducted at the National Bureau of Standards, are memories. Typically, such registers are utilized experimenting with a linear file consisting of basic for index words, base registers, real-time clocks fixed-field information about the entry, plus links to and similar relatively infrequent usages." (Gluck, satellite information files for a chemical retrieval 1965, p. 663). system. As part of the same effort, Marron et al., "The incremental compilation achieved by developed programs for chemical structure or sub- indirectly addressing all operands through their structure searching based on the Hayward notation. reference entries suggests a small very-high-speed Since no screening techniques, which serve to memory, functioning much like the index registers, categorize the structures, are used in the imple- to be used by all identifiers' reference entries." mented system, a linear search of the entire file is (Lock, 1965, p. 471). required. The authors recognize this hmitation for 3.27 "The key to the system is the presence an operational system." (Minker and Sable, 1967, of an Associative Memory Subsystem. The asso- p. 140). ciative memories are used to provide dynamic 3.25 "The rationale behind the inclusion of local control over processor assignments, to mechanize scratchpad memories in the B 8500 computer an automatic page turning scheme, and to provide

module encompasses . . . the need for buffering of important functions concerned with the execution four-fetches of instructions and dat'a in advance of I/O commands." (Gunderson et al., 1966, of their use, i.e., lookahead. Also important are its abstract). uses as storage for intermediate results, as an eco- 3.27a These authors comment: "The parallel nomical implementation for registers and counters, search function of associative memories requires and for the extension of the push-down stack." that comparison logic be provided at each memory (Gluck,1965,p. 663). word cell. The APP, by moderate additions to this "The storage queue contains words destined for logic, allows the contents of many cells, selected storage and their absolute address. Since the storage on the basis of their initial content, to be modified function has the lowest priority in the communica- simultaneously through a 'multiwrite' operation." tions unit, words are retained in this area until (Fuller and Bird, 1965, p. 105). 3.27b "The ILLINOIS Pattern Recognition service time is available . . . [but] data being fetched are checked against the contents of the Computer (ILLIAC III) ... is designed for storage queue by the use of the associative mem- automatic scanning and analysis of massive amounts of relatively homogeneous visual data. In particular, ory . . . Fetching of data from a main memory the is of studies at this location about to be updated by a word awaiting design an outgrowth Labora- a system capable of scanning, service in the storage queue would provide incorrect tory of computer measuring, and analyzing in excess of 10" bubble- information to the program." (Gluck, 1965, p. 666).

chamber negatives per year. . . . 3.25a "The power of the machine is achieved by "A Pattern Articulation Unit (PAU), whose duty using a small fast scratch pad memory for manipu- is to perform local preprocessing on the digitized ." lating address fields . . (Wigington, 1963, p. 707). raster, such as track thinning, gap filHng, line ele- 3.26 "In the beginning, small associative ment recognition, etc. The logical design of the all- memories will be provided as an adjunct to com- digital processor has been optimized for the idealiza- puters of conventional organization to permit fast tion of the input image to a line drawing. Nodes table look-up, facilitate the performance of certain representing end points, bends, points of intersec- classes of information retrieval applications. Later, tion are labeled in parallel by appropriate they may be employed to facilitate memory assign- programming. The abstract graph describing the ments, bus assignments and processor assignments. interconnection of labeled nodes is then extracted Still later, we may see the whole organization of as a list structure, which comprises the normal the memory changed ..." (Clippinger, 1965. output of the processing unit." (McCormick, pp. 208-209). 1963, p. 791). "Many systems applications of high-capacity, 3.28 "The interrogation, which may consist of high-speed, random-access, read-only memories a number of descriptors, each containing many bits appear to stem from table lookup operations. of information, causes an appropriate group of 63 solenoids to be driven . . . The solenoids interact plane is connected to its own detector-driver which simultaneously with all enclosed loops on all data tests the output in comparison with aU the other planes, resulting in a simultaneous voltage on data plane outputs to find that output containing the output of each data plane that is the cross- the best correlation. Alternatively, the detector- correlation between the driven input solenoids and driver can be set to test for some predetermined each individual data plane. The output of each threshold.*' (Pick and Brick, 1963, p. 245).

4. Mass Storage Considerations

4.1 "Documents will probably be always with it — especially in an open-stack library. In such us because the material in them does not easily a library it would be desirable to place a dummy or economically lend itself to machine handling. after the last bound volume with some such state- Also, for some purposes (browsing, frequent refer- ment as 'vol. to date on microfilm.' ence) documents are more easily and more "For extended reading it is certainly true that efficiently handled." (Brown et al., 1967, p. 64). print is much easier on the eyes than film, and the

4.2 "One . . . factor is the likehhood of mutila- bound volume is more convenient for notetaking. tion or theft of a given title. Until 1984, when In short, from the point of view of the user, the every home will have its own microviewing device, only advantage of film (and it is not an insubstantial no patron is as likely to remove reels as he is to advantage) is that the title is never absent at the remove issues. Every library can provide its own binder or never on a processing shelf awaiting list of titles on which the razor and the scissor receipt of a missing issue so it can go to the binder. have wreaked frequent havoc; for such titles film "Fourth is the element of cost, and here a com- would diminish the problem. puter would be necessary to put into a proper "A second element to be considered might be formula all the factors involved. University Micro- called the nature of the material — including both films takes many of them into account in stating physical aspect and content. A weekly magazine that to store a current periodical for a 30 year of the size of Business Week or the New Yorker, period by binding would cost $240.00, while to in which the advertising matter cannot be removed acquire and store the same periodical on film would will take up large amounts of shelf space in bound cost $106.65. Ford suggests that even more can form. On the other hand an abstract journal is much be saved by using a microfilm splicer to combine more convenient to use when it is bound; one must in one box on one reel three or four volumes of a turn from index in the back to abstract(s) in the periodical. He states that his library in this manner middle, and it is far easier to do this with a printed succeeded in reducing 72 boxes of film to 40 which volume than to reel and unreel continuously. A saved 'almost 45 per cent of housing space plus the periodical with many illustrations, maps, and cost of splicing.' The clerical costs of preparing diagrams, especially if they are in color, will not a volume for binding and checking the bound be as suitable for microreproduction as one with volume on its return no longer exist. The missing text only. A periodical already in hard covers issue problem mentioned above is also a factor (Horizon) or one which can be purchased in hard here. Frequently a missing issue is not noted until

cover form (American Academy of Political and it is time to send the periodical to the binder. Social Science Annals) can go directly from current In many libraries this means that the title, sans shelves to stacks without being sent to the binder. missing issue, must be set aside until a replacement "A third element in the decision is the user of arrives, which means more storage space, more the material. Students at Staten Island, for the adjustment of records — in short, more cost. Further, most part, never have seen a microfilm reader when microreproduction is used, one can, in the before coming to the college and are usually en- case of some titles, sell the superseded issues to thralled to use one. Not so their instructors. . . . a second hand dealer and thus recoup part of the represents the attitude of the professor; he prefers initial investment." (Peele, 1964, pp. 168-169). to take the periodical to his home or office rather 4.2a In the Annual Review of Information than to be tied to the machine. The same is true Science for 1966, Van Dam and Michener comment of the serial cataloger. Good cataloging requires as follows: "The improved accessibility of micro- that each back volume be inspected; it is far easier form was a significant development in 1966. There to do so at one's desk where the ALA-LC rules were three aspects of this development. First, in- are at hand than to have to go to some other area creasingly more microform was being produced and of the library where the film readers are. The distributed, by the government as well as by manu- division of location that microfilm requires may facturers. To do the research for this chapter, for also be a hardship. If one has a long run of a periodi- instance, some hundred microfiche were ordered cal in bound form and then switches to micro- from the Clearinghouse for Federal Scientific and reproduction, the title is now shelved in two places. Technical Information or through interlibrary loans, Even though the information about the switch is and were scanned on a $100 viewer. Most engineer- in the card catalog, the reader wiU often overlook ing drawings and other documentation must now

64 be supplied to NASA or to the Department of De- documentation operations using the Kodak Re- fense in aperture card form. Thomas Publishing cordak or 3M Filmac reader printers. VSMF will Company publishes product information in the form make accessible to QDRI its own extensive file of microcatalogues. Aperture cards, particularly of commercial catalogs. Also the VSMF Vendor microfiche, were the most popular medium. Some Selector questionnaire is identical in use and has minor revisions to the National Microfilm Associa- major similarities in format to the Standard Form tion's microfiche standard specifications were pub- 129 and the DD Form 558-2 employed by the Army lished." (Van Dam and Michener, 1967. p. 201). in establishment of R&D source data. The need for "In summary, then, there are three basic dupli- improved language in present Army and DOD forms cating media — film, paper and tape. As indicated, is being met by the new R&D Category Index, magnetic tape requires facilities for interpretation, coordinated by the Armed Services Procurement paper creates unmanageable storage problems, and Regulations (ASPR) Committee with all of DOD flooding, a major problem, could destroy the rec- and NASA." (Peirce and Shannon, 1967, p. 265.) ords. Microfilm has a definite edge on both, as 4.5 "We (Ford Motor Company Engineering) readers provide rapid interpretation, and flood . . . changed over to 35 mm aperture cards . . . damage can be controlled by a simple washing of About 99.9% of our engineering drawings are the microfilm reels." (Butler, 1962, p. 63). now on aperture cards and are being distributed 4.3 "Up-dating or additions and changes to the throughout the world on that basis." (R. H. De Rosa images in roUs of film used to be considered a in "Systems Magazine's Third Annual Microfilm contra-indication to the use of roll film in applica- Seminar", 1965, 35). tions where up-dating is an important requirement. Four new IBM microfilm products were intro- Because of the facilities afforded by magazines we duced in 1965: "Key to the system is the microproc- can and do seriously propose roU microfilm today essing reusable aperture card which may be used for some of these applications. And, moreover, such to project information onto screens or to obtain proposals are not at all necessarily based on film paper copies. One of the products, a copier/ splicing or other non-dark methods of adding images reproducer is a high volume machine that re- to existing film. I refer, of course, to a more limited produces up to 2,100 cards an hour from the master type of market, but my purpose is to illustrate some micro-processing card. It enables the transfer of of the many new implications for the microfilm both punched holes and microfilm images from one industry represented by the magazine concept and punched card to another in a single operation." the methods and functions this concept facilitates." (Systems 6, No. 6, 6 (June 1965).) (Brunelle, 1962, 91.) p. "A Texas architectural firm, Caudill Rowlett 4.4 "The Microstrip System, offered by Re- Scott, has developed a computerized microfilm cordak Corp., uses foot-long plastic holders for file system enabling architects to stop 'reinventing storage and retrieval of 12 in. strips of 16 micro- mm the wheel'. By converting drawings and file material film. . . . File categories are distinguished by color to IBM micro records, needed information can coding on the end tab, which contains an index by be retrieved and displayed, or reproduced, in a content. The plastic strips are also indexed length- few minutes. wise to facilitate location of the desired image." "In the past, CRS, like other large architectural (Bus. Automation 12, No. 7, 47 (July 1965).) firms, was constantly redesigning many common "The Micro-Thin Film Jacket, marketed by the construction features such as ships ladders, flash- Recordak Corp., accepts single microfilm images as ings and curbings. The only alternative was the well as strips, and permits the entire contents or even more time consuming task of retrieving a any part to be copied by direct contact printing particular drawing from the company's file of without removing the microfilm." (Systems 6, No. completed projects (bins containing all records 8, 6 (Aug. 1965).) pertaining to the completed project) — located in a "It is clearly uneconomical to put aU data into distant warehouse. Since there is no statute of the computer. The data elements established so limitations on the responsibihty of architects in far in the design are essential. In addition, the Texas, the firm retains its files permanently. In documentalist knows that sometime, somewhere, addition to the time consumed, the usual results someone will want and need additional corroborative were frustrating failure. information. Some type of microfilm appears most likely to meet the need for hard copy backup "CRS's system employs both aperture cards and documentation, storage, and retrieval for all five microfiche. Aperture cards are standard sized punch QDRI RODATA files. At this point, we are con- cards with a film frame mounted in a window on cerned only with the facility file. Now, QDRI is the card. The CRS microfiche are sheets of micro- not attempting to develop new techniques. Economy film containing up to 90 images. Projected data, requirements have dictated use of a proven com- reduced to microfilm, will be readily available in mercial system. About a year of search and dis- an area adjacent to the computer room. If the archi- cussion led to the conclusion that the VSMF tect wants to study the drawing, he drops the cartridge or cassette system should satisfy QDRI aperture card into a viewer. If he wants a hard RODATA needs. This is the same system that is copy, one can be produced in thirty seconds for in use by Chemical Abstracts and several library him to study at his desk. 65 "Under its developing program, CRS also is in that with increasing reduction ratio the film analyzing all of its design functions, coding them transport mechanisms in cameras, readers, and and entering the data into an IBM 1130 computer. printers become extremely complicated. The When completed, the computer will contain refer- greater the number of images stored within a given ences to projects by design elements. An architect area of film, the more difficult is the problem of seeking to learn how a particular problem was solved indexing, locating, and accurately registering in the past will cite the factor and the 1130 will list any single one of them. Eventually a point is reached the projects containing that particular element." at which the cost and complexity of the image- (Computers and Automation 16, No. 10, 54 (Oct. registration mechanism discourage the use of 1967).) higher reduction ratios." (RADC Summary of the 4.6 "Currently, the 4 by 6 inch microfiche has State-of-the-Art, 1967, p. 16). been standardized and popularized by both govern- "Special systems based on document reduction

ment and industry. Two fundamental system re- ratios of 60 : 1 (Kodak Minicard), 140 : 1 (AVCO

quirements are filled by the microfiche unit record. Corporation), and 200 : 1 (National Cash Register First, it fits the human hand well, a mandatory re- Company) have been successfully used for high- quirement since the carrier must operate in manual density document storage, but they require special as well as automatic systems. Second, it fits the (and expensive) equipment exclusively designed information packages well, and therefore it will for these purposes." (Ibid, p. 19). provide a least-cost file dissemination tool. This is "The Verac, constructed by the AVCO Corpora- achieved because the microfiche is usually relatively tion, 1958-1961, could store a million pages in full. Thus, using microfiche, the user can obtain reduced facsimile in a cubic foot of space and the lowest cost per page of any comparative method could find and display or reenlarge any one of them in pubhshing appHcation." (Tauber, 1966, p. 280). in seconds. It demonstrated the engineering "Microfiche makes the duplicating library con- feasibility of high-density photostorage; in economic cept possible. In this concept the basic library is a terms, however, it was a failure." (Council on file of microfiche. The user receives a duplicate Library Resources 10th Annual Report, 1966, p. 21). microfiche, either negative or positive, made while "The highly complex hardware necessary to he waits. Other optional outputs will be made avail- handle ultra-microimages and the quality of the able such as selectively producing hardcopy on a input material must be the two principal criteria reader-printer or completely producing hardcopy against which any plans to use such high reduction through the use of a step-and-repeat enlarger." ratios need to be checked. I discussed some of (Tauber, 1966, p. 281). the apparent but not real compressions of document "Indexing acquisitions on tape, however, is only systems in the ultra-micro range before the National a part of processing. In establishing an operating Microfilm Association last year and will not repeat philosophy for handling of materials, NASA was the arguments here. Project WALNUT, the CIA/ bound by the need for timeliness and the need to IBM high activity document store about which you give individual service in providing the documents have read, utilizes a net 35X reduction ratio. The themselves. The best solution seemed to be micro- proportion was reached only after having experi- fiche. A single processing results in a micromaster mented with and found impractical reduction from which any number of production copies can be ratios above 50X." (Howerton, 1962, p. 23). obtained. The production copies are readable as is, However, it is to be noted that "Republic Aviation with readily available equipment, and if the user entered the microform field because of the large

wants hardcopy, a copy can be blown back from number of technical manuals it provides with the microfiche and handed to him. We also con- its airplane weapons systems. Its equipment 4" 5" sidered the sheer bulk of files, the cost of handling records 9801 frames on a x fiche at 260 : 1 and mailing tons of paper annually, and the vast reduction ratios in a two-stage photographic process. storage space our NASA Centers and other service A 10-fiche automatic loader for the keyboard-con- points would need." (Day, 1965, p. 130). trolled X—Y positioning "Micro-vue" viewer gives 4.7 "Reductions greater than 35X have to have access to any of about 100,000 pages in less than highly specialized associated storage and retrieval 30 seconds, at a projected cost per page of .OOlc};." hardware unless these high reductions are being (Van Dam and Michener, 1967, p. 204). used to compress records for archival storage and 4.8 "The Council has this year commissioned thus not in high activity files. Perhaps we could a new investigation from the Republic Aviation consider any image reduced more than 35X as an Division of the Fairchild Hiller Corporation. This 'ultra-microimage'." (Howerton, 1962, p. 23). study will seek to identify and estimate the effect "The computer-controlled Magnavox 'Magnavue' of factors (size, contrast, optics, photosensitive and the IBM 'Cypress' 1350 photoimage retrieval materials, etc..) affecting the usefulness of microsystem are both in the $1,000,000 range and there- copy at high reduction. Film chips will be made fore are for special high-priority operations only." demonstrating optimum use of high reductions (Van Dam and Michener, 1967, p. 202). up to 300, a laboratory model of a viewer will be "Even if the system will be used under conditions constructed for evaluating such microimages such that successive regeneration of document under various conditions, and an engineering study images is not a limiting factor, there is a problem will investigate the feasibility of procuring hard

66 copy from the microimages." (Council on Library action so that proper measures may be taken to

Resources 10th Annual Report. 1966. p. 50). prevent blemish formation. . . . 4.9 "The importance of high quality in the hard "Research at XSB. at the Armed Forces Institute or soft copy output, results from the intensive use of Patholog}". and at Kodak Research Laboratories which is made of the patent documents. The docu- has led to the conclusion that biological mecha- ments are searched for several hours at a time and nisms are not directly contributorv". .AH findings the selected documents are carefully studied. thus far indicate the spot formation is a chemical A system providing difficult to read copy under process, and as usual in such cases, the probabihty these conditions would clearly and rightfully of reaction depends on many factors. These include meet strong customer resistance. The maintenance the exposure and processing conditions as weU as of high quahty throughout the system provides a various aspects of film storage." (Research on considerable challenge because of the many genera- Archival Microfilm. 1966. pp. 154-155 1. tions involved in reaching the output copy. Figure 2 4.11 These techniques were developed, for shows the steps that might be involved in the pro- example, in the Microfilm Rapid Selector as duction of hard copy in the system. The camera originally proposed by Bush and in the Eastman negative would probably not be used in the store Kodak Minicard System. For the historical back- because of format requirements and the need for ground, see Bagg and Stevens. 1961. pp. 9-36. an archival master. The output of the central 4.12 "The merged-record approach, wherein file, a third generation microform, would be an the machine-readable, coded representation of a intermediate whether used for printing, as the input document is fused together M"ith the full text to a Copyflo or Electrofax. or as a master for the document image, results in a clumsy, slow-response production of classified release microforms. The system which inflicts excessive wear on the docu- use of video printer does not ehminate these steps, ment images, and involves great inefficiency and because there is still the requirements for an inflexibihty in index searches." {\ ogel. 1962. p. 27.) archival master and because of the need for a In contrast, "physical independence of the index buffer between the central file output and the from the related photo images . . . makes it printer. Finally, in the pubhc search appHcation. convenient to: the remote user should be able to produce hard a. Search a great amount of compressed data copy from his soft copy viewer. This would be a (indexed information) efficiently, without print-out from a 4th generation • microform." scanning or physically passing over the great (Urbach. 1962. p. 159.) volume of extraneous information contained 4.10 "What are the impHcations of the Register in fuE copies of the documents indexed. for a national preservation program? One of the b. Reclassify, update and purge sections of the first effects wiW be to bring to the attention of H- index without altering the image file. brar\- administrators the need to establish policies c. Establish as many cross-indexes as may be concerning preservation of the materials in their desired. collection — all of which are deteriorating at one d. Index and refer to relevant holdings stored in rate or another. The hbrary community has not yet other forms or in other locations. fully faced up to this need. Some Ubraries are using e. Add to and purge documents from the image their only microform copies as service copies, and, file without regard to fixed categories or file in some instances, these represent the most com- locations. plete files available, or even unique and irreplace- f. Use the mechanized index in combination with able copies." (Applebaum. 1965. p. 492.1 a manned librarv'. or the mechanized image "C. S. McCamy . . . [said that indications based file without the computer-based index. on an extensive survey] point conclusively to exces- g. Replace either the index or image file with sive humidity as an important element in the pro- different units without disrupting the svstem." duction of blemishes on microfilm. Laboratory work (Ibid, p. 28.) also indicates that chlorine (as found in urban water systems) is a factor to be considered. Further re- 4.12a "Posting document numbers in a peek-a- search must be done, however, before definitive boo system involves locating, in a rectangular array conclusions are reached or any new recommenda- on an index term card, the coordinates correspond- tions are made regarding the production and storage ing to the document number and punching a hole at of archival-qualitv microfilm." (LC Info. Bull. 25. that point on each card applicable to the document. App.. 285 (19661.)' A search wiU then let light pass through the same "Scientists at the XBS Institute for Basic Stand- hole locations in several stacked cards indicating, ards are investigating aging blemishes that some- with the aid of a grid. lor. as described below% by times form on microfilm years after the films are use of the Microcite search machine) the numbers placed in storage . . . The discovery of aging of those documents indexed by all the terms selected blemishes on some microfilm has generated exten- for the search, hence the documents to be retrieved. sive research into the causes of blemishes both at "The peek-a-boo or search cards . . . used NBS and throughout the photographic industry. The are 5" x 8" x .007-. 010" polyvinyl stock, of which objective of these studies is to obtain a clear under- there is one each for everv" priman." index term standing of the physics and chemistry of the re- used . . . Output may be limited to this display 67 or the user may, by pushbutton, command readout activity and cost. Although complex centralized into an IBM punch. The basic elements required storage and retrieval systems have been advocated for provision of hard copy output have been infor many years, (2) the shortest time cycle from corporated in the machine, but this capabihty has question to answer is frequently found to be the use not as yet been used in routine operation. of decentraUzed microfiche files. Therefore, we can "In Microcite, the location of each indicated hit draw the following conclusion: when manual micro- is used as a reference point from which to position fiche files are automated, the ideal of having an and project a microfilm image of the abstract of entire file of information concerning one individual's the document, or of data that can be accommodated area of interest can be on his desk, available on within the limits of a 5 x 8 inch card in the case of simple command, and capable of easy update." (3) data retrieval systems, for visual inspection by the (Hoadley, 1966, p. 30). querist. The searcher may then (1) determine the "Microfiche report dissemination or publishing probable relevance of the projected items to his may reduce the need for remote physical access to question prior to retrieval of the full document central files simply because you literally can store text, (2) selectively sample the results to ascertain a small Hbrary in your desk." (Herbert, 1966, pp. whether his search prescription is too broad, or 34-35). narrow, and, with collateral use of an indexing "Obviously convenience of look-up on the prem- term dictionary (coding vocabulary), (3) either re- ises of his own building will revolutionize the re- formulate his search prescription in the former case searcher's work and should improve the quality of or add and/or remove index term cards in the latter. his work and prevent unwanted duplication; this, These operations are possible because the searcher after all, is the raison d'etre of the special library." becomes part of a closed feedback loop when he (Kozumplik and Lange, 1967, p. 67). is conducting a machine search; that is he receives 4.14 "GPL Division of General Precision, Inc., prompt indications of the machine output from has developed a system that allows microfilm to be which he may modify the input. In terms of up-dated and corrected without film reprocessing. "directed browsing" (where the user poses a delib- Only a standard pencil and eraser is needed to add erately broad question), a convenient tool is thus or delete up-dated information. The new system, provided. In this way the machine provides browsing PARD (for Precision Annotated Retrieval Display), capability equivalent to access to a very large stores up to 500,000 management information items number of libraries each organized in a different on 5,000 EAM-cards, automatically selects any one

basis at the whim of the user. card within six seconds, then 300m magnifies it up "Microcite is therefore a special-case system to 250 times for television viewing, while providing involving search type selection that not only retrieves continuous updating capability." (Data Proc. Mag. address locators, but is coupled with a direct 7, 44 (Feb. 1965).) facsimile output of document abstracts or of data 4.15 "Erasibility is beginning to be offered with (projected image and, if desired, hard copy). It certain microforms. For some photosensitive dyes thus uses two basic storage media: vis., film on there is a gradual loss of image density with return which the abstracts or data are copied and peek-a- to reusable state, unless the image is protected or boo cards for indexing. The process for storing developed. Another erasable medium is thermo- the abstracts or data involves a special camera plastic, in which an application of heat smoothes that records each image at the precisely correct out previous modulation and renders the plastic location in an array. At a reduction ratio of 64 to 1, ready for a new recording." (Heilprin, 1961, p. 217). 20,000 photoreductions of 5 x 8 inch cards are "Thermoplastic recording is permanent within accommodated in a 15" x 15" area on a 20" x 24" reasonable temperature ranges, and erasures (with

film sheet . . . degradation) are possible." (Gross, 1967, p. 6). "In operation of the Microcite machine for Another proposed recording-storage development search, the film matrix appropriate to the peek-a-boo reported with erasure capability is that of the set being searched is automatically selected and "Photocharge" system invented by Gaynor and mounted on a drum. The operator places selected Sewell of G.E.'s Advanced Technology Laboratories. peek-a-boo cards on an illuminated area of the This technique utilizes a photoelectronic potential machine and manipulates two control wheels to of material in special film so that light and heat position a set of cursors on each illuminated, together produce a picture consisting of minute that is, unobscured hole in turn. Turning the control deformations on the surface of the film, which can wheels cause rotation and translation of the drum be erased by reheating. (Business Automation 12, carrying the film matrix, so that as the cursor is No. 7, 50). set on each hole, a fullsize focused image of the 4.16 "One of the major accomplishments of the citation and abstract corresponding to that hole NCR PCMI process is that it not only permits in- is projected on a screen in front of the searcher." spection to occur at any step of the process but it (Adapted from Bagg and Stevens, 1961, pp. 83, also allows the operator to correct errors." (Tauber * 85—86; additional information from J. Stern, and Myers, 1962, p. 403). Sept. 1969.) 4.17 "It is probable that the single most impor- 4.13 "The principal factors that determine tant impetus to the trend of 'five' micro-image sys- - whether or not a file should be automated are file tems was the approval and subsequent pubUcation

68 in 1960 by the Department of Defense of a group of "The development of any new art usually begins specifications covering a DOD engineering data rather awkwardly and, with learning and experience, microreproduction system ... In late 1962, these it and the associated instrumentation is refined. specifications were revised and now represent a very While at the present time we must read the reflec- important standard for the entire microfilm in- tances of several fines and perhaps determine some dustry." (Hanlon et al., 1965, p. 5). optimum exposure based on the range of reflec-

"In June 1963, COSATI endorsed a . . . limited tances, there may come a time when we will be able set of microfiche standards, specifying reduction to scan the entire drawing and have an integrated ratio, image frame size, and separation between exposure which will be optimum for all the lines on frames. The Committee recommended their adop- the drawing. The objective should be to find a tion by all Federal agencies. In August of that year. method of holding line densities through proper Dr. Jerome Weisner, then President Kennedy's exposure, development and drafting techniques to Scientific Advisor, made government-wide adoption a spread of approximately 0.20 density. While we of the standards a matter of Federal pohcy and are placing the emphasis on the control of line den- urged development of further standard specifica- sities, we have not entirely ehminated the impor- tions for reading and printing the miniaturized prod- tance of background because there must be suffi- uct and for reproducing full-size copy." ("The New cient difference between densities of linework and Microfiche Standards". 1966, p. 36). background to provide adequate contrast both for "A microfiche standard for documents by 11 producing reproductions and for viewing on a reader. inches or smaller has been established by the Na- If we can hold to a maximum line density of 0.25, tional Microfilm Association (NMA). The NMA stand- including the density of the base, the 0.25 would ard identifies four acceptable microfiche sizes: 75 represent the maximum Hne density on a micro- by 125 mm, 105 by 148 mm, 5 by 8 inches, and 3.25 film. If we then establish that a minimum hne to by 7.375 inches. A reduction ratio of 18: 1 to 20: 1 background contrast difference of 0.75 is required is specified with a fixed frame size. It was an- to insure usability, the minimum desirable back- nounced that all Government agencies would stand- ground density level would be the sum of the maxi- ardize with the 105 by 148 mm size (approximately mum line density plus the desirable contrast or 4 bv 6 inches) early in 1965." (RADC Summary of 0.25 + 0.75=1.0 density. It is a procedure similar the State-of the-Art. 1967, p. 53). to this that will be employed to determine the back- "Further refinement of the standards which led ground density level, once the line density standard to the governmentwide adoption of the 4"'x 6" micro- has been established." (Frey et al., 1962, pp. fiche has resulted in the publication of Federal 110-111). Microfiche Standards, issued in September, 1965." 4.20 "National Microfilm Association members including all major equipment producers conferred (Veaner. 1966, pp. 204-205). with government representatives on September 4 "For the first time the federal government is on a and 5 and are close to agreement on basic standards: standard before too many noncompatible systems get started. Users will be able to purchase reading a. Microfiche Sizes — 75 x 125 mm (Ubrary card), equipment with assurance of total system compat- 105 mm X 6 inches, 5 inches x 8 inches, 3.25 ibihty and without fear of early obsolescence." inches x 7.37 inches (tab card). (Yerkes,1963.p. 129). b. Single frame size: 11.25 mm x 16 mm, double 4.18 "Librarians with a responsibility for the 23 mm x 16 mm with one-half mm separation. service or custody of microforms were urged to ac- c. Reduction ratios 18x to 20x variable. quire and maintain a collection of the latest revisions d. Reference point on microfiche for automation of appHcable photographic standards approved by requirements — lower left-hand corner. ." e. Layout — Eye legible title on top of microfiche the American Standards Association . . (LC Info. (along longest dimension, Bull. 25, App. 483 (1966).) micro pages running in sequence below from left to right and down- 4.19 "The need for a shift to a linework density ward from row to row." (Yerkes, 1963, 129). standard arises from the fact that the lines which p. we wish to photograph vary widely in reflectance and "The NMA Standard specification for microfiche produce lines of very wide density differences on the is being revised and wiU include almost all of the film. In fact, measurements have shown that the line agreements reached between the government and densities on negative film of engineering drawings the industry and user representatives." ("The New vary all the way from 0.08 for some ink hues to 0.60 Microfiche Standard Standards", 1966, p. 39). for light pencil lines. This wide range of hne density 4.21 "The third major objective, i.e., encouraging is the cause for most of the problems in making manufacturers to reduce costs of microfiche equip- satisfactory reproductions and providing a readable ment, will not be achieved until a high-volume image on the reader. It is with the intention of im- market is assured." (Schwelling, 1966, p. 36). proving this condition that a control over the density 4.21a For example, "standards for microfilm are of linework is needed. Preliminary studies have provided in a new pubhcation — Microfilm Norms — already indicated that a maximum line density of issued by the American Library Association. Sub- titled 0.25 with adequate background density would bring 'Recommended Standards for Libraries' . . . substantial improvement." (Frey et al., 1962, p. 105). [it includes] a list of pertinent standards of the

69 American Standards Association, the National hbraries with periodic reports on practically any- Microfilm Association, and the U.S. Government. It thing they are Hkely to buy including microform also provides target descriptions, target and page reading machines. All machines are given an im- arrangements, reduction ratios and image orienta- partial examination so that the subscriber receives tion, quality requirements for first-generation films, not only specifications about the machine but also a and a section on duplicate films." (LC Info. Bull. critical evaluation. All tests to date have been 25,552(1966).) conducted by WiUiam Hawken, formerly head of 4.22 "The first attempt to hst microfilms was the Photographic Services at the University of California Union List of Microfilms initiated in the spring of Library, Berkeley, and a recognized expert in the 1941 by the Philadelphia Bibhographic Center and field of microreproduction. Mr. Hawken has also Union Library Catalogue. The list was published in written Enlarged Prints from Library Microforms 1942 with Supplements 1-5, 1943-47, and was cum- (Library Technology Project Pubhcation No. 6), a ulated into the Union List of Microfilms, Revised, detailed study of reader-printers. Enlarged and Cumulative Edition, 1951, followed by Another important source of equipment informa- Supplements 1949-52 and 1952-55. There was then tion is the Guide to Reproduction Equipment, edited a final Cumulation, 1949-59, pubfished by Edwards by Hubbard Ballou, Head of the Photoduplication Brothers in 1961, which includes more than 52,000 Laboratory at the Columbia University Libraries. entries representing microfilm accessions reported The Guide is pubHshed by the National Microfilm by 215 hbraries in the United States and Canada Association under a grant from the Council on from July 1, 1949, through July 31, 1959. It excludes Library Resources and is now in its third edition newspapers, dissertations, and a number of special- (1965). It is probably the best reference source for ized series such as the Vatican Manuscript Codices. specifications; it does not, however, give any evalua- In 1948, another important reference source ap- tions. The current edition contains 550 pages and peared. Newspapers on Microfilm, a listing which includes descriptions of 292 machines." (Diaz, has gone through several editions and is now pub- 1967, p. 211). hshed by the Library of Congress. ''The International Micrographic Congress has

"Both the Union List of Microfilms and News- received a grant . . . [from the Council on Library papers on Microfilm limited themselves to materials Resources] to prepare and pubhsh an 'International reproduced on roll microfilm. The first Usting of Guide to Microfilm Equipment'." (LC Info. Bull. 25, pubUcations on unitized, opaque microforms (e.g. No. 10, 137 (1966).) Microcard, Microprint) was the Union List of Pub- 4.24 "Let us now turn to the problem of good lications in Opaque Microform, compiled by Eva to excellent microimage production equipment Maude Tilton and pubfished by Scarecrow Press in for problem document collections in which the material to micro- 1959. There is a 1961 supplement and a 1965 revised spectrum of best to worst be edition. These are essentially listings, in dictionary proportioned is very great. Just now are cameras order, of entries appearing in publishers' catalogs. becoming available which will accommodate such "In 1961, Microcard Editions, Inc. issued the first variance of input quality in resolution ranges vehicles. list of microform publications to include all methods commensurate with the newer storage adjustment devices of microreproduction. It is called Guide to Micro- Cameras with coupled exposure for forms in Print and appears in February of each year. to reheve the operator of the responsibiHty judging exposure characteristics of problem docu- Basically, it is a comprehensive list, in alphabetic order, of materials which are available on microfilm ments may resolve this problem with amortization and other microforms from United States publishers, of the added cost of the equipment to be written off fairly quickly in reduction of film spoilage and exclusive of theses and dissertations. It is not a will of the clientele of the service." union list but rather a list of publications offered for added good (Howerton, 1962, p. 25). sale on a regular basis." (Diaz, 1967, p. 212). "The National Register of Microform Masters 4.24a "Eastman Kodak Co. has announced

should fill an important need as it provides a central Kodachrome color microfilm for use by business, listing for the type of microform that has been most education and industry. In a recent demonstration, difficult to keep track of— that made by a library for Kodak showed examples of 16mm color microfilm its own use. Since the library may have little or no in rolls and unitized in jackets, as well as in micro- interest in selling duplicate copies, no publicity is fiche— 4 X 6 in. film-cards with rows of 16 mm ex- given and therefore the existence of the microform posures totalling up to 60 individual images." does not become widely known, leading to the type (Bus. Automation, p. 69, Dec. 1967). of situation wherein one year after Library A has 4.25 "The essential feature that is missing from completed filming a lengthy set. Library B begins the currently available machines is the ability to filming the same set because it does not know of scan directly from the pages of a bound book." Library A's activity." (Diaz, 1967, p. 213). (Schatz, 1967, p. 19). 4.23 "Perhaps the best source for information 4.26 "Data Reproduction Systems, Inglewood, and evaluation of current equipment is Library California, brought out two portable cameras, the Technology Reports, a subscription service offered DRS 8514 and DRS 1117, which could be used to by the Library Technology Program of ALA. This photograph bound volumes owing to a unique design service, which costs $100.00 per year, provides feature: an 'open bottom' which permits placement

70 of the camera directly over the pages of an open next conditional upon the results of previously book of any thickness." (Veaner. 1966. pp. 202-203). performed analyses ... A fourth objective was to 4.27 "Capable of being operated by untrained incorporate as fuUy as possible into the logic of personnel, the camera can copy single page docu- each routine whatever automatic decisions could be ments, periodicals, or books in a face-up rather made strictly on the basis of problem and data than a face-down position. An even-pressure descriptions provided by the user (e.g.. whether to system holds bound volumes precisely positioned compute a binomial sampHng distribution or to for exposure, preventing damage to the binding. approximate it with a normal distribution, depending An electronic flash source supplies uniform illumi- upon the sample size)." (Miller, 1967, pp. 1-2). nation without heat." (Houston Fearless brochure, 4.29a "The great interest that has been aroused "Books Microfilmed \^ hile You Wait by New- by the proposal of a National Data Center holds Camera-Processor". B-102-65A. 1965). great promise and great problems. At the heart of — "Houston Fearless . . . announced plans to the issue is the juxtaposition of conflicting values build a microfiche Camera-Processor expressly the protection of privacy as opposed to the need to designed for library materials . . . The opened discover new knowledge. We have (Considered one book is placed /ace up on the platens and the book possible means of partially reconciHng these con- holder moves upwards against the glass wedge. flicting values, namely, the use of partial aggregation This flattens the pages and prevents motion during which obscures individual identification. Partial exposure. Presumably the optical system is designed aggregation has the great advantage of satisfying to "open" the book page a fuU 180°. FuUy processed all but the most severe of requirements for the microfiches (called Filmcards by Houston Fearless) protection of privacy while still offering great im- are produced at a reduction of 18 to 1. Capacity of provement over the data which are currently avail- the camera is 250 precut film sheets of size 105 mm able to the research communitv." (Feige and Watts, x 148 mm. about 4" x 6". Work on this unit is being 1967, p. 16). supported by a grant from the Council on Library Many of these issues of privacy and protection Resources." (\ eaner. 1966. p. 204). will be discussed in a later report in this series. 4.28 "There have been all kinds of medical 4.30 "The first Rapid Selector was constructed studies and tests about which is better, negative at M.I.T. during the period 1938-1940 by Bush and or positive film. We can"t prove much one way or his assistants. Figure 8 is a schematic diagram of the other." ("Microfilm Experts Scan the Future'". the original Bush Rapid Selector. In this original 1967. p. 28). version, the microfilmed file consisted of document 4.29 Cf., for example, the following: "During frames accompanied by specific patterns of clear the past four years, a concerted effort has been dots used as a selection code. The code pattern made to compile a library of working computer consisted of 12 fields, each of which contained 27 programs appropriate for analyzing social science discrete indexing areas. Each area represented a

research data . . . The routines are problem- different indexing character. This pattern could be oriented rather than technique-oriented. The organi- used to identify each document by serial number, zation of the library and its documentation reflect name or title, author, or subject content indicia.

an assumption . . . that users will start with a In this machine, the code appeared on one longi- statement of their overall research objectives and tudinal half of a 35 mm film and a photographic infer from these objectives which routine or set of image of the original document was recorded on the routines would be appropriate to analyze their data. other half, as shown in Figure 8. In fact, several routines ask the user while he is "In searching, the code pattern for each document conversing with them on-Hne to select the kind of frame was projected onto a bank of photocells. The analysis he would like to perform ... It is assumed photocells were masked by a cutout that was the that social scientists will tabulate or cross-tabulate exact physical image of the selection criterion code their data in ways which will facilitate a test of pattern. Recognition would therefore occur when- their research hypotheses. The tables so created ever match resulted in all photocells being illumi- will then contain direct observations or category nated. Such a match fired a strobotron that projected frequencies arranged by variables, by categories, the selected frame (document-record half) onto a or by sub-samples, whichever is appropriate to section of hitherto unexposed 16 mm microfilm testing the research hypothesis previously formu- advanced to proper position, thereby allowing a lated. The routines, in turn, reference and discuss replica recording to be made. Thus, the result of these data according to the structure previously finding a desired documentary item was to copy it

set down by the user . . . Once the routine has when it was appropriately positioned in the selection- been initiated, detailed instructions are given by mechanism. This preliminary Bush Rapid Selector the routine itself concerning its purpose and scope, was used experimentally by the U.S. Department of its restrictions, the proper way to enter data (with the rVavy during the early years of World W ar II and

error diagnostics if errors occur), and the proper subsequently was returned to M.I.T. . where it is

way to interpret results . . . Analytical flexibility now stored." (Bagg and Stevens. 1961. pp. 19—20). is provided by frequent choice points programmed 4.31 "Randomatic Data Systems of Trenton. into the routines themselves such that the user N.J., has introduced a completely new fine of may decide on-Une what kind of analysis to perform random card file equipment featuring a 2-second

71 automatic retrieval of randomly filed cards. A unique create one bit, the vacuum temperature of the laser concept of coding ordinary cards, by notching them image if 3 x 10^ °K produces a vapor pressure of along the bottom edge in such a way as not to inter- 1000 atmospheres within the bit volume, leaving fere with their use in most other card processing an ellipsoidal hole in the Unidensity layer . . equipment, is used. Both notching and selection are (Becker, 1966, p. 711). performed on a simple 10-button electrical key- 4.31c They conclude: "While it will take con- board." (Data Proc. Mag. 7, No. 11 (1965).) siderably more work to make systems of this type 4.31a "The Microstrip System, offered by that are operationally foolproof, experience has Recordak Corp., uses foot-long plastic holders for indicated that the principles of operation are sound storage and retrieval of 12 in. strips of 16 mm and that by using this type of system, much of the microfilm . . . File categories are distinguished high-storage potential of fine-grained photographic by color coding on the end tab, which contains an materials can eventually be realized." (Lamberts index by content. The plastic strips are also indexed and Higgins, 1966, p. 733). lengthwise to facilitate location of the desired 4.32 "Actuators may become very important image." (Business Automation 12, No. 7, July 1965). items of output equipment for the information re- 4.31 "An information bit is represented by a trieval field in the near future. When combined with diffraction-limited 'hole' within the Unidensity film digital to analog converters, actuators can enable layer. It results from the evaporation of the Uni- an automatic computer to control the operations density medium by means of imaging the aperture of machinery and processes. To date, they have of a laser to a three-dimensional ellipsoid of revolu- generally been applied to industrial use. However, tion (Debye ellipsoid). During the time required to they may become quite powerful in the future in

72 linking digital information retrieval systems to IHS classifies, indexes, microfilms and distributes graphic document retrieval systems. It is not the data. The information is indexed by multiple difficult to conceive of a literature search being terms which describe the parts, materials and sub- done by traditional computer techniques with the systems and manufacturers needs. The film car- results of the search being used to produce micro- tridges and VSMF index are updated regularly and form copies of documents from a mechanical graphic airmailed to subscribers to the system. Data is retrieval system, the necessary hnkage being located on a given product in about half a minute. provided by a computer-driven actuator." (Austin, At the moment, VSMF essentially serves product 1966. p. 243). design, plant engineering, and purchasing groups 4.33 "Data within a given magazine may not in defense, aerospace, commercial and industrial

be sequentially ordered by any criteria. Serial manufacturing and support organizations. . . .

numbers may be entered on each page of the data VSMF . . . consists of a copyrighted product index by various numbering techniques — including auto- which lists products by type; a microfilm cartridge matic numbering coincidental with feed input at the library of technical data; a full-size viewer, which microfilmer. Alternatively, we may place a number- enables a user to locate required data, and an auto- ing counter in the photographic field and sequentially matic print copier." (Christian, 1966, pp. 12, 15). number images that way. During filming, we may 4.35 "For 16 mm film only, 3M offers another enter also what call an 'image counting we mark' reader/printer, the Filmac 4()0 . . . With a car- for each image. The marks set up the possibihty tridge system designed to compete with the

for entirely automatic finding of images. . . . This Recordak Lodestar." (Veaner, 1966, p. 202). is the Lodestar Reader with an automatic image 4.36 "The most successful of the systems built finding device coupled to it. There are a variety around roll microfilm is the Recordak MIRACODE of quite practical ways that the operator may know which combines the search function with document the image frame numbers desired. She enters the retrieval. It uses 100-foot cartridges of 16 mm micro- number any number up to 10,000 on the key- — — film. Each cartridge contains 2000 images, any one board . . . Then she depresses the 'go' button of which can be retrieved by attribute(s) in 10 and the correct image appears on the screen in an seconds and either viewed or copied." (Berul, average of five to ten seconds — depending on the 1968, p. 30). location of desired image within a particular roll of "The Microstrip System . . . Foot-long strips of

film . . . 16 mm microfilm are inserted into grooved plastic "Before leaving the subject of magnetic tape and sticks after a small plastic button is first fastened to computers for the storage, and retrieval of infor- one end of the strip. The plastic stick is inserted mation, a few comments may be appropriate. Up to into the reader and a pointer which engages the 6 full 2400' reels of magnetic tape are required to plastic button is pushed along the stick. This allows store the data which can be recorded in human the reader to scan the contents of each strip on the language and stored in a single 100' microfilm screen. The system is said to be designed for quick magazine. Magazine High Speed Retrieval of desired look-ups and 'frequently changing' files." (Veaner, information may be accomplished in a matter of 1966, p. 202). See also note 4.31a. seconds and at a very low relative cost. Compare 4.37 "A new records retrieval system called for a moment the paper print-out volume repre- is available from Remington Office sented by approximately 2500 8^" x 11" pages with Remstar now With Remstar, closed circuit television a magazine of microfilm having the same capacity. Systems. with automated records retrieval units These and other cost and systems advantages are is combined not only locate records, but to transmit them made possible through the use of film magazines." to as well. (BruneUe. 1962, p. 87,91). 4.34 "The VSMF System, a copyrighted infor- "With this system, the files are centralized and mation retrieval system owned by Information housed in Remington's electro-mechanical units. Handling Service, Englewood, Colorado, or IHS, After retrieval, the record is transmitted on a has three main elements — microfilmed data in Remstar Transmitter to a Remstar Monitor which cartridges or cassettes, a printed index for locating can be placed wherever the record is needed. After groups of desired data, and the microfilm reader viewing the record is released, or, if needed, a copy Printer. variety of printer. As the microfilm is scanned rapidly a heavy can be made on the Remstar A black line appears on the reader-printer screen, automated file equipment for the Remstar System is traveling up the face of the screen, which has a available to meet the needs of any size business." (Feb., 1966).) ruled index beside it. As the index line approaches (Systems 7, No. 2,6 the desired number group on the ruling, the scan 4.38 "As many as six million records can be rate is slowed until the operator is able to stop at stored in 100 square feet of floor space with the the desired number. Skilled searches can go from Randtriever system . . . Any one . . . can be fast scan to the desired number almost instantane- retrieved by pushing a button on the input keyboard ously." (Peirce and Shannon, 1967, p. 266). or by inserting a punched card into the request "More than 1.5,000 suppliers are supplying data reader. Electronic memory banks in the system take for VSMF files, including scientific and technical over to retrieve and deliver the request record." data on almost every product used in industry. (Systems 8, No. 2, 6 (1967).

73 4.39 "The Selectriever, introduced by Mosler and associated documentation. The computer con- Safe Company of New York in 1965, provides for trolled DARE system provides automatic collection, 6-second needle-sort retrieval of any one of up to storage, retrieval, and preparation of punched Diazo 200,000 aperture, microfiche or tab cards, which copy card outputs from a rapid-access file with a are stored in 100-card cartridges. This wholly capacity of 750,000 microfilm images. On a random automatic unit . . . can also be interfaced with a basis, the equipment provides an average access computer, hardcopy printer, remote displays, or time of 50 seconds. In the normal sequential 'batch facsimile transmission system." (Van Dam and processing' mode, a file of 750,000 microfilm images Michener, 1967,p. 202). can be processed in approximately 7 hours. The 4.40 "The difficulty of reel film for updating daily sequential processing includes, on the average, might be overcome by the use of microfiche as in the output of 5,000 Diazo copy cards, the input of the Houston-Fearless CARD (Compact Automatic 2,000 new Magnavue film chips to the system, and Retrieval Display) system. Such a system could the removal of 1,800 Magnavue film chips from the also be computer manipulated, affording faster system. scanning access and a more dynamic interrogative "Magnavue film chips, held in 2,500-chip maga- process." (Taylor, 1967, p. 31). zines, are 35 mm wide by 3" long and have a mylar

"Houston Fearless Corporation . . . announced base. The film chip has a high-resolution image for delivery in 1967 a desk-top microform retrieval portion for the storage of a standard 35-mm micro- unit and reader, with any of 73,500 (750 fiche with film image and a coded portion for the storage of 98 images each) accessible in less than 5 seconds. 80 alphanumeric characters that identify the asso-

The unit . . . will be available with an optional ciated microfilm image. The system also can use hardcopy printer or with a teletype or CRT for on- magnetic chips of the same size to store additional line hookups. Edge-notched identification numbers information associated with a given document." are stored on the metal frame surrounding the micro- (Van Dam and Michener, 1967, pp. 202-203). fiche, and the search request for a random page, 4.42 "Alder Electronic & Impulse Recording or a next page, is entered via keyboard, paper tape Equipment Co. is marketing a 'dial-a-document' or computer, thus maintaining file integrity." (Van system for retrieval by telephone of full size docu- Dam and Michener, 1967, p. 202). ments stored in a centralized microfilm center. "Houston Fearless demonstrated a breadboard "Called the Alden/Miracode System, the equip- model of CARD (Compact Automatic Retrieval ment is a combination of Eastman Kodak Co.'s Device), a desk-top unit capable of displaying within Miracode microfilm retrieval system and an Alden 4 seconds any one of 10,000 pages of filmed data. Alpur-Fax Facsimile Scanner which scans, onto The recording medium is standard 4" x 6" micro- high-speed magnetic tape, a facsimile picture of the fiches, mounted in metal holders housed in maga- original document. The tape then is used to transmit zines holding from 128 to 1,024 microfiches. Each the picture over standard telephone circuits to metal holder is edge-notched for retrieval purposes subscribers." (Systems 8, No. 1, 7 (Jan. 1967).) and is magnetically suspended so that microfiches 4.43 T. C. Bagg, private communication, Feb. do not scrape against each other as they are with- 13, 1968. drawn and replaced." (Veaner, 1966, p. 208). 4.44 "Ampex Corp.'s new Videofile document "The Compact Automatic Retrieval Display filing and retrieval system, a completely automated (CARD) system is produced by the Houston Fearless micro-filing system, stores documents as recordings Corp. Selling at under $5,000, it has a capacity of on magnetic video tape and provides fast, automatic 73,500 pages and a response time of less than five access and unusual flexibility in file organization." seconds. Document address is entered by keyboard, (Data Proc. Mag. 7, No. 11, 56 (1965).) however, it could be adapted to accept paper tape, "The Ampex Videofile system is now capable of or signals directly from a computer." (Berul, 1968, selectively recording a single page at any predeter- p. 30). mined spot on magnetic tape ... It is understood 4.41 "Magnavox defivered to the Army Missile that at least one major university library is consider- Command its DARE, a high speed, unitized system ing the Videofile as the next step beyond the for storing and retrieving engineering drawings. computer-produced printed book catalog." (Veaner, The DARE (Documentation Automated Retrieval 1966, p. 207). Equipment) is equipped with several hundred "Ampex's Videofile system, used in a NASA 3" magazines, each holding 3,000 35 mm x chips engineering fibrary, is particularly suited for real- containing a drawing plus coded data. A chip is time retrieval of documents with a very short life accessed directly, and copied onto an expendable span. Indexing can be done numerically or alpha- aperture card within one minute. Besides rapid betically using 18 numeric or 12 alphanumeric retrieval, the main advantage of the system is that characters stored contiguously to images on (mag- the master copy need never leave the main file." netic) video tape. Access to a document can be (Veaner, 1966, p. 207). remote, with hard- or soft-copy output, and the "The Magnavue System is utiHzed by the U.S. average access time ranges from 11 seconds for the Army Missile Command in their DARE program for 5-inch diameter, 10,800-page-capacity reel, to 115 the processing of a large file of engineering drawings seconds for the 14-inch, 259,000-page-capacity reel.

74 One page is stored on about a third of a linear inch second announcement describes improvements in of standard 2-ineh wide tape." (Van Dam and Photoplastic Recording, a process first brought out Michener, 1967, p. 203). by General Electric in 1963. Both processes are 4.45 "At Republic Aviation, we have determined dry and produce high resolution images almost that many users want the ultrahigh density that a instantaneously; both require further development system of our type offers ... A test program which before becoming commercially available." (Veaner, will be starting shortly will have an ultimate require- 1966, p. 208). ment for updating 50,000 pages of repair parts 4.48 "IBM's 'Cypress' system ('Son of hstings every three weeks and distributing the "Walnut" '), announced (and withdrawn) in 1966, ultrafiche to 3,000 centers. For a program Hke that, has storage facilities for half a million images on ultrahigh density is a must. A portable device that Mylar-based chips. For retrieval, a cell (cartridge) can give the customer parts and pricing information of 35mm x 70mm chips is located under computer within a maximum of 30 seconds from a 50,000 page control and is transported pneumatically to a internally stored library is a real plus ... It is a copying station, where the selected chip is copied truly automated random address device that onto an aperture card, which is in turn, transported displays the image full scale after it has been for subsequent viewing or reproduction. Total reduced 260 diameters. Its trade name is Micro- retrieval time for this random access system is Vue ... In order to shoot down original material between 4 and 6 seconds." (Van Dam and Michener, 260:1 and then bring copy back 260 times, auto- 1967, p. 203). matically screens out much existing poor quality For a discussion of the earlier "Walnut" system, microfilm material. As for the hardware, auto- see Vogel, 1962. For example: "The image converter matically handhng very high density information incorporates eight exclusion switches which repre- is not achieved at low cost. Since Micro-Vue is sent eight document classifications, each assigned basically a digital logic driven platform, moving a numeric group code. The classification codes, in row/column coordinates under high quality punched in the input control card, cause the input microscope optics, dictates that the price of the converter either to skip over these images or to viewer will never get down remotely into the $100 transfer them, as desired. This permits grouping range . . . The process which Fairchild Hiller is like documents in a given cell. In addition, an willing to sell to users allows the laying down of X-punch in the card provides for excluding poor 10,000 images on a glass master within a one-day quality microfilm images." (Vogel, 1962, p. 37). period with an accuracy of ± five microns on frame 4.48a "Data is stored on a 1.38 x 2.75-inch silver centers." ("Microfilm Experts Scan the Future", hahde photographic film chip. Thirty-two chips, 1967, p. 29). containing a total of approximately 150 million data bits, 4.45a "Ford Motor Co. 's Autolite-Ford Parts Div. are stored in a small plastic cell. . . . The has signed a contract with the National Cash first cell contains 2250 cells and houses the pneu- Register Co. to transfer parts catalogs to a micro- matic blowers for cell transport. Additional files image dissemination and retrieval system. containing 4500 cells each may be added to increase "The parts catalog system, based on NCR's total on-line capacity. A manual entry station on PCMI process, provides for filming 14,000 catalog one or more of the files permits entry and removal pages covering automobile and truck parts. With of cells from the system. Any cell can be accessed the PCMI process, 2,560 pages can be recorded and delivered through the pneumatic tube system on a single 4 x 6-in. transparency; the 14,000 pages to a reader in less than 3 seconds." (Kuehler and can be put on seven transparencies. Kerby, 1966, p. 735). "Initial distribution to dealers will require up to 4.49 "Magnetic cores have remained the most 70,000 transparencies, with a set of one to seven widely used main memory element and current for each dealer. Readers will be leased to Ford by research is developing increasingly smaller cores, NCR. After original conversion, NCR will provide making possible faster, more compact memories. up to 500,000 additional transparencies a year IBM has recently developed a core with an outer covering catalog revisions." (Business Automation diameter of only 0.0075 inches, small enough to fit 15, No. 3, 88 (Mar. 1968).) inside the center hole of most currently used cores. 4.46 "A novel technique devised by Aeroflex Also under development are new organizations of Laboratories, Plainview, N.Y., permits recording cores within large memories. Innovations in core up to 500 different pictures on a single negative. shapes have been introduced; for instance, the Film is processed in the usual way but is read on "bi-axial" core is a rectangular block with two holes a special viewer which allows the user to select through it that can be read in about 100 nano- any 'page' at random." (Veaner, 1966, p. 207). seconds. Magnetic core memory elements in general 4.47 "General Electric's Advanced Development will be in use for at least the next few years because

Laboratories . . . announced two other develop- of the large investment in existing and developing ments. One is Photocharge, a new photographic techniques." (Van Dam and Michener, 1967, p. 207). recording process. Unlike other electrophotographic "A new low-cost mass core memory has been processes such as those using selenium, or coated announced by the Ferroxcube Corp. of Saugerties, paper, Photocharge material requires no external New York . . . The new memories, dubbed 'Megabit', charging to activate the recording process. The are offered in modules of five million core capac-

75 366-108 O - 70 - 6 ities . . . This 5-million-bit size can be expanded Information is retrieved either through fixed whenever desired by adding on additional 5 million address or record-content search. According to the bit modules. For current systems, according to company, search by record-content is their exclusive Ferroxcude, Megabit 53.12 can cut sorting and technique that permits any field to be used as an merging time by one third thereby releasing vital access key. Only the nature, and not the location, computation time and extending system efficiency." of data sought need be known. This feature is said ("New Mass Core Memory", 1965, p. 40). to eliminate costly flagging and table look-up pro-

4.50 "Bryant Computer Products . . . has grams, saves central-memory space, and permits announced the availability of a new online, real-time simultaneous off-line search. Librascope Group, mass storage disk file system, the Model 2A series General Precision Inc., Glendale, Cal." (Computer 4000. capable of storing 3.8 billion bits of data with Design 5, No. 2, 66 (Feb. 1966).) an average access time of 100 msec." (Commun. "The [Control Data] 6608 Disk File system ACM 9, 404 (1966).) contains a double-actuator 72-disk unit with a "The disks allow more than 2.5 million 16-bit capacity over a billion bits. The transfer rate is words or 5 million 8-bit characters of on-line storage 1,680,000 characters per second. Minimum access

simultaneously." (Data Proc. Mag. 9, No. 6, 8 time is 34 msec. , maximum, 100 msec. Over 5,250,000 (1967).) characters are available at one actuator position. "A large capacity disk storage subsystem The two actuators operate independently to provide (DS-25) storing up to 800 million characters. 64 simultaneous seeking and read/write operations." simultaneous record 'seek' operations are combined (Commun. ACM 8, 723 (1965).) with four concurrent data transfer operations in "Sperry Rand's Univac Div. has introduced the each subsystem." (Commun. ACM 8, 343 (1965).) 8410 Disc File for use with its 9000 series computers. "A series of large disc memories features storage The 8410 provides removable direct access storage capacities up to 6.4 billion bits on a single trunk of 3.2 to 12.8 milhon bytes, or 6.4 to 25.6 million line and a data-transfer rate of a billion bits per digits, in packed decimal format. The system in- second. Basic 6-disc configuration has a storage cludes a master dual disc drive expandable one capacity of 400 million bits, a data transfer rate up drive at a time to a total of eight drives. Each drive to 150 million bits per second, and an average access holds a reversible disc cartridge with two storage time of 35 milliseconds. By combining 16 of these surfaces, one of which is on-Une. Each disc drive files on a single trunk line, an on-line storage capac- can access 10,000 160-byte records plus an 8,000- ity of 6.4 billion bits can be achieved. With special byte fast access track. Average time to locate and electronics, a data transfer rate of a billion bits per read a record on a random basis is approximately second is available. The new memories, designated 135 milliseconds. A high speed buffer permits all the Librafile 4800 series, also feature a fixed-head disc reading, writing, checking or searching to be per track, two methods of search and retrieval, and performed simultaneously with 9200/9300 process- retractable head plates. The 4800's rotating memory ing and peripheral operations." (Business Auto- element is composed of six aluminum discs coated mation 15, No. 3,68 (Mar. 1968).) with a cobalt alloy. The discs measure 48 inches in "Caelus Memories, Inc., has announced that its diameter and i-inch in thickness. Mounted on a six-high magnetic disc {^cks are now in production. common shaft, they rotate at 900 rpm. Data is stored Each unit is designed to function with IBM 1311, on both disc faces. Information is retrieved either IBM 2311 and compatible equipment. The 10 through fixed-address or record-content search. coated surfaces in the CM VI disc pack each has Search by record-content is said to be the company's 203 tracks of information, capable of storing up to exclusive technique for permitting any field to be 14.5 milhon bits. Access time is in the 75-85 milli- used as an access key. Only the nature, and not the second range. The pack rotates at either 1,500 or locations, of data sought need be known. This 2,400 revolutions per minute. The units, weighing feature is said to eliminate costly flagging and table 10 pounds per pack, are shipped in sculptured and look-up programs, saves central-memory space, and ribbed styrofoam containers." (Business Auto- permits simultaneous off-line search. Average mation 15, No. 3, 80 (Mar. 1968).) access time is 35 milliseconds. Librascope Group, "A new, low-cost, direct access storage system General Precision, Glendale, Cal." (Computer designed for the UNIVAC 9000 series computers Design 4, No. 11,83 (Nov. 1965).) has been announced by Sperry Rand's UNIVAC "New disc file memory provides a storage capacity Division, Philadelphia, Pa. The new system, known of 200 million bits, a data-transfer rate of up to 42 as the UNIVAC 8410 Disc File, makes it possible million bits-per-second. and an average access time for users to bridge the costly gap between punched of 17 milliseconds. Called the Librafile 3800, the card and direct access processing. Suitable for system was designed for use in large scientific field-installation on any UNIVAC 9200/9300 System, computing centers, in message-switching centers, the 8410 Disc File provides removable direct access in military command-and-control installations, and storage of 3.2 to 12.8 milhon bytes, or 6.4 to 25.6 in intelhgence data processing systems. The mass million digits in packed decimal format. memory is available in a basic 6-disc configuration "A basic 8410 System includes a master dual which features a fixed-head per track, two methods disc drive expandable one drive at a time to a total of search and retrieval, and retractable head plates. of eight drives. Each drive holds a reversible disc 76 cartridge with two storage surfaces — one of which and the IBM 2314 Direct Storage Facihty. The is on-line. By interchanging disc cartridges, un- latter represents a significant development since, limited storage is provided for applications which with 1600 million bits of information, it is com- require serial processing. parable to some of the large fixed disc in capacity, "Each disc drive can access 10,000 160-byte cost per bit, and average access time for a single records plus an 8000 byte fast access track called piece of information (90 milliseconds); however, the Fastband. All disc drives contain a fixed head since there are eight independent disc packs, each for reading and writing on the Fastband and a with its own accessing arm, eight different pieces of movable arm with two heads for the remainder of information can be sought simultaneously; average the disc surface. The average time required to locate access time is reduced from 90 to 40 milliseconds. A and read a record on a random basis is approxi- comprehensive comparison of bulk storage devices mately 135 milliseconds. Fastbands can be used for appears in Craver." (Van Dam and Michener, 1967, program storage, index tables, or storage of fre- p. 206). quently accessed data files. 4.51 "[UNIVAC] Fastrand II Mass Storage- 129, "Another feature of the UNIVAC 8410 Disc 761, 280 alphanumeric characters per unit; up to File is a high-speed buffer permitting all disc 8 units per computer channel; 92 milliseconds aver- reading, writing, checking, or searching to be age access time; sector addressable." (Computer performed simultaneously with 9200/9300 process- News 9, No. 7, 7 (1965).) ing and peripheral operations." (Computers and "Replacing the model 7320 for use with the Automation 17, No. 2, 46 (Feb. 1968).) 360/40, 50, 65, and 75 is the model 2303, which can "A high-capacity random access DISCflLE store almost 4 million bytes accessible in 8.6 milli- memory system designed specifically for time shar- seconds. Transfer rate is 312.000 bytes/second. ing applications has been announced by Data Capacity is five times that of the older unit, and Products Corp., Culver City, Calif. The Model speed is doubled. IBM DP DIV., White Plains, 5085 time/share DISCflLE stores over 5 billion N.Y." (Datamation 12, No. 3,81 (Mar. 1966).) bits, and has lower access times and greater "A magnetic drum (MD-30) which provides operational flexibility than has been previously high-speed storage and retrieval of frequently used available to the industry. data and instructions. Transfer rate is 370,000 "The modular concept of the new DISCflLE characters per second. Average access time is permits preventive and corrective maintenance to 8.5 msec, and it stores up to 12 million characters." be performed upon individual components while (Commun. ACM 8, 343 (1965).) remaining modules are operating on-line with the "The Honeywell random access drum file pro- computer. Each positioner module can be with- vides a medium for extensive file storage. One to drawn, and individual heads, arms, or associated eight drums can be connected to a series 200 system, electronics can be serviced without shutting down the storage capacity of each drum being 2,621,440 the system. Preventive maintenance devices are characters which may be divided into 20,480 records, incorporated so that failures can be anticipated and each of 128 six-bit characters. Average access time avoided. to data stored on the drum is 27.5 milliseconds and "The system includes eight independent posi- information is transferred to and from the drum at tioner modules accessing 32 discs. Each positioner an average rate of 102,000 characters a second." .", module carries 32 heads which access data on the ("Five Compatible Computers . . 1965, p. 289). eight surfaces of a four-disc group. Each positioner 4.52 "The RCA 3488 is a mass storage and re- is independently addressable and electrically de- trieval unit for on-line operation. The data is stored tented, so that all eight positioners can be simul- on 16" by 4|" magnetic cards which are removed taneously placed in motion, effectively overlapping from their magazines for read/write operations. seek delays. An option is available to operate up to In one unit there are eight (8) magazines with a total 10'' eight heads in parallel from each positioner module. storage capacity of some 0.34 x alphanumeric "Random average position time is 60 milliseconds; characters. In certain arrangements, the total capac- 10-'* track-to-track time is 15 milliseconds; full-stroke ity can be as high as 5.450 x characters. The time is 100 milliseconds. These values include time average access time is 350 milliseconds with a data required to move the positioner and completely transfer rate of 80,000 characters per second." settle on track; reading or writing can then proceed (Bones and Kros, 1966, p. 60.) immediately. "A data cell is divided into 20 sub-cells each I containing ten strips of magnetic tape 13 inches j "Control units wiD be available that are compat- or re- 1 ible with existing systems using either fixed or long by 2 inches wide. Data are read from, of a rotary positioning , variable length records. These provide address corded on, a strip by means system which aligns a selected group of ten tapes ; processing, data formatting, comprehensive error beneath a read/write head. Access to a specific , checking and off-line testing capabilities. Thus the

- 5085 directly interfaces with existing hardware and record takes between 95 and 600 milliseconds software." (Computers and Automation 17, No. 5, depending on the position of the required strip and

61 (May 1968).) the arrangement of data in the file . . . "Two removable-medium devices that have been "The 2321 data cell drive has a storage capacity of recently marketed are the Anelex 80 series disc 400 million bytes ... up to eight 2321 units may 77 be connected to a computer, via a storage control card, about the size of an 80-column punched card, to characters each. unit, providing an overall storage capacity of more which can hold up 117,500 ] than three thousand million alphanumeric char- These strips are housed in lightweight cartridges acters." ("IBM System/360", 1965, p. 300). each of which contains 512 magnetic strips ... * "A magnetic strip mass storage unit (MS-40): Data are recorded on the magnetic strips in tracks Each unit can hold more than 532 million characters, in a serial-by-bit format. Each strip contains a and eight of those units may be used together as a series of notches and when a cartridge is loaded subsystem. This provides random access to more into a unit a series of selection rods are auto- , than 4 billion characters." (Commun. ACM 8, matically inserted through the notches. The selec- 343 (1965).) tion rods are actuated and a specific magnetic ' "A CRAM control unit interfaces up to 8 CRAM strip is retrieved by a single program command." .", units to a single position of a Century 100 or 200 ("Five Compatible Computers . . 1965, p. 289). I/O trunk. Each CRAM unit stores up to 124 miUion 4.53 "This electronic system makes possible bytes or 248 million packed decimal digits in a high-speed and high-density storage (and subse- ; single changeable cartridge. The cartridge holds quent retrieval of digital information on recording

384 cards made of Mylar with a magnetic oxide media such as thin metallic foils or semi-conductor ; coating. One selected card at a time is dropped from films). the on-line cartridge and vs^rapped around a revolving "The heart of the system is an electron beam tube f drum; this takes about 125 milliseconds. Then any which produces a spot size of five microns at the or all of the data on the card can be read and/or recording surface by magnetically focusing the written. Data is transferred at the rate of 71,250 beam. Electrostatic deflection is employed to posi- characters per second." (Hillegass, 1968, p. 49). tion the beam wherever required. High-speed j

"Take the face of a king-size cigarette package, accessibility for read-out operations can be ac- 1 coat it on both sides with a magnetic film, and you complished by special circuitry. The device is a !

have a 'file-card' with a potential for storing over rugged and completely demountable unit." ! 100 times as much computer information as the (McDonald et al., 1964, abstract). conventional punched card. Based on a new concept 4.54 "A large scale memory system, which uses f proposed by the United States Army Electronics film chips contained in plastic cells to store more : Command, Fort Monmouth, a new technique in than a trillion bits of digital information is being unit record information storage developed by Sperry developed for . . . AEC [by IBM] . . . Information ! Rand Corporation's UNIVAC Division (Philadelphia, will be recorded by means of an electron beam which Pa.), is almost that simple. will 'write' digital data on the 35 mm x 70 mm film f

"The new technique employs an all-fluidic (air) chips in the form of microscopic black and white ; transport and a pocket-sized card with magnetic coded spots. recording surfaces. Instead of the familiar 7x3 inch "When a piece of information must be retrieved paper punched card, the new unit record in the from, or stored in, the mass storage systems,

' experimental system measures 2V2 x 4 inches and the cells move automatically, under program

' is made of a plastic material overcoated on both control — to photoelectric reading and writing * sides with a magnetic film on which digital data is stations. Data is transferred on-line at high speed recorded by magnetic techniques. between the storage systems and the central "The plastic magnetic record can store 1,000 computer." (Bus. Automation 12, No. 8, 62 (1965).) '; characters — more than twelve times as much "The triUion-bit IBM 1350 storage device, an ] 7" 3" — ' information as a standard x punched card offshoot of the 'Cypress' system . . . uses 35 mm x using only 12.5% of its magnetic surface for machine- 70 mm silver haHde film 'chips'. A total of 4.5 ^ readable data. (The rest of the card is used to million bits are prerecorded on each chip by an the feasibility of incorporating printed, electron beam. For readout, a plastic cell containing demonstrate ; handwritten and pictorial information.) Other 32 film chips is transported to a selector, which picks benefits of the plastic card include a more rugged the proper chip from among the 32; average access '

construction, a reusable record, and considerable time to any of the 10'- bits is 6 seconds. After a chip ' better resistance to heat and humidity. is positioned, information is read using a flying-spot '

"The new method of handling unit records takes CRT scanner. Two IBM 1350 units are scheduled "i full advantage of the new but rapidly developing for mid-1967 delivery to the Atomic Energy Commis- f

fluidics technology. Instead of electrons — the sion at Livermore and at Berkeley for use with ' controlling medium in electronics — fluidics uses bubble chamber data." (Van Dam and Michener, , air pressure to perform control functions. Simply 1967, p. 205). ' stated, fluidics is the use of a small, easily manipu- 4.55 "Another scheme for storing digital in- lated control stream, such as an air current, to formation optically is the UNICON system, under control the action of a much larger stream of the development at Precision Instrument Company. same medium. The source of the low pressure This system uses a laser to write 0.7-micron- air flow is a simple rehable device." (Computers diameter holes in the pigment of a film. Informa- and Automation 1 7, No. 5, 66 (May 1968).) tion is organized in records of at most a million "The basic storage medium of these [Honeywell bits; each record is in a 4-micron track extending

mass memory] units comprises strips of magnetic about a meter along the film . . . Each record is

1 identified by information stored next to the begin- high-speed, high-reliability devices and media with ning of that record, in an additional track at the edge no moving parts." (Hobbs, 1966, p. 38.) of the film. Readout of a particular record involves "The goals for one development program of this scanning the identifier track for the proper code and type are 4 million characters per [solid-state then scanning the track with a laser weaker than storage] module, read-write rates in the order of that used for writing. It is predicted, on the basis 2 or 3 million characters per second, and costs of of an experimental working model, that one approximately 0.015 cent per character for off-line UNICON device with 35 mm film could store a storage. A further advantage that would be offered trillion bits on 528 feet of film, with an average by this particular approach is random access (in access time to a record of 13 seconds." (Van Dam 1 /a sec) to any block of data within a storage module ." and Michener, 1967, p. 205). on the read-write unit . . (Hobbs, 1966, p. 38.) 4.56 "At least two different programs are under- "The serial data processing applications of the way to develop soUd-state storage modules that future could well use removable-medium, mass could be plugged into read-write electronics ... If storage devices in place of tape units. On-line, this proves feasible and economical, the input/ random-access applications could be handled during output and off-Une storage functions presently the day and serial processing at night in a typical provided by magnetic tape could be provided by system." (Bonn, 1966, p. 1866).

5. Output and Post-Processing Requirements

5.1a "A basic principle of systems design states Then we note also the following: "General- that no system is better than its final products. It purpose displays and, to a lesser extent, alpha- therefore follows that output requirements establish numeric consoles can provide an author/editor a framework upon which all other elements of sys- with implementation of a file organization such as tems design must concentrate. Output products that first described by Bush and more recently by must be specified in terms of both content and form. Nelson as "Hypertext": "the combination of natural Unfortunately, too many automated systems have language text with the computer's capabilities for not given sufficient consideration to the latter, and interactive, branching or dynamic display ... a although the system may be able to produce the nonlinear text . . . which cannot be printed .". necessary items of information, these are frequently conveniently ... on a conventional page . . presented in such a way as to be either extremely Englebart's extensive work on computer-aided difficult or in some cases virtually impossible to use. text manipulation using a flexible editor's console It is also unfortunate that, in this era of modern illustrates this concept. Another summary of on-line computer technology, the development of sophisti- process building, information retrieval, and text cated output devices has lagged seriously behind manipulation is found in Corbin. Magnuski's other types of data processing equipment. This master's thesis gives overall design specifications paper will discuss these problems, and in addition for the capabilities of an Information Retrieval and to presenting a detailed review of currently available Display terminal, using existing equipment in the equipment, will describe some of the more important context of the MULTICS GE 645 time-sharing elements of systems design to be considered in system. Research continues on the BOLD (Biblio- relation to the output of systems." (Austin, 1966, graphic On-Line Display) system." (Van Dam and p. 241). Michener, 1967, p. 197). 5.1b "University of Pittsburgh. The Computa- 5.1c Thus, "it is not unreasonable to imagine an tion and Data Processing Center (CDPC) whh author, sitting at a typewriter in his own office to support from the Advanced Research Projects be able to edit his manuscript by entering changes Agency of the Department of Defense, has developed for the computer, type requests for bibliographic a system for book and journal printing. Under the references from the library, control the format of leadership of Kehl (now at M.I.T.), this group at this publication, and request the computer to first carried on the work begun by Barnett at M.I.T., produce an index automatically." (Kehl, 1965, p. later developing a separate system, which included 26). use of hght pen, on-line typewriter, display scope, 5.2 "One of these [CRT] displays was available ." and a general-purpose text-editing program for in 1953 on the ILLIAC . . (Frank, 1965, p. 50). handling the processes of editing, formatting, and 5.3 DYSEAC, later PILOT, and to a certain introducing author's alterations. Large amounts of extent the latter's halfsister machine, Burlesc, at 31-channel monotype tape generated in composing Aberdeen, included, for the relatively early period The Physical Review were accumulated for conver- of their design, unusual features of multiple, sion to magnetic tape via a special reader built at variable-format input and output, priority interrupts the Center, to serve as a data base for experimenta- and control interlocks, and multiprogram swapping tion. These developments are described by and sharing characteristics. Ohringer, and Roudabush et al." (American "As an example, before digital computers can Institute of Physics Staff, 1967, p. 351.) be successfully incorporated into the often-predicted

79 'automatic factory' and 'automatic office' of the "One further device that may be attached to future, such equipment must possess ready means DYSEAC for special experiments is SEAC itself. for sending inteUigence to and receiving inteUigence As the two machines employ the same digital from a variety of external devices performing language, this attachment can easily be made many diverse functions. Some of these devices will through their regular input-output terminals. By have to store, tabulate, file, convert, display, and use of a coordinated pair of programs, the two sense information; still other devices will have to machines can be made to work together in common actuate mechansims such as servo equipment in harness on a number of interesting and potentially response to signals sent out by the computer as a useful tasks. Indeed, this mode of operation can, result of information being processed within it. by the appHcation of available technology, be The computer will have to direct all these devices extended to a widely dispersed group of information- and coordinate their activities into an ensemble processing machines that are interconnected by operation. Indeed, to achieve the full effect of an means of a communication network. With suitable ensemble operation, the system needs to have the feedback facilities and correlated processing characteristics of a generalized feedback loop. programs these machines could even engage in That is, the computer must not only exert control cooperative tasks for which the supervisory func- over these external devices, but they in turn must tions are transferred back and forth, as the need be capable of calling for alteration in the course of arises, among the several machines in the network. action of the computer. Such requests enter the These new approaches to the problem of automatiz- computer as special signals or information transfers ing industrial and commercial operations are reasonable extrapolations of current trends and can from the external devices. . . . "This feedback arrangement can be used to be expected to lead to practical results if vigorously introduce human monitoring and selective inter- pursued. . . . vention into the normal operation of the system. "As a second example of a potential application In its simplest form, pertinent information is dis- which requires these capabilities, consider an played to the human monitor who can elect to installation engaged in the control of air traffic at respond by actuating other external devices that an airport terminal. For this appMcation, the human supply the computer with either new data or operator can be coupled to the machine system (1) instructions. This feature will doubtlessly be a through the medium of a visual display device which highly significant factor in the efficient exploration exhibits graphically the numerical data stored inside of the new areas for which digital techniques appear the internal memory and (2) by means of a battery

so promising. . . . of manual control switches capable of inserting new "The kind and extent of the external devices to numerical data or certain new control commands be added will necessarily depend on the particular into the machine. The numerical data displayed by application which is being explored. Nearly all the machine represent in real time a map of the tasks need some printing and external storage predicted traffic patterns in the neighborhood of an facilities; hence, the initial installation will have a airport, e.g., the predicted locations, based on latest directly connected electromechanical typewriter scheduling information, of all aircraft inside a cer- and one or more magnetic wire cartridges for tain area. The human operator can, by means of the speedy loading and unloading of the machine. manual input-controls, enter newer data into the External storage in the form of magnetic tape equip- machine as the information becomes available, ment is also to be included initially in order to handle interrogate the machine for more detailed informa- the more usual computing and data-processing tion concerning certain individual flights as the tasks. More demanding tasks requiring somewhat need arises, and instruct the machine to exhibit faster access to a rather large volume of data will the effect to be expected from issuing revised probably lead to the annexation of one or more flight-control orders. In the latter case, parts of the magnetic drum units. Still more extensive problems internal information of the machine must be pre- concerned with means for handling masses of sented to the operator in a rather complex pictorial paper work will probably lead to the subsequent form in order to be meaningful to him. Applications addition of experimental magnetic disk memory of this sort, like the previous one, require a constant assemblies and experimental versions of the auto- interchange of intelligence between the computer matic magnetic file. For the exploration of real- and its operator, and between the computer and time problems, including simulation and control other devices." (Leiner et al., 1955, pp. 39, 40, 51). aspects, it will be necessary to annex input and 5.4 "A computer system, to work in partnership output converters to permit translation of infor- with a designer, must have several clearly definable mation back and forth from digital to analog form. capabilities. It must be able to accept, interpret, For example, digital-to-analog conversion is used and remember shape descriptive information when visual display of information stored inside introduced graphically. When such a graphical the computer is provided externally by means of input capability is properly designed, the man- special cathode-ray tube devices. This listing of computer combination can manipulate the elements external devices includes only those for which of a drawing in an entirely new way, with a freedom serious attention and development are already in and precision far surpassing what is possible with

progress. . . . pencil and paper." (Coons, 1963, p. 301). 80 5.5 "A display scope capable of generating image of each user when his program is not running. three dimensional curvilinear figures with con- A user's program in run status is run for 140 venient control of translation and rotation would milliseconds, then if there is another user also in be extremely useful. The approach that has been run status, the state of core memory is stored in the taken is to provide a function generator capable first user's core image on drum and simultaneously of generating straight lines or second-order curves, the second user's core image is loaded into core and and to have special purpose computing equipment the second user's program is started in the appro- in the display unit which can perform the three- priate place." (McCarthy et al, 1963, p. 51). dimensions to two-dimensional axonometric pro- 5.8 "AESOP is an experimental on-hne informa- jection computations. . . . tion control system realized in the Systems Design "A system is presently being designed for the Laboratory of the MITRE Corporation ... It Electronic Systems Laboratory which will display is a CRT display-oriented system in that the user on a scope a standard two-dimensional plot or a experiences the information system primarily rotated axonometric projection of a three-dimen- through his CRT displays and exercises his control sional line drawing. The system wiU be capable of through his light pencil. The current version of the generating first- and second-order lines and will AESOP prototype operates on an IBM 7030 (Stretch) provide easy means for the computer to translate computer (65 K memory with 64-bit words) with a or rotate the picture, or to magnify or de-magnify it. 353 disk storage unit holding two million words. Techniques are incorporated to prevent plotting Each of the four user stations consists of an on-line when the limits of the scope edge are reached." Data-Display-13 display console with a photo- (Stotz. 1963, p. 325). electric light pencil, an on-line typewriter, and a 5.6 '"Each time a drawing is made, a description Stromberg-Carlson 30370 medium-speed printer." of that drawing is stored in the computer in a form (Bennett et al., 1965, p. 435). that is readily transferred to magnetic tape. library A "The AESOP system, developed at the Mitre of drawings will thus develop, parts of which may be Corporation, employs an interactive display show- at fraction used in other drawings only a of the ing data structures and other aids in query formula- investment of time that was put into the original tion." (Minker and Sable, 1967, p. 137). drawing ... A drawing in the Sketchpad system 5.8a "As we have noted, the MCG [Man-Com- may contain explicit statements about the relations puter Graphics] concept is relatively new. It first between its parts so that as one part is .changed the began to achieve significance when I. E. Sutherland implications of this change become evident through- presented his 'Sketchpad' work at Lincoln Labs to out the drawing. For instance. Sketchpad makes the Fall Joint Computer Conference in 1963. Lock- it easy to study mechanical linkages, observing the heed-Georgia began its work in MCG early in 1964 path of some parts when others are moved . . . The without knowledge of any other industrial activity abihty of the computer to reproduce any drawn in MCG. In the fall of 1964, Lockheed-Georgia symbol anywhere at the press of a button, and to ordered a Univac 418 and a Digital Equipment recursively include subpictures within subpictures Corporation 340 scope. At that time, no computer makes it easy to produce drawings which are com- manufacturer had announced an MCG system. posed of huge numbers of parts all similar in shape." Shortly thereafter, IBM announced its Alpine (Sutherland, 1963, p. 332). system, while Control Data Corp. announced 5.7 "'Sketching is facihtated by the computer, MCG which posts a background outHne plan having the its Digigraphic system in 1965, and the General proper dimensions and showing existing structures Motors DAC-1 (Design Augmented by Computers) was unveiled at the Fall Joint Computer Conference that cannot readily be altered . . . On their in 1964. (The work had begun about 1959, sketches, which they can readily file away and recall GM for revision, the planners label the various depart- but had been undisclosed until this conference.) . . . ments and the stairs, elevators, dumbwaiters, etc. "One of the first applications envisioned for MCG Each label, typed on the typewriter, appears at was the numerical control parts programming the top of the oscilloscope screen, and then is ad- function. Conventionally, the parts programmer justed to desired size, trapped by the light pen, starts with a blueprint of a drawing and redraws it moved to its proper location on the plan, and according to his own needs. He then gives a coded dropped there. Each label serves as a storage and name to each point, line, and arc, and writes a

retrieval tag for the sketch to which it is attached. computer program in some stylized language such as The plan can therefore be made up in small parts APT (Automatic Programmed Tooling) to describe and displayed as a whole." (Licklider and Clark, the path a cutting tool must take to cut out the part 1962, p. 118). on a milling machine. This requires conventional "The purpose of the BBN time-sharing system batch processing and the usual wait is required to is to increase the effectiveness of the PDP-1 com- get a plotted output from which the program may be puter for those applications involving man-machine verified. Mistakes require corrections to the program interaction by allowing each of the five users, each and additional waits for 'batch' output. at his own typewriter to interact with the computer "With MCG, the drawing is put onto the display just as if he had a computer all to himself. ... A scope by input from cards, light pen, tape, or com- 4096 word drum field is allocated for saving the core binations thereof. The cutter path is depicted on the

81 scope as the operator maneuvers it with the light such output displays as maps, charts, and diagrams." pen. Any mistakes in path definition are instantly (Direct Communication Between Man & Computer", apparent and may be immediately corrected. As National Bureau of Standards Tech. News Bull. the operator describes the cutter path on the scope, 50,53-54 (Apr. 1966).) a tape is automatically defined. This tape, when "ACCESS, A DATA-PROCESSING SYSTEM post-processed to adjust to the format for a par- RECENTLY DEVELOPED BY THE NATIONAL ticular milling machine, is then used to cut out the BUREAU OF STANDARDS' INSTITUTE FOR depicted part as often as is required." (Chasen and APPLIED TECHNOLOGY, represents an advance Seitz,1967,p. 25). in communication between man and computers. 5.8b See also Gallenson as follows: "A graphic This system will receive and process data from tablet display (GTD) console has recently been local and remote sources and can present its output designed to operate with the Q-32 Time-Sharing in a form immediately intelligible to the human operator. (so-called for System (TSS) at System Development Corporation. Access Automatic Computer The successful completion of the task has provided Controlled Electronic Scanning System) was devel- for use the Planning new insights into system design problems involving oped by Office of Emergency highly interactive consoles in a time-sharing environ- to help provide rapid access to digital and pictorial ment; some interesting innovations in the design of a data. The ready availability of these data will aid the in evaluating situations during a national graphic tablet console; and some solutions to graphic OEP emergency. Access accepts input information tablet interface problems. . . . directly from microfilm records of hand-marked "Employing the tablet display as a projection on documents and digital information either from other the writing surface has improved the 'naturalness' machines or directly from its keyboard. It has been of this I/O device and has encouraged more users. used in experimental work at the Bureau to accept The display hardware is more expensive for equiva- such graphical material as specially prepared maps lent quality of direct-viewing displays, but the | and charts. Access scans marked documents by additional advantages gained may be justified for means of an advanced version of Fosdic which is a given applications. Additional innovations can be "Film Optical Sensing Device for Input to Com- introduced to improve the flexibility of the GTD puters." Fosdic was initially developed jointly by configuration. These include: the projection of NBS and the Bureau of the Census for machine slides in conjunction with the CRT image for tablet reading of census documents." (Computer Design 4, manipulation; the adding of a color wheel to achieve No. 12, Dec. 1965, p. 12). computer-generated color displays; and the use of a "A machine which has been developed ... as a camera to achieve hard copy output from the CRT. research tool for the investigation of man-machine Significant improvement in light output of the techniques involving CRT displays . . . has been projected CRT image must be achieved before these designated MAGIC (Machine for /Automatic Graphics additional capabilities are introduced." (Gallenson, /nterface to a Computer). This machine combines 1967, pp. 689, 694-695). large-diameter cathode-ray displays with a specially We note also the following: '"Work is proceeding designed programmable digital computer. It is on the implementation of an experimental network designed as a remote display station and is intended involving the APEX time-sharing system running on to be connected to a large ADP system via voice the TX-2 computer at M.I.T. Lincoln Laboratory in quality communication systems." (Rippy et al., Lexington, Massachusetts, and the time-sharing 1965, 819). system running on the Q 32/PDP-l computer p. "A presentation can be obtained on MAGIC's complex at System Development Corporation in display in three ways: a previously mem- Santa Monica, Cahfornia. Initially, a 4KC four-wire primary dial-up system will be used with 1200-bit-per-second orized figure can be obtained from the machine's asynchronous modems." (Marill and Roberts, 1966, memory, the operator can draw a straight-line figure of the light pen, or a drawing can be p. 430). by means 5.9 ACCESS (/iutomatic Computer Controlled assembled from components in the machine's Electronic Scanning System) "was developed for "library" of symbols. In the last method the position- use by the Office of Emergency Planning to help ing operations are used to manipulate components

provide rapid access to digital and photographic and the light pen to connect them. . . . can not only position material; he data . . . [It] accepts input information directly "The operator from microfilm records of hand-marked documents can also shrink, expand, and rotate it. When the and digital information either from other machines drawing has been edited to the operator's satisfac- or directly from its keyboard. It has been used in tion, it can be returned to the machine's memory. experimental work at the Bureau to accept such The digital information can be used to produce graphical material as specially prepared maps and 'hard' copy also; this has been done by placing it charts. It will store the information, perform a on magnetic tape to drive an X-Y plotter. This variety of operations on it, and present outputs capability lends itself to making engineering changes either in digital form for use by other machines or on drawings and keeping up-to-date drawings on the in a form requiring no further translation for man. tape for ready inspection and reproduction. It also The system includes an X-Y plotter which prepares permits making additions to maps, charts, or floor

82 plans, a capability which should be useful in chart- Console — designed as a man/machine interface ing routes and courses, adding isobars and symbols for information-handUng systems. The console to weather maps, and recording occupancy assign- provides aU the necessary tools for graphical data ments." (NBS Tech. News Bull. 50, No. 8, 135 analysis, data generation and computer control. . . . (Aug. 1966).) "The console can interface with almost any com- 5.10 "The Shared Memory Computer Display puter or digital communications system. It uses System described by Ball et al., used a DEC 340 keyboard buttons that can be labeled by an overlay CRT whose display instructions are stored in an in natural, problem-related terms to query an attached PDP-4, which is in turn connected to a associated computer. . . . larger, batch-processing CDC-1604-A. The display "The operator presses these buttons to actuate computer receives input from three keyboards, computer-programmed subroutines; the computer pushbuttons, shaft encoders, three-dimensional responds in the form of alpha-numeric or graphic trackball (rolHng sphere control), Ught-pen, a displays on a cathode ray tube. . . . RAND tablet, and has "bell, two-tone gong, telegraph "It is possible to reorient the console and the sounder, and other small noise-making devices, computer program it controls for virtually any operated by relays under program control'. The application through keyboard overlays. The two DEC 338 is a commercial version of this type of new and significant features of the BR-90 VAC system: a small computer dedicated to a CRT to are stored-program logic and combined electronic provide flexible display programming." (Van Dam and photographic displays on a rear-ported cathode ray tube." and Michener, 1967, p. 196). (Technology Week, Nov. 21, 1966). 5.11 "Bell Labs has devised a simple and eco- 5.14 "A display system that provides continuous nomical console that displays line drawings, letters, visual presentation of data from high-speed digital and numbers in answer to problems submitted to a computers has been developed by the Strand Div. computer. Known as Glance, the console can display of Datatronics Engineers, Inc. . . . The system computed data such as graphs, plots, IC masks, and consists of an input buffer unit, display programer unit, character generator, line generator, and 12- drawings." (Data Proc. Mag. 9, No. 6, 3 (1967).) inch screen monitor. In typical applications, more 5.12 "Benson-Lehner Corporation . . . has developed a new sohd-state Large Table Llectro- than 1,000 flicker-free characters and line segments

may be generated and displayed . . . Special plotter (LTE) . . . [which] affords a complete contouring package for large-scale computer microfilm output from the D-2(30 recorder is suited to of applications for exact delineations in minute detail intermediate storage computer output data." (Bus. No. 14, 2 of surface contours, weather maps, and topo- Automation News Report 2, (1962).) 5.15 "The Gerber Scientific Instrument Co., graphical features. It produces report-quaUty Inc., has announced the design and dehvery of the graphs and maps rapidly and accurately from digital first Stored Program Graphic Display System for computer generated output tapes. From either pro- automatic drafting. Now in operation at Lockheed- gram or operator control, it plots points, symbols, or Georgia Company, the System will be used initially alphanumeric characters, and draws straight or to lay out full-scale undimensioned structural draw- contour graphs from digital input data". {Com- ings and contour-lines drawings for aerodynamic munications ACM 8, 343). structural 5.13 [Bunker-Ramo 85 Control/Display Console] testing and studies during design of the logistics transport aircraft." (Comm. "Text, symbols, point plots, and lined drawings C-54 ACM 9, may be produced on the 12- by 16-inch active display 781 (1966).) 5.16 "Information Displays' high capacity area of the 23-inch aluminized TV-type screen . . . display 10001 uses three consoles Symbols are generated at the rate of 100,000 per system Type CM a display generator, the second, may be produced in two sizes, and may be driven from common and entire operates on-line with the made to bUnk on the screen." (Dig. Comp. News- system UNIVAC 1107 computer. Provisions are included for vector letter 16, No. 4, 39-40 (1964).) circle "A new video display system that provides com- writing, character writing, and drawings, as puter processed information about developing well as programable line texture, programable market trends on the New York Stock Exchange character size and programable intensity." (Data Proc. 45 has been installed by the Bunker-Ramo Corporation Mag. 7, (Feb. 1965).) at Shields & Company's new Stamford, Conn., "A new computer-controlled display system, the office." (Data Proc. Mag. 7, No. 5, 12 (1965).) Type CM 10093, will be shown by Information "A new family of display devices, the 200 series, Displays, Inc. . . . The 10093 features a fast is being offered by the Defense Systems Div. of response, large screen CRT display, along with the Bunker-Ramo Corp. The display units combine high speed character and vector writing capability a TV-like screen with a keyboard similar to that of and program control of size, intensity, and line a standard typewriter. Information being entered structure. The completely solid state CM 10093 into the computer or the computer's response are features a light pen to facilitate man-machine simultaneously displayed." (Data Proc. Mag. 7, dialogue." (Commun. ACM 9, 782 (1966).) No. 10,63(1965).) "Up-to-the-moment company operating data can

"Bunker-Ramo's console — Bunker Ramo landed be presented on the Information Displays, Inc. . . . an Award of Honor for the BR-90 Visual Analysis Management Display System, IDI Type CM 10009, 83 shown for the first time at IFIP Congress 65. Driven a highly effective method of real-time use of time- ." over Dataphone from a UNIVAC 490 computer, shared computers . . (Commun. ACM 8, No. 11, the system can be many miles from the computer. A 715 (Nov. 1965).) large (4096 x 30) core memory in the display "In the READ System, the single monitor and refreshes the display rapidly at flicker-free rates, light pen is replaced by a group of up to 15 consoles while at the same time, accepts low speed Data- with both keyboard and light pen. These consoles phone signals from the computer." (Commun. are driven from a single set of display logic which ACM 8, 479 (1965).) consists of a character generator, vector generator, 5.17 "Informatics Inc., has received over format generator, control logic unit and computer $400,000 contracts from the U.S. Air Force's Rome interface. This system is unique in that the presently Air Development Center to evolve a system for high speed of operation of digital logic is utilized to on-line computer use. The system to be known as generate characters and vectors at a speed which DOCUS (Display Oriented Compiler Usage System), exceeds the amount of data that can be displayed wiU employ sophisticated display equipment, on a single cathode ray tube. By using this high operating on-line with modern, large-scale com- speed, a single set of generation elements can be puters." (Comp. News 9, No. 7, 5 (1965).) used to drive several parallel consoles. This results "DOCUS, developed by Informatics for the in very great economy since the generation units multicomputer complex at Rome Air Development form a major portion of the total cost of a display Center, provides a display-oriented man-machine system. The READ System has the ability to dis- system with some built-in data management func- play over 200,000 characters per second and over tions and a capabihty to assign functions to control 100,000 vectors per second on the remote consoles. keys at the display and define and execute com- The organization of the system is such that a unique pound functions based on earlier defined tasks." display can be formed on each of the remote (Minker and Sable, 1967, p. 137). monitors or an identical display can be shown on "The development of software techniques for each of the monitors. The common display logic on-line information processing and man/machine to monitor separation in this system cannot exceed communication is then the focal point for the 75 feet. The computer to common display logic Display Oriented Computer Usage System (DOCUS). connection could be made considerably longer since are to facilitate the Techniques being developed the data rate is relatively slow between these implementation of on-Une procedures and the terminals with an average word time of approxi- operation of such procedures through a man/ mately 15 microseconds. The entire READ System users' areas of computer conversion mode. The interfaces to a 131 data channel on a DEC type interest may be related to information retrieval, PDP-1 computer. A portion of the 20K memory scientific research, management, or command and of the computer is utiUzed to refresh the display It users control. is desirable to make these on-line with approximately one third of the available less dependent upon the professional programmer memory cycles used for this purpose." (Scholten, by simplifying the implementation and reducing the 1965, p. 128). time for programming new applications." (Corbin 5.21 "Sanders Associates, Inc., has announced and Frank, 1966, p. 515). a new line of information display systems primarily "DOCUS has been implemented for a CDC designed for use with commercial computer in- 1640B computer to which is attached a Bunker stallations." (Data Proc. Mag. 7, No. 11, 56 (1965). Ramo 85 display console. The techniques which 5.22 "Development of a versatile-display Com- were developed are, however, not restricted to puter Input/Output Communications Console this hardware which has served primarily as a combining multiple input controls, visual display of test bed." (Corbin and Frank, 1966, p. 515). entire messages before transmission, and com- 5.18 "A display console combining electronically patibility with a broad range of digital data process- generated symbols and slide projected images has ing systems has been announced by Tasker Instru-

been announced by ITT Federal Laboratories. . . . ments Corporation . . . The Model 544 console The console is called MACC-Modular Alter and can be tied in with up to 15 similar slave sub- Compose Console." (Commun. ACM 9, 707 (1966).) consoles . . (Communications ACM 8, M^). 5.19 "'3-M', the Minnesota Mining and Manu- "A high-speed, real-time modular CRT display facturing Company, offers a 'Visual Communications system that permits the customer to 'build his own' Center' utilizing a desk top projector whose display functional configuration is announced by Tasker can be 'written on' and a transparency-copier." Instruments Corp. of Van Nuys, Calif. Designated (Bus. Automation 12, No. 10, 60 (Oct. 1965).) Series 9000 Modular Display Console, the new 5.20 "A computer-driven display system is system features customer options of a customized scheduled for delivery to the Carnegie Institute of or any standard keyboard, five other control devices, Technology Computation Center in Pittsburgh four display devices, and four different memory this Fall by Philco Corporation. The complex subsystems. The Series 9000 Display Console wiU

cathode ray tube system . . . was adapted from be shown to the industry for the first time on Novem- Philco's Real-Time Electronic Access and Display ber 8 at the FJCC in San Francisco." (Commun. (READ) system. The advanced system will provide ACM 9, 831 (Nov. 1966).)

84 5.23 "The IBM graphic data processing system With this new keyboard/display station, programs takes advantage of the power of a computer to can be displayed in a format identical to the pro- handle information displayed in the form of sketches, grammer's coding sheet, and in a data acquisition drawings, diagrams and graphs. This system allows system, the station can also be used for display of scanning of an existing microfilm image or calling raw or reduced data as the test progresses." (Sys- out an image stored digitally within a computer, .tems 7, No. 2,8(1966).) displaying it on a screen, changing the image with an "New data display devices for the RCA 301-3301

electronic 'pen', and, within seconds, recording the processors . . . [consist] of a keyboard, display new image permanently on microfilm and reviewing and memory-logic element connected to a central a projection of the image 19 times its actual size." processor through data sets, voice grade communi- ." (Data Proc. Mag. 7, 45 (Feb. 1965).) cation circuits and communication buffers . .

. . [IBM] announces the new, low-cost IBM (Data Proc. Mag. 7, No. 2, 44 (1965).) 2260 display station, which can be Unked to an IBM "The CRT sets to be designed and manufactured

I System/360. The new device will make vital informa- by Univac wiU be able to display over 1000 char- tion stored in the computer available for review on a acters on a 5" X 10" viewing screen. The i" char- 4 X 9-inch screen. The information also can be acters are now formed through a stroke generation I r revised or updated with a display keyboard." method. Since the sets will be the major interface (Commun. ACM 8, 342 (1965).) point between the users and the computer, a great "'The IBM 2260 enquiry display terminal com- deal of time was spent during the development and prises an input keyboard which is used to enter an negotiation phases of the study to design a set enquiry, and an output screen which displays the that incorporates features that recognize human

answer . . . The information is written on the view- j engineering factors." (Porter and Johnson, 1966, ing screen at a rate of up to 2,560 characters a I p. 81). second and a total of 960 characters arranged in 5.24 "The automatic drawing of PERT charts 24 lines each of 40 characters, can be displayed at was demonstrated recently by North American

one time. . . . Aviation . . . The chart is reproduced by General "An even more advanced display unit is the IBM Dynamics/Electronics' SC4020, which prints an 2250 which has the ability to display graphic images image from a cathode ray tube onto a 9" x 9" hard- as well as alphanumeric data derived from the copy or 35 mm film." (Datamation 9, No. 6, 50 computer. The basic console contains a 21-inch (June 1963).) cathode-ray tube on which 3,848 characters can be "Other applications of visual display . . . altera- displayed simultaneously in 52 lines with 74 char- tion of a PERT network to determine the effect acters per line [using a character generator and ." of different target dates . . (Silveira, 1965, p. 37). optional buffer for image regeneration]." ("IBM ". 5.24a . . Human engineering in the broad System/360", 1965, pp. 297-298). sense is a basic consideration in this system con- "A user can transmit messages to other display cept; the system is required to present to the user stations or to the central computer complex from the interface which is best matched to the problem 1 )i ; either an alphanumeric or numeric keyboard. The and the user's approach to its solution." (Mills, f messages are composed in their entirety, displayed 1966, p. 196). h for verification and editing, and then transmitted "More than any other single concept, the man- I. ! for computer processing and possible display at machine system concept is the one around which 1- other locations." (Commun. ACM 8, 343 (1965).) j most human factors engineers rally. It is perhaps e '"The cursor can be moved to any point on the now the core of the field. Thus, all human factors if screen, giving the user freedom to position informa- engineering can be thought of as dealing with the I I- ; tion wherever he desires or to erase any displayed appropriate allocation of tasks between assemblages •r data. This symbol gives the user the ability to edit of and machines, in which the two anomalous h and correct alphanumeric charts or listings on the men components the and the machine are in * I screen, and update information stored in System/360 — man — such an intimate cooperative relationship that the ). with these changes." (Commun. 8, 343 (1965).) j ACM "A computer-directed can designer is obliged to view them as a single system. I drawing system which Rather than being called upon to adapt a finished ly : I produce detailed construction drawings 25 times machine so human operators can live with it more i' faster than a draftsman was demonstrated recently I easily, the human factors specialist wants to be in !r by Control Data Corp. The demonstration involved I on the design from the very beginning to insure that nl computers in the company's Chicago and Min- j » neapolis data centers and the use of long distance best trade-offs between men and machine functions I 4 telephone lines over which the problem and solution are chosen." (Lindgren, 1966, p. 135).

*. were submitted." (Bus. Automation 12, No. 7, 55 5.25 "One of the more promising areas of

ry ; (1965).) computer use involves the coupling of a man to a il "Raytheon Co. has recently announced that it computer system for real-time problem-solving

B- has made several improvements on its 520 Computer where the procedure for solution of the problem is

]i System. These include a 1-microsecond main either unknown or involves complex tasks such as

! memory and a new keyboard-display station pattern recognition, that can best be performed by

i which makes possible on-Hne display and editing. humans." (Brown, 1965, 82). I p. 85 "Use of the graphical input-output console allows "At the present stage of this project, the quantities the designer to pause and study his design at all of displacement . . . and wetted surface can be stages in its process without unnecessarily tying up evaluated for any hull or portion of a huU; the frame- the computer. Furthermore, he can make modifica- work also exists for future subroutines to calculate tions at any time and be appraised within seconds waterplane inertias, mass inertias, weight of shell of the effects of each change." (Hamilton and Weiss, plating, and eventually (given some loading patterns), 1965, p. 3). stress distributions in hull plating." (Hamilton and "Accessibility, provided through some type of Weiss, 1965, p. 11). remote computing facihty such as time-sharing, "Arthur D. Little, Inc., is developing under the permits the engineer to interact with the computer sponsorship of the Naval Ship Engineering Center, during the problem solution." (Roos, 1965, p. 423). Naval Ship Systems Command, a computerized "The complex of recent developments centered system for the design detailing of hull structures for around time-sharing systems and aimed at exploita- naval ships. The system takes its input data from tion of man-machine interaction is, in the view of the ship's contract plans, standard design specifica- many, the most exciting current trend in the tions, and parameters; it produces as output working computer field, and the one with the most far- drawings for use in the shipyard, APT part programs reaching implications to each of us as an individual for numerically controlled flame cutting machines, and to society as a whole." (Fano, 1967, p. 30). and auxiliary information such as center of gravity 5.26 "The IBM graphic data processing system statements and bills of material and weight." takes advantage of the power of a computer to (Cohen etal., 1967, p. 341). handle information displayed in the form of sketches, Research at the University of Glasgow is also drawings, diagrams and graphs. This system allows aimed at the computer design of ships, "from the scanning of an existing microfilm image or calling initial conception of the shape of the huU to the out an image stored digitally within a computer, production of detailed drawings of the steelwork." displaying it on a screen, changing the image with (Commun. ACM 9, 539 (1966).) an electronic 'pen', and, within seconds, recording 5.29 "CADET [Computer-Aided Design Ex- the new image permanently on microfilm and review- perimental Translator] systems are aimed at mixed ing a projection of the image 19 times its actual size." verbal and graphical man-machine problem-

(Data Proc. Mag. 7, 45 (1965).) solving. . . . 5.27 "A display system that provides continuous "An initial version of the CADET system is sched- visual presentation of data from high-speed digital uled for operation in the fall of 1965 and will be computers has been developed by the Strand Div. applied in such areas as three-dimensional shape

of Datatronics Engineers, Inc. . . . The system description, nonlinear electronic circuit design,

consists of an input buffer unit, display programer etc. . . . unit, character generator, line generator, and 12- "The designer communicates to the computer by inch screen monitor ... In typical applications, moving the Display Console light pen to various more than 1,000 flicker-free characters and line positions on the CRT. He indicates the meaning of is segments may be generated and displayed . . . these positions by pushing buttons while the pen

Special microfilm output from the D-200 recorder at the desired locations. . . . is suited to intermediate storage of computer output "The PLOT operator is used to create a picture data." (Bus. Automation News Report 2, 2 (1962).) on the oscilloscope corresponding to the graphical

5.28 "The . . . characteristics of the design language statements ... It builds a list of 'console problem strongly suggest that the process would be commands' (plot a point, or a line, reposition the greatly enhanced if tools were made available by beam etc.) known as the display file, which are which the designer could be quickly appraised of interpreted by the display console hardware so the multifaceted effects of fixing any particular producing the desired picture." (Lang et al., 1965, variable. Such is the capability of the high-speed pp. 3,4,5,21,39).

digital computer. . . . Similarly "an experimental computer program "The idea that the designer can (in a CAD system) which determines how to interconnect 80 electronic explore a much wider range of answers to the circuits on a wafer of silicon, and then controls decision-makers' questions may be an appealing fabrication of the interconnections, was described one, particularly because of the vagueness and recently by an IBM scientist. changeability of the ship's missions at the inception "The program for the first time permits automatic of design." (Hamilton and Weiss, 1965, p. 5). by-passing of any faulty circuits on the wafer, so "The designer or engineer is interested not only that most wafers are usable despite the occurrence in the surface shape, but also in most of the various of faulty circuits at random positions." (Data and surface properties: area, volume, mass moment of Control 4, No. 4,5 (Apr. 1966).) inertia, cross-section inertia, center of gravity, 5.30 It is noted that "Lockheed-Georgia uses a resultants and equivalent points of application of Univac 418 and a DEC 340 display scope with light external loads (both point and distributed loads), pen for research design applications for aerospace." cross-sectional views, and views of the surface, (Data Proc. Mag. 7, No. 5, 8 (May, 1965).) including 'hidden lines', as seen from various 5.31 "Master dimensioning gives a computer the orientations." (Hamilton and Weiss, 1965, p. 13). ability to derive math definitions of complex sur- 86 faces, and to combine many complex surfaces by rivets. He then drew a support at each end of mathematically to produce a perspective view the girder and a load at its center. The sketch of

of an object . . . North American Aviation, Inc., the girder then sagged under the load, and a number is reported to have used master dimensioning in its appeared on each member indicating the amount of design of the B-70 supersonic bomber. Boeing also tension or compression to which each member was used master dimensioning for perspective art on a subjected." (Teitelman, 1966, p. 11). new fighter aircraft." (Gomolak, 1964, p. 68). "An IBM program, ALDEP (Automated Layout "At General Motors in Detroit, the computer-aided Design Program) aids the plant or builder designer design system, GEM, is used operationally in the in developing optimized layouts for various activities design of automobile bodies. Among other things, to be accommodated within constraints of the space GEM coordinates the artistic and engineering available, including the graphic display of layouts aspects of design, checking, for example, the as they have been evaluated." (Seehof et al., 1966, strength and manufacturability of a fender as soon p. 191). as the designer has sketched it on the oscilloscope "Graphical I/O appears to have very great screen." (Licklider, 1965, p. 66). potential for the engineering of transportation "Using the 1965 models as a starting point, the facilities. A scope with input as well as output designer inserts still another piece of information capabilities allows easy, positive and rapid change into the computer. Looking at this year's car, he in any assumptions or designs which can be ex- says, 'Let's see what it would look hke if the body pressed graphically — a highway network, a roadway were two inches longer, the greenhouse an inch cross-section, vertical and horizontal roadway lower, and the fenders wrapped around an addi- ahgnment, intersection layout. Plotters make tional three inches'. This information in almost this possible automatic drafting of final or intermediate exact language is . . . loaded into the computer designs, even construction plans." (Breuning and

. . . the styHst tells the When computer what view Harvey, 1967, p. 261). he wants, the computer takes the modified master "A noted architect is experimenting with the dimensioning data, combines it with the graphics technique for designing apartment houses, office program and comes up with the coordinates needed complexes and ideal communities. A major bridge for the perspective drawing." (Gomolak, 1964, 69). p. builder is also looking at drafting machines. And 5.31a "Rowland Brandwein, director of research one of the leading makers of home appliances for Span Arch Structures, Inc., Conn., Nevtown, plans to use the graphics program for selecting has designed a new kind of arched roof with the models, along with numerically controlled machining aid of a communications terminal in his office linked to turn out washers, dryers and refrigerators." to an IBM computer in New York City. The metal- (Gomolak, 1964, p. 71). and-plastic arch is cheaper and easier to put up 5.32 "In current practice, the woven design is than conventional ones, and so Ught that it does represented on point paper (graph paper) which away with many of the supports normally used. has the number of horizontal or vertical lines per "After hitting upon the idea of making an arch inch in the same ratio as the number of weft and by sandwiching an adhesive plastic between thin warp threads. Each rectangle represents the inter- metal, Brandwein used his terminal linked to an section of one warp and one weft thread, i.e., one IBM QUIKTRAN center to see if his theory was interlacing. At each intersection the warp thread practical. He not only used the computer to work may pass over the weft thread or vice versa. This out the basic methods for constructing the arch, binary choice now represented by painting or not but also to design special equipment needed by painting the point paper rectangle corresponding contractors for the job and to create instructions

to each interlacing. . . . for them." (Computers and Automation 17, No. 5, 54 (May 1968).) "In reference 1, the first approach, the pattern 5.31b "At the Atlanta meeting of the American is scanned by a 'photohead'. Then it is dissected, Association of Highway Officials in December, a within a computer, into small equal parts called civil engineer designed a superhighway interchange shape elements. Each of these parts is then sub- in ten minutes through an IBM 1050 data communi- jected to a pattern recognition routine in order to cation system connected a thousand miles to the classify it into a 'primary standard image' category. MIT 7094 in Cambridge." (Data Proc. Mag. 7, The set of primary standard images represent No. 2,7(196.5).) various versions of shape elements prepared by the Another system designed by Nordisk ADB and artist. When the shape element is classified, its the Swedish Board for Road Construction has been corresponding secondary standard image is chosen reported as capable of making automatic "a large to represent it in woven form. The secondary amount of the work of road design." (Langefors, standard image represents the same shape element 1965). with account taken of the laws of weaving (inter- Other civil engineering examples are represented lacing rules). Each shape element is analyzed in by the ICES system and by the following: turn, thus producing, in one pass, a complete set

"In one impressive demonstration, Dr. Sutherland of fabric interfacings. . . . sketched the girder of a bridge, and indicated the "After the design has been introduced into the points at which members were connected together computer by tracing or scanning, the CRT is used as

87 the primary display device. By manipulating the graphical and alphanumeric displays of the results display with function keys and light pen, the provided within minutes. designer can use the computer as an experimental "A 'demonstrator' version of this system is designing tool." (Lourie et al., 1966, p. 539). available for the IBM 1620 computer. The demon- "He may display the design on a grid with different strator uses only the standard computer console ratios of horizontal and vertical lines which repre- typewriter and card reader/punch, and does not sent different ratios of warp and weft. He may also require any special equipment.'* (Clem 1966, alter the number of warp threads per inch to see Abstract, p. 116). how the design would look woven with different 5.32b "The AMTRAN (Automatic Mathematical degrees of coarseness. He may refine the mesh by r/?y4A^slation) System has been developed at the enlarging the design on the same grid or holding the NASA Marshall Space Flight Center in Huntsville, design constant and imposing a finer grid. The Alabama and has been available on a limited basis designer may manipulate the display in other ways. to users there since early 1966." (Ruyle et al., 1967, He may 'repeat' the design with different spacing p. 151). See also Reinfelds et al., 1966. between centers; he may specify different Unes of 5.32c "The Culler-Fried System is one name symmetry and thus develop a full design from a for two physically separate but direct descendants small segment. Many other techniques can also be of the system developed by Glen Culler and Burton made available to him. . . . Fried at Thompson Ramo Wooldridge, Canoga Park, "When a set of weaves has been determined, rules California, beginning in 1961. The first of these is concerning the interaction of adjacent weaves must the On-Line Computer System (OLC) at the Uni- be applied. For example, for structural purposes versity of California at Santa Barbara (UCSB) there is a rule that no warp thread may float over which operates on the same computer as the original more than n weft threads (or vice versa). A violation system. The OLC system is used for research and of this rule will be pointed out to the user who may teaching on the Santa Barbara campus as well as correct the situation by altering a specific interlacing from remote terminals at UCLA and Harvard. The with a light pen or by changing the starting position second, an expanded version of the original system, of one or both of the weaves in question. Another has been implemented at TRW Systems (formerly example of such rules concerns a special set of Space Technology Laboratories) in Redondo Beach, interlacings which delineate an area and which California for the use of scientists and engineers are 'shape binders'. and has been operating since late 1964." (Ruyle "Programming can incorporate many of the prin- et al.,1967,p. 151). ciples of this technique. Refinements may be made 5.32d "The Lincoln Reckoner system has been by the designer with the light pen. It is this ability developed at the Lincoln Laboratory of the Massa- to alter and 'touch up' the automated output that chusetts Institute of Technology to operate within gives this system the required flexibility." (Lourie the APEX Time-Sharing System for the TX-2 et al, 1966, pp. 537, 539, 542-543). computer and has been used by the staff of the 1966." 532a. See also Ruyle et al. (1967) as foUows: laboratory since early (Ruyle et al., 1967, "Martin has developed a PDP-6 LISP program that p. 152). converts complex mathematical expressions stored "The Lincoln Reckoner is a time-shared system in a list structure form into a CRT display that is for on-line use in scientific and engineering research quite close to textbook mathematical notation, and . . . It offers a hbrary of routines that concentrate with which the user can interact by means of a on one particular application, numerical computations on arrays of data. It is intended for in light pen." (Ruyle et al., 1967, p. 166; ref. is to use Martin, 1966). feeling one's way through the reduction of data from a laboratory experiment, or in trying out "The AMTRAN computer system is a time- theoretical computations of moderate size." (Stowe sharing, keyboard computer system which permits et al.,1966,p. 433). a scientist to enter mathematical equations into a computer in their natural mathematical textbook 5.32e "The MAP (Mathematical /Analysis with- format and, barring complications, to obtain im- out Programming System, which operates within mediate graphical and alphanumeric display of the the M.I.T. Compatible Time Sharing System solutions on a cathode ray scope and a typewriter. (CTSS) at the M.I.T. Computation Center or at The system can be used for straightforward prob- Project MAC, has been used in research and teach- is access- lems by a scientist with no previous computer ex- ing at M.I.T. since mid 1964. The system teletype or 1050 terminal that perience while, at the same time, it provides the ible from any IBM the physical flexibility required by the experienced programmer can dial into CTSS, regardless of et al., 152). to solve non-routine problems. Great improvements location of the terminal." (Ruyle 1967, p. in convenience and accessibility as well as in 5.32f "The MATHLAB System of Engelman, programming and turnaround times accrue to the which operates under CTSS, is an important start user in comparison with existing computer systems. toward providing a system capable of providing For example, a typical nonlinear differential equa- both analytical and numerical assistance." (Ruyle tion can be entered and solved, and permanent et al., 1967, p. 166; ref. is to Engelman, 1965).

88 5.32g "A new user-oriented programming system communication and feedback, which greatly for the purpose of facilitating the programming and simplifies the user's task, is an important function of analysis of well-formulated problems has been most on-line systems — particularly graphic ones. designed and implemented at Columbia University, This provides a 'forgiving' system in which mistakes Hudson Laboratories. This system, consists of a by the user are not catastrophic but are called to standard Flexowriter modified to construct two- his attention for appropriate action." (Hobbs, dimensional mathematical expressions and a new 1966, p. 1873). programming language." (Klerer and May, 1965, 5.34 "The present approach was to define an p. 63). algorithm capable of one-pass operation on a small Other aspects of the program at the Hudson (IBM-1620) computer, using a Calcomp digital Laboratories of Columbia University include the incremental plotter online to produce the annotated development of auxihary consoles for "pushbutton" flowchart." (Anderson, 1965, p. 38). programming, R&D efforts with respect to auto- 5.35 "Applied Data Research, Inc., . . . intro- matic mathematical text typesetting, and investiga- duced AUTOFLOW . . . [which] accepts assembly tion of capabihties for two-dimensional program or higher level languages and produces quality and text editing. (Klerer, 1966). flowcharts for online or offline printing . . . The 5.33 For example, "using an NCR Elhott 4120 AUTOFLOW system allocates the necessary size with a display and light pen equipment, the NPL for a symbol, rearranges groups of symbols and will develop methods to aid the computer program- draws lines between symbols." (Commun. ACM 9, mer by allowing him to write, test and edit programs 469 (1966).) in a high-level language, on-line to the computer. "1966 additions to Apphed Data Research Inc.'s The computer and display will also be used in ." AUTOFLOW systems were for system 360 COBOL patten recognition research . . (Data and Control and FORTRAN. (Datamation 12, No. 11, 93 (Nov. 3, 11, (1965).) ". 1966).) . . Recent rumors suggest flow charts to be used on oscilloscope terminals for on-line program- "The proprietary package AUTOFLOW was ming." (Burkhardt, 1965, p. 7). released in February 1967, having been developed "Other, more radical uses of display devices in under NASA support. This system produces flow- debugging are now being investigated. Flow-chart charts from program units written in MAP/FAP, languages, where programs are created on-line by FORTRAN, and MYSTIC. MAP/FAP and generating a flow chart with a light-pen, are being FORTRAN decks may be optionally augmented studied at Lincoln Laboratory and at RAND. A by special comment cards and codes which will dynamic display of the program state at any point alter the interpretation of the source and thus in terms of the flow chart is expected to be a useful modify the quality of the flowchart produced. Of debugging tool." (Evans and Darley, 1966, p. 44). course the use of these options requires extra effort "William Sibley's and Thomas Ellis's on-line on the part of the user. AUTOFLOW also produces programming system, FLOWCHART, in which a Table of Contents and a Cross Reference Listing. the flow diagram is the source-language computer Output may optionally be obtained on the SC-4020 program and through which the programmer writes, and/or the line printer; the internal tables may be edits, debugs, and operates the program by sketch- employed to drive either output. Table overflow can ing and printing with a stylus on a RAND Tablet. occur, but a capacity of 30,0()0 input cards and 1700 The Tablet is a graphical input device that looks statement numbers is claimed. The output employs like a sheet of paper on a blotter pad but underneath variously shaped boxes to represent the action and is actuaUy a grid of wires, through capacitative pays much attention to the location of all boxes, coupling to which an electronic device can sense especially transfer points. All boxes are identified the position of the stylus." (Licklider, 1967, p. 10). by page and sequence number. The contents of "Use of on-line graphical displays for flow the boxes are identified by page and sequence charting and programming is an interesting appUca- number. The contents of the boxes are language tion in the computer field. An example of work dependent, but clear, and in some cases expanded, of this type is that done at the RAND Corporation in reproductions of the source card content." (Abrams. which a rough flow chart can be drawn with a 1968, p. 745). graphical input device and displayed on the screen. 5.35a ''AUTODIAGRAMMERISohware Re- A computer provides such niceties as straightening sources Corp., Los Angeles, Calif. This proprietary lines and squaring boxes. However, the more im- program, developed by ARIES Corporation and portant function of the computer in this case is to being marketed throughout the western United incorporate into a program the subroutines or func- States by Software Resources, produces accurate tions called for by the individual symbols on the and standardized flowcharts of programs written flow chart. If the flow chart as drawn by the pro- in IBM System/360 Basic Assembly Language. grammer does not make sense or if necessary in- The use of AutoDiagrammer eliminates a significant formation or linkages are omitted, the computer part of the time needed for program documentation can indicate the discrepancies and ask the program- and shifts the burden of flow chart production mer for clarification. This ability to provide two-way from the programmer to the computer itself."

89 (Computers and Automation 17, No. 2, 47-48 the output copy can occupy several pages if the J

(Feb. 1968).) size of the program being flowcharted requires it. J 5.35b "A new program now available from IBM Even so, they do report a limitation of 200 state- | enables flowcharts to be produced by a System/360 ments; furthermore, they do not indicate how their a computer. Compared with earlier programs of this program goes about recognizing the branches. | type, System/360 Flowchart has more automatic Since they claim an application for their program 1 features for labelling and for linking elements in drawing PERT networks, it is inferred that pre- | of the chart: it also provides for footnotes." (The processing is required by the user." (Abrams, 1968, | Computer Bull. 1 , No. 2, 159 (Sept. 1967). 744). 1 p. I 5.35c "It is hoped to generalize the Flowlister "A program by Stelwagon called FLOW 2 which so that it indexes a complete program, producing produces flowcharts of FORTRAN and FORMAC an alphanumerically ordered list of all common programs directly from the source decks became assignments and subroutine calls, and a map of part of the open literature in August 1966. Actually the program as a whole. It could also be extended this data is deceptive, since the instructions for to handle subroutines written in various assembly using the program are dated one year earlier. An languages. amazing degree of flexibility is available to the user "Except for a few small character-handling of this program, including the option of producing routines, the Usercode Documenter, Auto-Diagram- an intermediate flow deck which directly controls mer and Flowlister are all written in Fortran." the graphical output produced and may be manip- (Roberts, 1967, p. 22). ulated by the user to improve the esthetics of the "One of the early efforts was reported in 1959 by topology. Since the graphical hardcopy is produced Haibt who was concerned with flowcharting machine by an SC-4020, Stelwagon had the problem of ' language. Although not completely clear, it appears page segmenting and referencing to contend with. that extensive of tables He has rejected the flow matrix approach as being use was required to produce \

output on the line printer. Since she felt it unde- too consumptive of storage and has, instead, de- ; sirable to attempt to describe an entire program veloped a reduced matrix which serves to position i; unit in one flowchart, probably because of the branches and boxes in quite a clear output. Using limitations of the line printer and the extensive auxiliary memory, he claims to be able to flowchart detail present in machine language, the decision to program units in excess of 2,000 statements. Various i produce multilevel flowcharts was necessary. The shapes of boxes are provided for identification of

j program featured a sketchy description of the type of operation. In solving the paging problem,

procedure represented by a box. Certainly this out- Stelwagon has brought his branch Hues out both i put was not at all language independent since cross horizontally and vertically. Thus, if a one-piece • references were supplied to program listings. flowchart is desired, one obtains a large rectangular Although mention was made in this paper of drawing mural when the pages are pasted together." flowcharts directly from a FORTRAN program, it (Abrams, 1968, p. 744). appears that this feature was never accomplished." "The present author has developed a flowcharting ii

(Abrams, 1968, p. 744). program MADFLO that produces continuous output i "Knuth, in 1963, reported on a system which on a CALCOMP drum plotter. The use of this

produced flowcharts on the line printer if the completely random-access graphical output device i; assembly code were augmented by descriptive allows the construction of a program which requires d comments. No doubt the inclusion of such com- no tables or arrays to determine the position of J ments is a valuable tool, but it does require special boxes in advance. Following what seems to be a effort on the part of the programmer and does also normal mode, single column output results from restrict generality. Furthermore, his flowcharts were non-branching, algorithms. Multicolumn output is cross-indexed to a flow outline which was in turn produced whenever the program contains decision ^ indexed to the assembly listing. Each level conveyed making branches; the number of columns is Hmited :

more precise information concerning the action by the width of paper available. The input scanner ; within each box. He felt that 'it is unnecessary to operates much like the statement recognition phase specify all the details of a program in the flow of a compiler. Only the syntax of the source language i

outline' and even less so in the flowchart. His statement is examined to determine the shape of I requirement of documentation effort on the part box required to properly represent the operation. c of the programmer does introduce the concept of Since it was desired to make the flowchart using computer-drawn flowcharts as creative tools essentially language independent, the scanner ;

in debugging." (Abrams, 1968, p. 744). deletes certain phrases when the statements are i

"An interesting flowchart-drawing program which lettered within the drawn box. Except for this edit- '- draws a picture of the (undirected) branches of a ing, the entire statement is always lettered out for ' program by disregarding the specific contents of the user. He can of course ignore any lettering that ' the boxes was reported by Hain and Hain in 1965. does not interest him. Some further development of A very elegant connection matrix is employed to the methods of Rich and Stone may be necessary to generate this 'connection graph'. Hardcopy output achieve less arbitrary Une division; also, the i is produced by an SC-4020 microfilm plotter; thus relational operators and assignment indicators li present in the source language character set should by the designer. These include scaling, rotations, be replaced with universally-acceptable symbols translations, contiguity assignments, insertions, perhaps taken from the USASCII character set." replications — all possible on both the parts (sepa- (Abrams, 1968. p. 745). rately within the whole of the drawing) and on the 5.36 "A computer controlled approach to the entire aggregate of features in the drawing." (Cheyd- logical block diagram construction . . . would leur, 1965, p. 174). have the following advantages: "Among the special faciHties available for operation of the display system are provisions for a. Computer storage of a complete logical displaying text, for drawing on the screen, and for design, either on paper tape or magnetic tape. rotating three-dimensional objects." (Fano, 1964, This means that the current design is in one place and prints reflecting this current p. 15). 5.38a "The program 'package' called CALCUL- design would be available for use by all concerned with the design. AID provides a convenient way to specify and execute requests for b. Easy correction or modification to the design mathematical computation and manipulations. It entirely and prompt issuance of revised prints reflect- runs within the OPS-3 ing these changes. system. In addition to a general assignment-statement capability, such as the execution of the c. The built-in checking of the logical operation statement of the design to see if, in fact, it does perform as expected. Also the analysis of the circuit X=Y+A*(Z + 5) for such items as the loading of the output(s) of each unit, and signal delay along various several more specific programs have been provided. signal paths." (Loomis, 1960, p. 10-11). These include a multiple linear or simple polynomial "Although a flow chart compiler is not yet avail- regression program, a solver of polynomial equations able, a graphical circuit simulator . . . has been in one unknown, a count-by-interval program, and programmed for the TX-2 computer at Lin- a program to rank-order a given sequence of coln. . . . This system should be capable of numbers. providing a circuit designer who has no pro- 5.39 "CALCULAID and MAP are two more sys- gramming experience, with a compreltensive simu- tems for using the computer as a mathematician's lation facility which he can use with Uttle instruction helper. CALCULAID is oriented toward writing to supplement his knowledge of the normal symbols, programs to solve large problems with much data. conventions and graphical properties of electrical It has built in FIT and REGRESSION operators, network theory." (Roberts, 1965, p. 216). and a convenient way of specifying matrix opera- "The MIT researchers plan to extend the tech- tions. MAP has facilities for performing convolu- nique from electronic networks to digital systems, tions, Fourier transforms, and other more treating logic gates as components. They will also sophisticated analytical operations." (Teitelman, apply it to distributed systems transmission and 1966, p. 15-16). acoustic Unes. . . . 5.40 "By means of displays, the user receives "CAD includes a unique circuit analysis tech- immediate feedback of the results of his actions, nique called circal, a computer program developed and he is able to modify his decisions in order to by an MIT graduate student. Charles Therrien." obtain a system response that is most relevant to (Electronics 39, No. 3, 39-40 (1966).) his needs." (Borko and Burnaugh, 1966, p. 1 "A typical use of AEDNET involves drawing (abstract).) the circuit diagram on the scope, typing or plotting "These programs can be used singly, to perform values and characteristics of the components, and individual tasks or they can be combined together requesting calculation of the response to a specified in 'programs', where each step of the program is input. At the user's request, the program displays one of these macro operations. The system can be the various network variables (currents, voltages, used either as a very large and fast 'desk calcu- charges, and fluxes) as functions of time or as func- lator' or as a powerful programming device which tions of each other. The user can at any time modify allows simultaneous creation and execution of the network and ask for a recalculation and redis- programs." (Wantman, 1965, p. 8). play." (Evans and Katzenelson, 1967, p. 1135). "By providing instantaneous response and a 5.37 "The major use of the console system so multiplicity of input controls, the operator, on his far has been for problems requiring at some stage CRT screen, is able to create and manipulate a 'dynamic scratch pad' capabilities. Included in model of his problem in the most descriptive and such problems have been printed circuit component meaningful manner. With a graphic display, the and wiring placement, schematic circuit design, user is given complete freedom in the presentation block or flow diagram design, text composing and of his data, choosing whatever pictorial form he ." editing . . (Ninke, 1965, p. 845). desires. Rich, problem-oriented, graphic languages 5.38 "With this [Sketchpad] system a wide can be tailored to each user's needs, allowing full variety of structural transformations and agglom- statements to be made with the flick of a fight pen. erative actions can be evoked extemporaneously Concepts can be depicted in entirely new ways.

91 366-108 O - 70 - 7 Data can be condensed and presented in clear "The term CAI should be reserved for those par- understandable graphs." (Stotz, 1968, p. 13). ticular learning situations in which a computer "Most graphic input-output systems are involved contains a stored instructional program designed in computer-aided design — a class of environments to inform, guide, control, and test the student until involving a high degree of interaction between a prescribed level of proficiency is reached." designer and problem-solving system. Computer- (Silvern and Silvern, 1966, p. 1651). supported drawings are a good medium for com- 5.42 "A much older CAI system is the PLATO munication between human beings and design System, developed at the University of Illinois. systems because drawings are easily understood In the present system, each of the 20 student sta- by people. The graphics problem is not so much tions connected to a CDC 1604 computer has an automating the drafting process, as automating the electronic keyboard for student communication interpretation of drawings." (McCarthy, 1967, p. 64). with the central computer and a CRT screen for

. . The engineer must be able to communicate viewing information selected by the computer. with the computer in an interactive environment This CAI system has been applied to the actual where he can: teaching of electrical engineering and nursing." (Van and Michener, 1967, 199.) 1. Request an operation. Dam p. of the project has 2. Examine his results. "The purpose PLATO been to 3. Determine the next operation to be performed develop an automatic computer-controlled teach- based on the previous results. ing system of sufficient flexibihty to permit experi- "This suggests that a problem-oriented language mental evaluation of a large variety of ideas in functions best in a time-sharing environment." automatic instruction including simultaneous tutoring of a large number of students in a variety (Roos, 1965, p. 425). of subjects." (Coordinated Science Laboratory, 5.40a Two examples are: "To correct an input 1965, p. 43). and to immediately signal the man in the loop for "Bitzer & Easley describe a particular automatic a correction or the need for a correction, is an teaching system (PLATO) in use at the University important time saving factor. Many needless hours of Illinois. The interface language of PLATO ex- of 'turn around time' can be saved by finding out hibits an interesting approach to the problem of immediately that a transcription error exists." smooth coupling to the problem environment (Bauer, 1965, p. 18). (example: the student can interrupt the main in- "A most important form of output is interim structional sequence for some special tutoring by feedback, which will sharpen the focus of the ques- pressing a button labeled "Help": a button labeled tion after the search has started." (Taylor, 1962, "Aha!" returns him to the main sequence). Among 393). p. the author's conclusions are an indication that the It is to be noted in particular that system is actually performing as a successful edu- "A communication network that does not let the cational device and that the same system imple- user know long it will be before his message will how on an appropriately large would mented computer i< be delivered to the end addressee may be theoretically provide this kind of teaching to as many as a thou- i oscillatory.'''' (Baran, 1964, p. 7). sand students simultaneously." (Mills, 1967, p. 238). i problem in 5.40b "The man-machine coupHng "PLATO, which stands for Programmed Logic its full breadth is the central issue in advancing

for Automatic Teaching Operations, is another i interactive-computer technology." (Mills, 1966, p. laboratory for educational research. It uses an 'i 196). ILLIAC computer-controlled system of slides, TV 5.41 "Application of man-machine systems in displays and student response panels for simul- ( support of education already promises to be among taneously instructing a number of students. ... It |i the most important in terms of future impUcations. is part of the Coordinated Science Laboratory of |'

' From a philosophical standpoint, problems of edu- the University of Illinois. ... cation can be included as part of the so-called "In the PLATO system, the computer accepts :: 'information explosion' problem area, since the each student's request in sequence, but because body of information to be transferred by the educa- of its high speed students can be served without tion process must ultimately reflect the growth of noticeable delay. Each student has a key set and a / (Mills, the aggregate of human knowledge." 1967, television set. The key set, used by the student to L

p. 235). communicate to the central computer, has keys of 'j

5.41a In another paper, these authors com- two types: character keys, such as numerals, letters ^ ment: "While Programmed Instruction, or PI, has and other symbols; and logic keys such as "con-

been evolving for a number of years, Computer- tinue", "reverse", "help", "judge", and "erase", t Assist Instruction, or CAI, is a much more recent logic are used the students to proceed These keys by |

arrival. This is because the computer itself is a through the instructional material. From the stu- '

relatively new device and time-sharing, which makes dent's response or from the position he has reached 1 CAI economically feasible, is only now being applied in the program, information is selected by the '

on a large scale in commercial systems." (Silvern computer for display on the student's television |ii and Silvern, 1966, p. 1651). screen. This information has two sources: (1) a h

92 central slide-selector, and (2) an electronic black- the instructor. A maximum of 32 stations can be board. .Although a single sHde-selector is shared by linked to the computer for simultaneous instruction. all students, each student has independent access This system has been used at the Brentwood School, to any slide. Each student has an electronic black- in East Palo Alto, California, to present mechanized board on which the central computer can write reading and mathematics (e.g., set arithmetic) to characters and draw diagrams. Information unique first grade classes." (Van Dam and Michener, 1967, to a student such as his answer to a question, is p. 199). written on his own blackboard by the computer. "The IBM 1500 Instructional System is an experi- The images from the slide-selector and the black- mental system for computer-assisted instruction, board are superimposed on the student's television designed to administer individual programmed les- screen." (Egbert, 1963, p. 622). sons to 32 students at once. Working through one 5.43 "CLASS stands for Computer-Based Lab- or more teaching devices at his own instructional oratory for Automated School Systems and is a station, a student may follow a course quite different facility developed and operated by Systems Devel- from, and independent of, lessons presented at opment Corporation. The CLASS faciHty is not other stations. Instructional programs stored in considered a prototype of a school system but is central files control lesson content, sequence, devoted rather to educational research in a system timing, and audio-visual medium, varying all of context. The laboratory permits the integration of these according to the student's responses." (Terlet, (1) individual student automated instruction, (2) 1967, p. 169). group instruction, both automated and conven- 5.45 "This means that a chemistry student can tional, and (3) centralized data processing for admin- sit at his desk and type directions to a machine that

istration, guidance, and planning functions. For will 'perform' chemical experiments step by step . . . flexibility, the central control of the laboratory is The student has a viewer on his desk that shows him exercised by a digital computer, the Philco 2000. what would happen, for instance, if he orders potas- Various innovations in instructional load, adminis- sium chlorate to be added to a solution of hydro- trative data processing, and counseling procedure chloric acid. If the solution turns yellow or the can be provided by computer program modifica- mixing causes an explosion, the student sees this tion." (Egbert, 1963, p. 621). on his viewer." (Bushnell, 1965, p. 19). 5.43a "PLANIT (Programming L.4A^guage for 5.46 "The CAI system has the capabihty of /nteraction and Teaching) is a system employing collecting student data, and this can be processed a flexible language designed for CAL PLANIT and regurgitated in the form of reports to the teacher provides for the designer a powerful and flexible and administrator. The system can provide rates of tool for entering communication material into the progress, areas of difficulty, problems, unexpected computer, for modifying the material, for presenting responses, time to complete, adherence to schedule, it to the user, and for prescribing the behavior of attendance facts, etc." (Silvern and Silvern, 1966, the computer as a function of the user's current p. 58). response and past performance. In addition, the "A cumulative record is kept of the student's language is simple enough to allow a nonprogrammer performance on each of the topics. These perform- to use it easily." (Feingold, 1967, p. 545). ance measures include the student's response "Materials produced via PLANIT are: Instruc- time, error count, and pattern of errors. If a stu- tional sequences in statistics (aimed at students of dent's performance falls below an acceptable level the social sciences enrolled in a first course in for a topic, the student is branched or detoured to statistics); spelling and vocabulary (for children a special set of remedial items on that topic." three to eight years); economics (for undergradu- (Egbert, 1963, p. 621). ates): introduction to computer programming (for 5.46a "A few commercial speech output devices persons having some knowledge of algebra). Agen- for computers or computer controlled systems have cies who have used or are using PLANIT are: been announced. All but one of these rely on the System Development Corporation, Southwest Re- playback of analog audio signals under digital gional Laboratory, University of California at Los control of selection, switching, and sequencing. Angeles, University of Southern CaHfornia, Uni- The analog recording has been both magnetic and versity of California at Irvine, United States Naval photographic. These devices generally have in- Personnel Research Activity (San Diego, California), dividual words in uniform time slots and can New England Educational Data System, and Lack- provide sequencing of these words under computer land Air Force Base." (Feingold, 1967, p. 551). controL They have been used in such applications 5.44 "The IBM 1500 Instructional System is as providing stock market quotations and telephone indicative of the complexity inherent in general- number change information." (Hogan, 1966, p. 92). purpose CAI hardware. The 1500 hardware consists 5.47 "The IBM 7770 audio response system

of CRT display units, alphanumeric keyboards and furnishes spoken replies to telephone inquiries . . . light-pens for student input, image projectors, tape- The inquiry is processed by the computer against recording units, a central processing unit (the new stored data. The reply is composed from a selected, IBM 1130), an audio adapter for sound transmission prerecorded vocabulary and transmitted back to to student stations, a station control, at least two the caller as a spoken message." (Licklider, 1965, disk units, and a printer and card reader for use by p. 102). 93 "In the case of the 7770 the enquirer enters a one language might be used within the same system request by dialling digits which are immediately at the same time. In the United States this is not a transmitted to the audio response unit and then sent major problem but in Canada and in some European to the processor. The computer processes the data, countries where people are often times multilingual, composes a coded reply, and returns the message it is only right to assume that one should have a to the 7770. This device then interprets the coded computer that would be too. Such a device is the reply, selecting the appropriate words from a stored IBM 7772 Audio Response Unit." (McDonald, 1966, vocabulary of between 32 and 128 words, and tells p. 54). the enquirer the answer to his query." ("IBM Sys- 5.49 "Cognitronic's 'Speechmaker' audio re- tem/360", 1965, p. 299). sponse units also have a fixed vocabulary of words

"The 7772 . . . generates audio on the 'vocoder' selected and combined by computer command. principle — that of energizing tone filters and com- These units have a maximum vocabulary of 189 bining the output result first to form words and then words. Each word is 1.6 seconds long and is pre- ." sentences . . (Urquhart, 1965, 857). Knauft recorded, by a technique similar to that used to p. [ et al. (1966) describe further developments with produce a sound track for a motion picutre, on a respect to the 7772 audio response unit involving cylinder of photographic film mounted on a drum hardware and software features for treatment of in the response unit. A narrow light beam, directed unvoiced components of coded speech signals. through the sound tracks on the rotating drum, is <

"The 7772 can have a vocabulary of several modulated by the prerecorded audio on each track | thousand words, stored on a disc device in digital and is detected by photocells. The output of the form. The audio response unit recreates the spoken photocells is ampHfied and fed directly to telephone voice from these digital records." ("IBM System/ lines. This system is in use at the American Stock j

360", 1965, p. 299). Exchange, enabhng stockbrokers to dial four-digit i 5.48 Vocoder techniques, which were initially codes on their office telephones and receive stock described by Dudley in 1936, involve devices de- quotations; it can answer 1200 telephone inquiries > signed to compress "the bandwidth of speech in a minute." (Van Dam and Michener, 1967, p. 193). signals in order to transmit them over channels of 5.50 "RCA announced that it has developed a ji very Hmited capacity. The vocoder measures the new voice response unit . . . [providing] vocal ; speech power in a number of frequency bands and replies to queries fed into the computer by tele- ; transmits these measures as signals over a set of phone from up to 100 remote points. It is designed ! narrow low-frequency channels ... At the re- to work with the Spectra 70/35, 45 and 55." (Com- ; ceiver the speech is reconstituted by modulating mun. ACM 9, 468 (1966).) the spectrum of a broad band source in accordance 5.50a "Burroughs Corporation has announced i with the frequency region and amplitude of each a data communications system which uses com- i of the measure-signals derived from the original puter-stored information to generate a spoken reply

. . . re- speech. Normally this reconstituted speech signal through a telephone hand set The voice ; is presented acoustically for a listener. Alterna- sponse generator's memory is a rotating, cyhndrical, i

tively, vocoded speech signals can be presented photographic film containing 64 parallel tracks on ij visually or tactually." (Pickett, 1963, p. 2). which numbers, words, phrases, and silence are recorded. Photoelectric cells read the 64 tracks as An alternative to the vocoder approach, also j they pass or numbers are selected, investigated by IBM, involves the development of by, and words ^ program, in the a library of analog synthesizer control signals cor- as instructed by the computer t responding to subword segments of speech. Thus, proper sequence to compose a message. One track • "segments of control signals can be made which is silent and provides timing and pauses desired in i correspond to segments of synthetic speech, and messages. The other 63 tracks can contain three i a hbrary of speech segments can then be stored in words or numbers, each of one-half second duration ^ or a phrase up to one and one-half seconds long. The the form of a library of synthesizer control-function \ tracks, into thirds, can contain a segments. The synthesizer may then be controlled 63 segmented , numbers." by composite functions assembled from the se- maximum of 189 individual words and s (Commun. ACM 11, No. 4, 289 (Apr. 1968).) quence of control-function segments corresponding [ to the desired utterance. In order to investigate this 5.51 "Audio equipment enabhng a computer approach to speech synthesis, a terminal analog to talk to students has been delivered to Stanford ; synthesizer and associated control system have been University by Westinghouse Electric Corporation, } constructed. Our goal is to generate a library of US. It uses a Prodac 50 computer that controls a ; reproducible control-function segments from which battery of 12 random access magnetic tape units we can construct connected speech." (Estes et al., on which verbal information is recorded." (Elec- ?

1964, p. 2-3). tronics Weekly, No. 258, 8/11/65, p. 11). i 5.48a It is to be noted that both the IBM 7770 5.51a "The RADC Information Processing Lab- 5 and 7772 units are available in languages other than oratory has sponsored research on a digital to analog i

Enghsh (Urquhart, 1965, p. 857). converter technique for speech. A breadboard 1!

"Of particular advantage is that the technique is model demonstrated feasibility of converting digital t language independent, and if so desired, more than information into spoken words using optical sound- !i track techniques and using a vocabulary of 385 even to 300 bps but with markedly reduced quality." words." (Shiner. 1962, p. 338). (Franco et al., 1965, p. 131). 5.52 "Pushbutton input and audio response can "Fairly good quality speech can be produced be combined in the home use of computers. IBM using about 48,000 bits per second. Even with the is now experimenting with the student use of com- vast economical digital storages available today, puters for homework problems, using home tele- the bit storage (using the minimum acceptable phones to link to a 1710 computer; spoken answers sample rate and resolution) for a 1000 word vocabu- are generated by a 7770 response unit." (Van Dam lary is not particularly attractive. . . . and Michener. 1967. p. 193). "Certain economies of storage can be achieved by 5.53 "A typical dial access medical tape re- the use of the channel vocoder mainly because of cording library is in operation today at the University the redundancy of human speech. That is to say, of Wisconsin Medical School. Staff physicians every instant of speech is not a wholly new complex record four-to-six minute commentaries on current waveform but is generally the repetition of the same information having to do with various subjects or complex wave form a number of times and then procedures. The list of subjects is circulated by changing to another. By taking advantage of this mail to physicians in the state. A practising physi- unique quality of speech, it is possible to digitize a cian may telephone ... at any time of day or portion of speech, a word, inspect it for its redun- night and request that a particular tape be played." dant aspects and rather than store all of the redun- (Becker. 1967. p. 8). dancy, establish a count of the number of times a 5.54 "The Bankers Trust Company noted both given pattern is used before it changes to another favorable response and rapid adjustment of bank pattern. . . . personnel and the public to its audio response "By the use of this technique it is possible to installation." (Swanson, 1967, p. 38). store reconstructable digitized speech so that 5.55 "Audio-response techniques are being approximately twenty-four hundred bits represent used at the Space Division of North American one second of speech [one to two words]. This is a Aviation for engineering-design status information considerable saving in space over the minimal on NASA's Apollo and Saturn projects. In this case, 24,000 bits and the desirable 48,000 bits." (Mc- the words in the audio vocabulary are the appro- Donald, 1966, p. 53). priate ones to provide information about engineer- 5.56 Under the general direction of G. Fant ing drawings and specifications." . (Whiteman, (1959), the work of J. Liljencrantz, U. Ringman, 1966. p. 67). P. Tjernlund and others of the Vocoder Group, "Another system is EDICT, which the space and Speech Transmission Laboratory, R.I.T., is pri- information systems division of North American marily concerned with speech analysis and speech Rockwell uses to provide the company's engineers synthesis. Both a speech analyzer and a speech with design data. Data from more than 75,000 engi- synthesizer are in operation can be used together neering drawings is available from any of the com- to improve the vocoding performance. A new spec- pany's 42,000 telephones via an IBM 7770." (Berul, trum analyzer has been designed for on-line opera- 1968, p. 29). tion with a computer, including such tasks as 5.55a "The ability of a computer to generate varying formant and anti-formant information, i.e., answers in audible words makes it possible to obtain to take samples of natural speech, to extract param- a great deal of information in a form understood by eters, and to try to replicate them with the speech anyone and with the use of unsophisticated and synthesizer. (See Stevens, 1968). therefore inexpensive terminal equipment. A 5.56a "A speech synthesis program was written number of computer companies have designed at Manchester (Mathers, 1964) based on the Para- inquiry devices. Telephones which use tones for metric Artificial Talker at Edinburgh University diahng can also be employed for this purpose by (Anthony and Lawrence, 1962). The eight param- using the pushbutton tone dial after the call is eters required by this program were measures of established to submit the inquiry. In this case, the the frequency of relatively dense energy bands answer is heard in the telephone receiver." (Gentle, in the spectrum, energy ampUtude measurements, 1965, p. 86). and measurement of the fundamental frequency 5.55b "It has been estimated that the informa- of the waveform. Two other synthesis programs tion content of the digital voice signal is as low as 50 were developed which required nine and eleven bits per second. Direct digital encoding of an analog parameters, representing a closer description of speech signal by sampHng, quantizing, or pulse the spectrum. Three analysis programs were written code modulation techniques requires a bandwidth which automatically measured the spectrum and of 56,000 bps. With fixed channel vocoding tech- produced the appropriate parameters from the niques, the bit rate required to transmit . . . speech waveform." (Lavington and Rosenthal, [speech] in digital form with adequate scores on 1967, p. 338). intelligibility, articulation, and speaker recognition "Routines r6 and r7 were developed for use in tests, has been reduced to 2,400 bps. There are a system of automatic speech analysis and synthesis other vocoding techniques which will eventually (Rosenthal, 1965). This system was based on the reduce this bandwidth stiU further to 1,200 bps, and assumption that the intelligible information in

95 speech may be expressed as values of a few slowly high degree of accuracy." (Van Dam and Michener, varying parameters, and that these parameters 1967, p. 194). may be measurements of characteristics in the 5.59a "It should be clear that with the addition spectrum of the speech waveform. The object of of a printer character-recognition unit, the system the exercise was to obtain a reduction in the data is useful as a reading machine for the blind. It rate of speech transmission without loss of infor- was originally with this purpose that our research mation. It was desired to produce analysis pro- into computer generation of connected speech from grams which would form the spectrum of the speech text source began." (Lee, 1968, p. 333). waveform and measure the values of several param- 5.59a2 "Formant amphtudes have not been eters which would then be used as input to a syn- studied much in previous investigations Formant thesis program. The synthesis program would frequencies and over-all intensities have received produce a waveform whose spectral characteristics much more attention. ... A correct reproduction were controlled by the parameters, and which would of formant amplitudes in synthesis is necessary hopefully, be judged similar to the original by for ensuring one of the elements of naturalness, human listeners." (Lavington and Rosenthal, 1967, such as, for instance, a voice source of correct spectrum shape." (Fant et al., p. 338). 1963, p. 1760). 5.59a3 "It is a mistake to suppose that, once 5.57 "A voice-excited vocoder has been suc- vowels and consonants have been simulated cessfully simulated on the IBM 7094 computer and in isolation, their conjunction will result in in- satisfactory quality of voice has been reconstructed telligible speech. If the formants are varied in a with 220—660 cps frequency used as the voice- stepwise fashion, the resulting excitation band. Experiments are under way to sounds are quite unintelhgible unless convert the vocoder output to a binary data stream one knows what was intended beforehand. In natural and investigate the effect of channel coding tech- speech, the formants vary slowly niques on the reconstructed voice." (Quarterly and smoothly, even persisting through seem- ingly colourless sounds like 'sh', they Progress Report JVo. 80, Research Laboratory for where form a smooth interpolation between the preceding Electronics, M.I.T., 179 (1966).) and succeeding values occurring in the speech. But > "The general research problem of artificial speech with sufficiently fast machines, all these things generation is under continuing investigation in could be allowed for by program." (Woodward, Professor K. N. Stevens' Speech Communications 1966, p. 262). Group and particular application to the reading- "Perhaps the intonation of the voice will be the machine system is being considered in coopera- most resistant problem; we have no notation for its > tion with his group." (Quarterly Progress Report inflexion and Httle understanding of the rules by i No. 80, Research Laboratory for Electronics, ' which inflexion conveys meaning. This, it may be M.I.T., 218 (1966).) suggested, is why computer voices are often pro- 5.58 is "Study of the speech process interesting grammed to sing rather than to speak." (Woodward, in its own right. It is, moreover, a route to practical 1966, p. 262). applications, not only in the engineering of devices 5.59b "The resulting speech is usually intelligi- \ for speech recognition, bandwidth compression, and ble, but not suitable for long-term use. Several s

prosthetic aids to the blind or deaf, but also to the problems remain, apart from those concerned direct- 1 more common problems that arise in connection rule from phonemic ly with speech synthesis by i, with the use of language by normal and handicapped specifications. First, many words can be nouns or /( people. The research on speech has necessarily verbs, depending on context [refuse, incline, sur- |i

been interdisciplinary, requiring close working vey], and proper stress cannot be specified until the j

cooperation among linguists, experimental psy- intended syntactic form class is known. Second, j chologists, engineers." (Cooper, and 1967, p. 1214). punctuation and phrase boundaries may be used l{ 5.59 "An interesting application of audio output to specify pauses that help to make the complete is the work at MIT's Research Laboratory of Elec- sentence understandable. Third, more complicated tronics blind to develop a 'reading' system for the stress contours over phrases can be specified which J

that will read printed English text aloud. The facilitate sentence perception. Finally, intonation > project is divided into three phases: recognition contours, or 'tunes' are important for designating of characters; translation of words into 'phonemes,' statements, questions, exclamations, and continu- the minimum units of speech; and conversion of ing or terminal juncture. These features (stress, ^ the phonemes into speech, using a speech synthe- intonation, and pauses) comprise the main prosodic f sizer. To date, phases one and three have been or suprasegmental features of speech." (Allen, completed. These modules have been built at the 1968, p. 339). , Laboratory and have cost only a few thousand 5.59c "The parser makes two passes (left-to- dollars, since special high-speed equipment was right) over a given input sentence. The first pass necessary. The character recognizer uses TX-0 computes a tentative bracketing of noun phrases and PDP— 1 computers to perform the recognition. and prepositional phrases. Inasmuch as this initial It can now scan printed input, recognize each letter, bracketing makes no clause-level checks and does and output the character vocally. The overall read- not directly examine the frequently occurring noun/ ing rates are about 60 to 80 words a minute, with a verb ambiguities, it is followed by a special routine j;

i 1 designed to resolve these ambiguities by means of "The speech signal being displayed is cut in the local context and grammatical number agreement same manner that it would be for splicing. However, tests. These last tests are also designed to resolve in this case the portion of the waveform between the noun/verb ambiguities that do not occur in bracketed "cuts' is played back to the operator on head phones. phrases, as [refuse] in [They refuse to leave.]. As a The selected portion may be played back a single result of these two passes, a limited phrase bracket- time, or (under operator control) it may be joined ing of the sentence is obtained, and some ambiguous to itself and repeated continuously as many times words have been assigned a unique part of speech, as desired. This feature is particularly useful for vet several words remain as unbracketed constitu- listening to repeated versions of single periods in ents." (AUen. 1968. p. 341). voiced speech which can be used to determine 5.59d "Another possible approach is to use boundaries of voiced dyads. partial words as the the basic unit to be stored. "Finally, a word should be said about the problem Research in this area has been very Hmited. Stowe of intonation. In this experiment, we have not and Hampton have constructed connected speech attempted to produce any prosodic features. This from syllables and have reported intelligibility problem could be handled in two different ways. scores of 90 per cent when the speech was pro- First rules could be developed for modulating the duced at the relatively slow rate of 2 words per pitch, amplitude, and duration of the stored dyads, second. hen the rate was increased to a more according to the intonation needed in a given mes- natural 4 words per second intelligibility dropped sage. These rules would presumably be similar to to 66 per cent. Harris attempted to construct those being developed in various speech model speech by using the basic sound of English as his synthesis systems. Secondly, the dyad inventory stored units: that is he stored (in analog form) sounds could be expanded to include several representa- representing P. T. A. and so on. these having been tives of each of the dyads, some with rising inflec- cut from words in which these sounds occur. Since tion, some with steady high pitch, etc. Wang and it has been long recognized that production and Peterson have calculated approximately 8500 perception of sounds are affected by adjacent segments would be required to represent all of the sounds, it was necessary in his method to have inflections used in English. Although this would several representatives of many of the sounds. That greatly expand the work required to construct a is, a different kind of O might be necessary if it dyad system, the technique is relatively straight- were to follow a T than if it were to follow a K, and forward and might be justified in the case of a user so on. Using a small inventory of such sounds, plus who needed very high-quality, natural speech." rather elaborate analog equipment for synthesizing (Becker and Poza, 1968 p. 795-800) References are monosyllables from the stored sounds. Harris ob- to Stowe and Hampton (1961). Harris (1953), and tained approximately 75 per cent intelligibility. Wang and Peterson (1958). "An approach somewhere between the use of 5.59e In this connection, Dolby et al. point out single sounds and the use of syllables is the use of that "solutions to the affix and syllable-counting dyads that approximately represent the last half of problems are essential to the determination of an one sound and the first half of a following sound. algorithm for transcribing written English into any The advantages of dyads are that they include the of the standard phonetic transcription systems now natural transitions to of speech from sound sound, in use." (Dolby et al., 1965, pp. 1-15). while still to keeping the required inventory a man- 5.59f "Speech synthesis by rule is the method of ageable size. \S ang and Peterson, on theoretical converting from a discrete representation of speech grounds, have calculated that an infinitely large in linguistic units, that is, phonemes and stress message set could be produced using approximately marks, to a continuous acoustic waveform . . . 800 dyads for monothonic speech and approxi- The discrete input is converted to continuous con- mately 8500 for dyads intonated speech. Assuming trol signals by the synthesis strategy. The synthesis that a dyad lasts for 50 to 100 ms, it can be seen strategy contains stored information about the that even using high-quality PCM coding of the phonemes and stored rules about the mutual effects waveforms, an 800-dyad inventory would require of adjacent phonemes. The stored rules operate only about 5,000,000 bits of storage . . . on the input sequence to produce the control signals "The decision to use natural rather than synthetic for the synthesizer. The speech synthesizer converts speech to obtain our dyad inventory necessitated the control signals to continuous speech. The the development of a system to extract portions of synthesizer may be a terminal analog, a dynamic a speech waveform and store them for later recall. analog of the vocal tract, or a combination. . . . It was necessary to isolate segments of speech "Each phoneme has a characterization indepen- as short in duration as 40 ms. Splicing speech seg- dent of adjacent phonemes. The characterization ments to this degree of precision was accomplished includes format information, source characteristics using a research tool developed at the Sensory in the production of the phoneme, a description of Science Research Center. A CDC-.3100 computer, whether it is nasal or fricative, and a set of fre- coupled to a high-performance, interactive display quency regions surrounding the formant positions. system, has been programmed to provide a visual "The formant information is a set of target posi- display of a previously digitized speech signal . . . tions for both center frequency and bandwidth of

97 formants one, two and three. The source character- vices. A good tune was, to the best of the author's istics describe the condition of the vocal cords knowledge, first played on the Ferranti Mark I during the production of the phoneme. If the vocal computer at Manchester, with Mr. C. Strachey at cords are vibrating the sound is voiced. The fre- the controls. Nowadays, with faster computers, it quency regions of a phoneme represent the degree is possible to play several good tunes at the same to which certain acoustic parameters must approxi- time. S. N. Higgins produced a common chord from mate the target values of these parameters in the a computer (TREAC) by time-muhiplexing as long context of connected speech. In an articulatory ago as 1954. The compilation of the necessary in- analog, the corresponding concept would be the structions made a more starthng sound than the to climatic extent which a given vocal tract configuration chord itself." (Woodward, 1966, p. 257). must approximate the target configuration for the "As the preparation of the tape for a piece of phoneme . . . music can be most time-consuming and the final certain "For consonants maximum durations are performance is of little serious value, the effort specified. Consonant duration (as measured from is unrewarding unless a suitably convenient code human speech) is not a fixed quantity but is very is devised. The code should allow the use of variable- dependent on context. (For example, initial con- length notes referred to a diatonic scale modifiable sonants are much longer than medial consonants.) by sharps, flats and naturals. Rests can be input as The synthesis strategy generates consonants whose notes of zero frequency. The score is best stored in duration is variable within certain limits. Maximum its entirety with all variables ready converted into durations are specified to prevent the consonant the form needed by the program at runtime. Dur- from being unnaturally long, hence objectionable. ing play, a pitch transition in any one part neces- sitates resetting a pitch constant in the program, "Among the topics that will be considered for and as the various parts have unrelated audio future work are the effects of stress and rhythm on phases, no time for such resetting can be found timing of an utterance, the inclusion of more than without interrupting the other parts. If all the other one breath-group in an utterance, and studies of parts move at the same time, the break is insuffi- further correlates of word boundaries." (Rabiner, ciently long to destroy a legato effect, but if one part 1968, 17, 23-24, 36). pp. is tied, its continuity is spoiled unless the resetting 5.59g See Lee (1968). In summary, computation is carefully timed and short enough not "F. F. Lee of the Cognitive Information Process- to interrupt it. The author has found it necessary ing Group has concluded that the technical advance to draw the fine before incorporating this last re- in low-cost high-density read-only storage has finement, which is only the first of what could made it feasible to consider the imitation of the easily become an evermore demanding pastime." human reading habit by machine and is now de- (Woodward, 1966, p. 261). veloping a scheme for automatic translation of 5.60b "At present, the generation of sound by English text from letters to phonemes. Specifi- computer affords the researcher and composer at cally, a lexicon containing on the order of 32,000 least three interesting possibiUties: (1) experiments selected morphemes and words can be used together in auditory perception — characterized by Babbitt with algorithms to give phonemic translation for as 'the most refractory of areas', (2) the study of a vocabulary equivalent to what is contained in musical 'grammars', (3) the development of original the Collegiate Dictionary." (Cornew, Webster New compositions. The greatest amount of activity has 88-89). 1968, pp. occurred in the latter category. Early credits go to 5.60 "Spelled speech offers a slower but simpler Max Mathews and the MUSIC IV program for auditory display than artificial speech. Under sound-generation and to Lejaren Hiller at the Uni- consideration are magnetic and optical storage versity of Illinois. More recently, computer genera- and read-out of the spoken-letter sounds, and under tion has been carried out at Princeton (by James investigation is the compression of spelled speech Randall. Godfrey Winham, Hubert Howe, and by sampling in sections, temporal truncation, and others), at Yale (by James Tenney), and at M.I.T. temporal-overlap superposition. . . . (by Ercolino Ferretti and by A. Wayne Slawson)." "Errors in phonetic speUing, arising from erro- (Forte, 1967. p. 328). neous character recognition and resulting in pho- Forte suggests further that "computer-generated netic nonwords, can be accommodated by spelled- resources for the editing speech display of the erroneous English word. graphic displays offer new procedures that are central to much work in musi- Spelled speech is also a reasonable possibility for phonetically intractable proper names." (Quarterly cology" (1967, p. 329). Progress Report No. 80, Research Laboratory for 5.60c "Experiments at Stanford University, Electronics, M.I.T., 219 (1966).) Stanford, Cahf., indicate that a computer fitted 5.60a "The work reported here was concerned with a special device can be used to teach pitch with what we might call the pianola problem, that to singers. Prof. Wolfgang Kuhn of Stanford's is to say, the use of the computer to reproduce a Music Department, writing in the fall issue of sequence of predetermined sounds from a pre- The Journal of Research in Music Education, viously handwritten 'score' without the use of reported favorable results with an experimental specialized peripiieral equipment or analogue de- 'pitch extraction device'. The device was developed 98 as a laboratory' research tool by two IBM engineers, an oscilloscope tube. This work is apparently ex- Jerome D. Harr and Bruce Moncreiflf, for use with tendible to a wide range of structural representa- IBM's 1620 computer at the company's Los Gatos tions. Consideration of the problems associated Advanced Systems Development Division. with storing, interlinking, retrieving, and trans- "The computer is programmed to evaluate a lating the data to compose a three-dimensional series of musical tones by comparing a student's model on the scope provides an intriguing view of pitch to the true pitch. It can distinguish between data-handhng capability that will be required in the and female voices, which it translates as future. a will useful , male How widely such technique be "high" or 'low'. It also can detect deviations from depends upon the availability of required data in perfect pitch from as much as four per cent to as the literature, but if the techniques of retrieving and little as half a per cent. When a student is ready manipulating the data are not developed, the data to practice with the computer, a tape-recorder can are apt never to be generated and used." (Tate, offer him a sample pitch or he may play a series 1967. p. 295).

of notes on an organ-like keyboard which produces "We . . . decided to develop programs that a true tone. Then he sings into a microphone the would make use of a man-computer combination to tones which the computer has printed out to the do a kind of model-building that neither a man nor beat of a metronome. a computer could accomplish alone. This approach "Prof. Kuhn programmed a series of sight- implies that one must be able to obtain information and introduce changes in the i singing exercises together with tests and options from the computer into computer's memory with the help of his way the program is running in a span of time that is I the assistant, Raymond AUvin. He used 10 under- appropriate to human operation. This in turn sug- graduate music students from San Jose State gests that the output of the computer must be pre-

College. They were instructed in the use of the sented not in numbers but in visual form . . .

; machine, and given a standard musical achieve- "In working with molecular models we are inter- ment test. Then they went through a three-hour ested in being able to obtain data quickly in order 1

I period on the machine, spread over a three-week to evaluate the effect of changing the input variables

period, after which they were again tested against of the program . . . We realized that our best hope the same (.AJiferis) achievement test. of gaining insight into unexpected structural rela- ' "The students noted an increased, awareness of tionships—relations that had not been anticipated —

i pitch and accepted the machine's criteria of perfect lay in getting the computer to present a three-

pitch. Prof. Kuhn said IBM's pitch-extraction dimensional picture of the molecule . . ."(Levinthal, I

ji device worked well: only once in 10.000 tries did 1966. pp. 49-50). ^ it give an erroneous report. "Each note in the per- "During the last year we have written a set of formance was sampled'. Prof. Kuhn reports, 'and programs which allows the construction, the dis- the pitch analyzed. The computer decided whether play, and the analysis of macromolecules using a a specific exercise was to be repeated, or whether digital computer and an osciDoscope display. With

' the student should go forward in the program or this computer controlled display and real time rota- repeat similar material for more practice.' Each tion of the projection of the molecule being dis- I

1 student's-records of performance are collected and played, it is possible for the observer to obtain stored in the computer memory so that his progress a true three-dimensional visualization of the

can be accurately measured and help given where molecule. . . . I ' needed. "Since even a small protein molecule may contain ' "Prof. Kuhn, who used a small Music Depart- 1500 atoms, the total number of pairs to be tested ment allocation for his study, foresees the day is very large. In order to avoid the enumeration of when a complete curricula in melodic, rhythmic all possible pairs, a set of programs has been written and harmonic sight-singing and dictation could which divides space into cubes and then makes a

be computerized and adapted for any level of list of all atoms in each cube. In this way it is pos- musical development. He also envisions instruc- sible to enumerate all pairs of neighboring atoms by tional systems for learning to play instruments." searching through the cubes and listing those pairs (Computers and Automation 16, 61 (Dec. 1967).) in which one member is in a cube and a second 5.60d "The display can also represent a three- member is in the same cube or one in the 26 sur- I dimensional object which can be automatically rounding ones. By using this list processing pro-

rotated around any given axis keeping the drawing cedure it is also possible to define 'holes' in a in proper perspective as it rotates. The ability to molecule by locating empty cubes surrounded by display a three-dimensional perspective drawing filled ones. In addition, 'insidedness' and 'out- and rotate this drawing to show the view from other sidedness' of the molecule can be determined by vantage points is particularly interesting in archi- finding cubes which have filled neighbors on one ." on the other side. tectural design. . . (Hobbs, 1966, p. 1873). side and empty neighbors The 5.61 "Levinthal has recently started to develop knowledge of these aspects of the topology of a a system for sequentially displaying two-dimensional folded protein are necessary in order to evaluate images of a protein molecule, to provide an apparent the contributions resulting from the interaction of three-dimensional molecular model on the face of particular amino acid residues with the water which

99 normally surrounds all protein molecules." (Levin- lows: "A computer at Brown University (Providence, thai, 1966, pp. 315, 317). R.I.) is creating pictures — as simple as a cube and "It is still too early to evaluate the usefulness of as complex as a refinery pipeline — that can be seen the man-computer combination in solving real in three dimensions. The 3D project has been under- problems of molecular biology. It does seem likely, taken by Charles M. Strauss, a graduate student, however, that only with this combination can the and Dr. Andries van Dam, assistant professor of investigator use his 'chemical insight' in an effective appUed mathematics, who is supervising this doc- way. We already know that we can use the computer torate work in the Division of AppHed Mathematics. to build and display models of large molecules and "Mr. Strauss has created a program for the uni- that this procedure can be very useful in helping versity's System/360 Model 50 which enables the us to understand how such molecules function." computer to construct a pair of images, differing (Levinthal, 1966, p. 52). slightly in perspective, side by side on a television- 5.62 "Multipurpose three-dimensional graphs like screen. When viewed through the stereoscope can now be produced from electronic data process- mounted in front of the IBM 2250 display screen, ing equipment with a newly developed spatial a person sees the two images merged into one with data plotter announced by Spatial Data Systems, the added dimension of depth. Inc. ... X- and Y-dimensions on the flat bed of "The images, geometric models stored in the the plotter are represented by the position of a computer's memory, can be manipulated on the series of metal pins thrust upward through the screen — enlarged, reduced, moved up or down board's surface; the Z-dimension is pin height. or rotated — by pressing various keys on the display All pins are step-motor positioned." (Commun. unit. In addition, a person can draw or alter the ACM 10, 253 (1967).) pictures generated by the computer through the 5.63 Some illustrative notes and comments use of a light pen." (Computers and Automation on three-dimensional data input, from an earlier 16, No. 11,52 (Nov. 1967).) report in this series, are summarized as foUows: In addition, we note that "an experimental Stereoscopic initial data recording and three-dimen- pseudo 3^D display at the White Sands Missile sional data depth reconstruction by machine proc- Range was reported to incorporate several 3-D essing and computational techniques are of sig- cues, including perspective, shadow patterns and nificance in such areas of information processing color." (Vlahos, 1965, p. 15). system application as aerial photograph interpre- 5.64 A recent example is summarized as follows: tation, map or contour matching, and analysis of "A perspective transformation is developed whereby bubble-chamber data. For example, a scanner an object in space as viewed from an arbitrary developed by Moore et al. 9" (1964), digitizes aerial point can be projected into a plane and plotted photographs for computer analysis of stereographic . . . An algorithm which eliminates the plotting of differences. hidden portions of the object is discussed." (Kubert interests in R&D the area of three-dimensional et al., 1968, p. 193). For earlier examples and dis- data input and manipulation are exemphfied first cussion see the first report in the present series. by the Sketchpad III system at M.I.T., which is a 5.64a "There are . . . considerable reasons computer program to facilitate entry and processing for developing techniques by which line drawings of three-dimensional drawings (Johnson, 1963). of sohds can be shaded, especially the enhancement Roberts reports that: "After a list of three-dimen- of the sense of solidity and depth . . . Shading sional objects has been obtained in some manner, can specify the tone or color of a surface and the it should possible be to display them from any point amount of light falling upon that surface from one

of view . . . The three-dimensional display program or more light sources. Shadows when sharply will do all this and more." (Roberts, 1965, p. 185). defined tend to suggest another viewpoint and Fano comments further that "special equipment improved surface definition. When controlled, developed by the Electronic Systems Laboratory shading can also emphasize particular parts of at M.I.T. it possible makes to draw with the light- the drawing. If techniques for the automatic deter- pen a three-dimensional object and display its mination of chiaroscuro with good resolution should projection on the face of the cathode-ray tube just prove to be competitive with line drawings, and this as if the object were continuously rotating in three is a possibility, machine generated photographs dimensions under the control of the viewer." (1967, might replace line drawings as the principal mode p. 31). of graphical communication in engineering and An idealized graphic input-output system, to be architecture." (Appel, 1968, p. 37). capable of processing "three-dimensional curve 5.65 "The examples given in the preceding and surface projections", is reported by Parker paragraphs are representative only of a few of the (1965). Developments in new technologies, such as many provocative visual presentations using depth those of holography, offer additional possibilities made possible by the computer technique described for three-dimensional data processing, storage, and in this article. The most obvious use is in the retrieval. presentation of curves and functions of three vari- A more recent example of practical 3— D data ables. When visualized in true depth, many im- processing applications has been reported as fol- portant trends in data become quite evident, as 100 for example, the formant structure of the speech "The scientists are working under a four-year, $5 spectra. . . . Here is a method for presenting million contract with the Advanced Research Proj- stress diagrams, the construction of beams and ects Agency, Department of Defense, Washington, bridges, the structure of molecules, functions of D.C., to increase communications and graphic tech- a complex variable, and much more — all viewed niques between man and computers. from any angle and any distance. It is apparent that "In the halftone research, which is part of the further applications are limited by only the imagina- project, the group is working initially with geo- tion of the prospective user." (Noll, 1966, p. 156). metric shapes, like cubes, pyramids, and spheres. "The present system definitely proves the feasi- These building blocks, once they are programmed bility of the real-time display of two-dimensional into the computer, are 'stretchable' and can be half-tone images. It is felt that the technique may manipulated into many different combinations of be easily extended to stereo representation of half- design forms and shapes. tone images. Furthermore, the algorithm is so con- "One of the practical applications of halftone structed as to allow computations to be executed in research may come in the field of architecture. The parallel (see the dotted section in Figure 11). As new computer graphics should allow an architect many scan lines as hardware permits may be calcu- to design a building on a large cathode-ray tube lated simultaneously. Also, much of the computa- instead of a drawing board. He will need to do so tion may be performed by incremental hardware. only once, and he will be able to do so in perspec- The parallel and incremental characteristics of the tive, right down to the last detail. In the process, algorithm lead us to believe that real-time movement the computer can rotate the design for a view from and display of half-tone images is very near realiza- any angle. The architect can magnify any section tion." (Wylie et al., 1967, p. 58). for close-up examination, and can remove walls to look inside any part of the structure. " (Computers 5.66 "A researcher at the Mitre Corp. . . . has developed a technique to make a three-dimensional and Automation 17, No. 2, 11 (Feb. 1968).) display with mirrors and an off-the-shelf loud- "A computer that makes television pictures out speaker . . . As the flexible mirror vibrates, becom- of numbers has been developed by General Elec- ing alternately concave and convex under the tric's Electronics Laboratory, Syracuse, N.Y. This influence of the speaker's pneumatic drive, the computer, designed for NASA's Manned Spacecraft change in curvature causes a reflected image vibra- Center in Houston, Texas, will be a part of NASA's tion also . . . One of the potential appUcations is in simulation faciHty. air traffic control for displaying air space in three "The purpose of the computer is to generate dimensions, range, azimuth and elevation." (Elec- Hfe-Hke TV pictures of objects such as NASA's tronics 39, No. 3, 41-42 (1966).) Lunar Module, Command & Service Modules and 5.66a "Utah computer researchers create 'Idea the lunar surface. The computer creates pictures Photographs'. Researchers at the University of of these objects from numbers stored in its mem- Utah are using a computer to create three-dimen- ory—numbers that mathematically describe the ional, halftone 'photographs' of a man's ideas. shapes and colors of the objects. There are no TV The research opens the door to a large area of crea- cameras, no films or slides, and no models or draw- tive visual electronics, with remarkable possibilities ings in the system. in nearly every field from entertainment and educa- "The numbers that describe a particular scene tion to architecture and medicine. are on tapes which are read into the computer's "The group of scientists working on the project memory initially and remain there until deliberately has made 'photographs' of objects that exist only erased or until a new tape is read in. Since the in the minds of those at the computer controls. The 'models' are only computer tapes, it is possible to photograph-like reproductions were 'materialized' change from one simulation task to another in a by the computer, which had been previously pro- matter of seconds. grammed in the mathematics appropriate to light "The operator can see the computer's visualiza- and shape. tion from any vantage point. In effect, he enters the "This development opens up many future pos- computer's world by manipulating an aircraft-type sibihties such as: control stick. The computer senses his location 1. Motion pictures in three dimension which give through the movements of the control stick, then the illusion of live action; computes and presents a TV picture of the scene 2. Computerized architecture, which would allow as it would appear for an instant in time. This the designer to spend ninety percent of his process is repeated up to thirty times per second, time on creative work; this would include three- giving the operator the same visual sensations he dimensional building design, ehminating the experiences in a 'real-world' situation. need for blueprints in the design process, and "In addition to space objects and scenes, the aDowing the designer to show viewers (through computer has been programmed to create pictures a special set of 'goggles') how the structure of an aircraft carrier, an aircraft, an airport, and a

will look on the site; and street scene . . . The computer is given a series 3. ReaHstic visual aids which wiU modify teach- of numbers that tell it the shape and colors of the ing methods in education. airport. From the numbers fed to its memory, the

101 computer draws the airport on a TV screen. As the multi-color displays from the user standpoint against viewer manipulates the aircraft-type control stick the penalties that may result in other performance to inform the computer where he would like to be characteristics, in cost and size, and in reliability." in relation to the airport, the computer re-computes (Hobbs, 1966, p. 1877). the exact perspective and size of the scene for the 5.69b "Current [command and control] systems viewer's location." (Computers and Automation call for displays that can handle large volumes of 16, No. 11, 53 (Nov. 1967).) data in real time and that are bright, in full color, 5.67 "Dr. Nathan [Jet Propulsion Lab.] says of high resolution, and highly flexible. There are no that he has been able to use the computer to draw existing displays that can provide this total capa- three-dimensional contour maps of the lunar sur- bility. New techniques must be provided to ensure face. The 7094 accomplishes this by comparing that the display does not become the weakest link light and dark areas in lunar pictures and com- in the data-processing chain." (Baker and Rugari, puting the angle of slopes and relative roughness 1966, p. 37). and elevations between them." (The Computer 5.69c "The technical and economic obstacles Bull. 9,62(1965).) that have prevented the use of color in console 5.67a "This paper describes Design Language displays are rapidly vanishing with the effect that I (DLI) — a prototype computer-aided design system. at least two manufacturers (ITT and Philco-Ford) DLI addresses the creative design of 3-dimensional are currently producing full-color, general purpose, structure, using a display console for communica- console displays. General Electric has also an- tion with the user. . . . nounced a small-screen color display designed for "The prototype DLI system handles design of NASA. 3-dimensional objects consisting of points, straight "The Sylvania tube produces two colors using a lines, and planes. Additional elements are scalars single gun tube and two phosphors, one for each (holders for single numbers), sets (collections of color. The color selection is performed by changing elements), views (2-dimensional displays of ele- the post-deflection accelerating voltage which ments) and modes (holders for definers to be applied changes the beam energy. A high energy beam to each element created when the mode is active)." penetrates the first phosphor layer and excites the (Denil. 1966, p. 527). second, while a low energy beam does not penetrate

"Results from the prototype implementation thereby exciting the first layer. . . . indicate that the language modelling approach "One recent system, developed by General Elec- to the design problem is sound. A designer may tric, produces a full color perspective picture of an start work by sketching a "basic object using the aircraft carrier deck and is used to train naval pilots. APPNT and SKLINE nouns. Once a rough outHne The position of the carrier is under computer control is obtained, the object may be modified into the and changes according to the trainee's actions." desired shape with no difficulties; it is found that (Mahan, 1968, pp. 20-21, 31). the process is very similar to the process one uses 5.69d "Radio Corporation of America (RCA) has in sketching and then refining this sketch into a modified its color scanners so they can produce finished design." (Denil, 1966, p. 535). four-color separations from reflection copy, as well 5.68 "The questions of color, 3-D, and size of as from colored transparencies. RCA has also display are matters of continuing controversy in announced the Graphic 70/8821 Composing Color our service laboratories." (Rosa, 1965, p. 412). Scanner. It has a three-part scanning cylinder. Two 5.69 "A technique that could permit radar or transparencies can be mounted on two of these, sonar screens to display hundreds of different colors with a mask on the third. The scanning head over has been developed by Sylvania Electronic Sys- the mask controls the switching-over from one color ." tems . . (Electronics 37, No. 18, 18 (1964).) transparency to the other to produce any desired 5.69a "It is necessary to make compromises composite picture. in some characteristics in order to accept a display "Crosfield Electronics, Ltd. also has a new scan- technology that meets other essential require- ner, called Diascan 2000. It scans one color at a ments more important to the particular application. time, with a resolution of 500, 1000, or 2000 Hues A decision as to whether to use a multi-color per inch." (Hartsuch, 1968, p. 57). system in a large-screen display is an example of 5.70 "The hardware is based on a 7090 computer, the compromises that must be made. The use of the Stromberg-Carlson 4020 microfilm recorder, several colors in a display offers definite advantages provision on the master film for designating se- in terms of the abiHty to distinguish different quence lengths for individual frames and standard types of items (e.g., in a simple case, friendly and optical printing. hostile forces); hence, from the user standpoint a "The programming involves a 'scanner language' multi-color system is desirable. However, a color which assumes 26 addressable scanners 'looking at' display usually involves a significantly greater specified cells of a 252 x 184 image matrix, where amount of equipment and hardware which implies numeric values 0-7 indicate gray scale." (Knowlton, an increase in space and cost (and probably adverse 1964). effects on reliability and maintainability). Hence, 5.71 "To completely automate the movie proc- the systems designer must balance the need for ess, one experiment used a General Dynamics/

102 Electronics SC-4020 cathode-ray tube as an elec- the problem of photographic inteUigence and the tronics draftsman. This tube, with its associated synthesis of this and other data into indications of

circuitry, was connected directly to the computer. hostile activity . . . The total network requirement

The different drawings were projected electronically [is] for presentation technology-in-combination. . . . on the face of the tube. The time of each projection "And why should we be concerned about mixing was 0.035 second. These projections were then media — text, audio, pictorial — into one single infor- recorded on film." (Gomolak, 1964, p. 70). mation network? For two reasons: one, to accom- 5.72 "Previously, animated movies were made modate the transfer of information to some users for a single frame at a time — a time-consuming and whom one form will not suffice; two, because of the tedious process. Now, however, special photo-optic very nature of information itself, because certain equipment under the control of a high-speed digital types of information are communicated more effec- computer is being used to generate animated movies. tively in visual or audible forms. . . . For example, such computer-generated movies "We must, therefore, consider ways to handle have been used at the Bell Telephone Laboratories digital data, video, analogue and any combinations to demonstrate the effects of different gyroscope of these to best affect the transfer of knowledge constraints on an orbiting satelHte. Also, an in- from originator to user. genious computer-programming language has been "The problem of mixing presentation media is

devised for producing computer movies. . . . "The extraordinarily complex if we consider the various computer method for generating a three-dimensional forms that data can take and the various sub-sets movie is first to calculate, at some particular instant of this data. The primary forms of data, for example, in time, the three-dimensional coordinates of the are original text, or the written word, audio informa- points in a line-drawing representation of the desired tion and pictorial data. If we draw these three forms object. The three-dimensional movie program then of data as three overlapping circles, we can see that calculates the points required for the two perspec- we have seven combinations of data to contend with. tives and also generates the instructions for draw- In medicine the written data would be the texts of ing the perspectives with a plotter. The three- authors contained in books, journals, and miscel- dimensional coordinates of the object at a small laneous monographs. Audio data would be heart time increment later are then computed, and the murmurs. Pictorial data would be surgical tech- three-dimensional program generates the instruc- niques. Audio data could be combined with visual tions for another frame of the movie. This procedure demonstrations of palpations. Written data could is repeated on a frame-by-frame basis until the be accompanied by diagrams or illustrations. desired movie sequence is completed. The movie "Thus, we have seven combinations of data pres- is then recorded on magnetic tape which is used entation to consider. Let us now look at the sub-sets as input to the plotter. The plotter decodes the of each of these. Full text can appear as abstracts instructions on the tape and repeatedl)^ advances the of the text, as shorter indexes, as titles and pagina- film and deflects the electron beam until the entire tions, or as evaluative data about the text such as movie is completed." (Noll, 1966, pp. 143-145). reviews or comments on the content of the text. 5.72a "This paper presents a generalized tech- For the text set alone we, therefore, have four sub- nique for generation of off-line CRT motion-picture sets of information, all or none of which may be displays of objects in three-dimensional space. available. All of these should be considered for the Initial implementation the called of system, VISTA* total information network." (Mayeda, 1967, pp. 4-5). (Visual Information for Satellite Telemetry Analy- 5.74 "It seems to us that there is some merit in sis), uses the Univac 1108 computer to generate trying to develop such large-scale, high-resolution time-sequence displays for output on the Strom- dynamic displays for group-computer interaction, berg-Carlson 4020 microfilm plotter. The system though at the same time we appreciate the difficulty accepts satellite orbit and attitude data and proof the technical problems involved." (Licklider, duces motion pictures illustrating the three dimen- 1965, p. 102). sional position and orientation of the spacecraft 5.74a "With the ability to scan preliminary or in orbit, relative to one or more celestial bodies." approximate results while his problem is still on (Chapman and Quann, 1967, p. 59). the computer, and to immediately make on-line 5.73 "Until now the designer has been con- modifications in method or detail of calculations, cerned with the many, detailed problems of design- the user need not plan his calculations so carefully ing and producing large, single-path information in advance, nor does he need to request fine detail systems. Mixing presentation media into one system over all ranges of a function simply because he can- has been limited to the efforts of the military, notably not tell in advance where the interesting regions intelligence, groups who were forced to cope with are." (Ward, 1967, p. 48). •VISTA was developed by Computer Sciences Corporation for NASA-GSFC. See also note 5.40, pp. 91-92.

6. System Output to Microform and Hard Copy 6.1 "First, the well-established tradeoffs improved in terms of capability and lower cost, between digital and microform technologies have These improvements have made equipment more changed httle; both technologies have steadily accessible to the 'average' user or installation.

103 will com- The two technologies increasingly be systems offer higher density than any magnetic i bined; for instance, with computers decoding tape device can provide. Image systems can also i complicated retrieval patterns into locations in be easily read by their human managers, whereas bulk microform memory that store difficult-to- 'reading' of mag tape data requires relatively ; digitize graphical information." (Van Dam and complex equipment. For these reasons, says the

Michener, 1967, p. 215). firm, manufacturers of EDP gear are gradually 5 "For a number of reasons, computer-based joining EDP and information retrieval systems, j hbraries that service a wide spectrum of users, while makers of other-than-digital storage equip- such as is found in a university, will be faced with ment are responding to the challenge with super operating on two basically different types of data — micro-image systems." (Kornblum, 1965, p. 35). !j that which is digitally stored and that which is 6.4 "The Stromberg-Carlson SC-402() micro- : photographically stored in some microfilm form. film plotter consists primarily of a cathode ray tube t latter will images of the original full text The be and a 35-mm camera for taking pictures of the ' of the documents contained in the library whereas information displayed on the face of the tube. In- : the digital data will constitute the augmented structions for the SC-4020 are written on magnetic ^ catalog of the library from which the library user tape; the tape is then decoded by the SC-4020 I gleans information about the library data and and used to generate commands for opening and ( documents by conducting on-line computer closing the shutter of the camera, for advancing the ! searches." (Haring, 1968, p. 35). film, and for deflecting the beam of the cathode ( 6.2 "Recent developments have been in the area ray tube. Development of the film produces a 35-mm i of microfilm readout devices which transfer data microfilm transparency which consists of lines con- I ." from magnetic tape to microfilm . . (Com- necting points, drawn, in effect, directly under the mun. ACM 10, 387 (1967).) control of a digital computer. In this manner, the "Right now the microfilm recorders — which perspective points computed by an IBM 7094 digital appear to be the most promising— show signs of computer are used as the input to an off-line SC- being successful in three major applications. 4020 microfilm plotter through an intermediate ji 1. Direct computer output for scientific com- magnetic-tape storage. After photographic devel- puter work where the plotting capability is opment, the microfilm can then be viewed directly required. in a stereoscope, and the final result is an illusion 2. Computer applications which are in the area of depth created by a completed computerized tech- considered primarily data retrieval, where nique." (Noll, 1966, p. 152-153). direct transfer to microfilm for efficient storage "This system is presently being developed and is is a primary consideration, and hardcopy scheduled for first dehvery early next year. It secondary. represents a combination of the knowledge gained 3. When the computer is used to prepare data in over 10 years of CRT display-microfilm recording for graphic arts reproduction. Here is a vast equipment experience. The S-C 4060 is a stored j

new field only beginning to be opened up. program, data recording system which translates i; (Webster, 1966, p. 33). digital data into alphanumerics and graphics and e For example, "we (General Electric) are moving records on microfilm and/or paper. The system can l it can i towards using computer output almost completely operate from the output of a computer or operate independently since it is equipped with its i in microfilm form." (F. B. Campbell in "System Magazine's Third Annual Microfilm Seminar", own computer. The S-C 4060 provides great flexi- j of options. To a great degree bility through a variety [ 1965, p. 36). its 'building block' modular design permits the s 6.3 "Bell & Howell's Micro-Data Div. underlines user to match his requirements as they change." f its 'yes' evaluation, that progress is being made, by (Peterson, 1966, p. 133). pointing out that modern recording devices are j "The significant trend in output technology that t placing computer information onto microfilm at began to develop in 1966 is the use of microform for lines At speeds of up to 30,000 per minute. the computer output to achieve substantial reduction same point in hardware technology, modern film in cost, in printing time of master and dissemination i readers are furnishing copies of film-held informa- copies, and in physical volume. Stromberg-Carlson's } tion to the ultimate user in from four to ten seconds." highly successful SC-4020 microfilm printer/plotter » (Kornblum, 1965, 36). p. is to be updated by an entire 'micromation' Hne 't "National Cash Register Co. puts forth some of equipment of varying capability and cost. The meaty and positive thoughts relative to the survey's microfilm recorders in this series — the SC-4060, > first query. NCR states that an increasing number the SC-4360, the SC-4440, and the SC-4460-all ' and variety of EDP systems will allow for computer convert digital codes from a computer to their alpha- I

output to micro-image forms. Some should even numeric and graphical forms on a CRT, where a ! permit micro-images to be used as input, with auto- camera records them. The SC-4060 printer/plotter * matic conversion to digital data. This trend, con- (the replacement for the SC-4020) is used for scien- ' tinues NCR, could bring micro-image systems into tific and engineering graphics and for frames for !

EDP communications, remembering that the image animated movies; it can plot curves, diagrams, j

104 maps, etc. This system leases for about $8000 per movies were being generated at a rate of about month. The SC-4440 converts alphanumeric com- 1,000,000 frames per year." (Christenson and puter output into microfilm at the rate of 9000 pages Pinson, 1967, p. 697). per hour. Hardcopy equipment is available with 6.6 Some further details are as follows: "After these recorders. Many other manufacturers, such an image is processed by the computer program and as California Computer Products and Benson- modified and verified by the operator at the graphic

Lehner. have similar equipment." (Van Dam and console, it is frequently necessary to produce a IVIichener, 1967, pp. 191-192). permanent output document of the processed image. 6.4a "The introduction by Stromberg-Carlson of Film is used for the output image media, as for the Micromation — or microfilm directly produced from input media, because of its compatibility with ." a computer, bypassing the paper stage . . computer speeds, flexibility of use, and high (Kalagher. 1968, 37). p. image quality. . . . "SC 4460 recorder converts magnetic-tape mate- 6.6a Thus it is noted that "seldom is there rial to roll film or microfiche at speeds of up to 90,000 control of original subject matter. It varies from cps. The characters are proportionally spaced and perfect, fresh, raised printing or litho offset to computer forms can be overlaid on the microfilm crumbling pages from archives; from clear black- to eliminate expense of forms printing." (Berul, and-white Hne copy to smudged spirit purple on 1968, p. 30). yellow sulfite paper; from sharp, dense characters facsimile tech- 6.4b "The system combines to burned out, illegible script, from thin, fine fines niques with digital processing to provide a scanning, to heavy soUd illustrations or even fine half-tones digitizing, and playback system. Particular emphasis with four and five color separations." (Dunning, has been placed on the requirements of high resolu- 1965, p. 14). tion and the ability to handle large image formats. "CRT recording on silver film permits images to The resulting device is capable of digitizing pictorial be generated by computer control at computer-data- information in 16 discrete gray levels, at a data channel speeds. The CRT used for output image density of 1000 elements per inch in both x and y recording is similar to the type used for input directions. Formats up to 9 x 18 inches can be image scanning. As in the scanner, a CRT image handled. is formed by vectors drawn on the CRT screen. 'Tmage data on tape, generated hy computer The vector trace is visible when the beam is un- processing, can recorded on film the be by system blanked, in order that the beam trace can be at the same resolution and gray-shade range as are recorded on the film. When the beam is blanked, used in scanning." (Gilman, 1966, 29). p. the trace is not visible and the beam can be posi- 6.5 "Minnesota Mining and Manufacturing's tioned over the image without exposing the film. "dry silver' Electron Beam Recorder (EBR) An image — for example, a drawing— is divided into provides strong competition to the more traditional straight line segments. Core storage contains the microfilm is written directly CRT recorders. Output X and Y coordinates of the end point of each vector. on the microfilm electron at a rate of with an beam Vectors may vary in length from the very short 30,000 characters per second (bypassing the traces required to form a smooth arc to longer traditional a display), filming of CRT thus yielding traces which are used for straight line segments of a sharper picture. Film is processed by a dry hot- the drawing." (Hargreaves et al., 1964, p. 367). air developing method (rather than the wet-stage 6.6b For example. Nelson found in a university development used with conventional microfilm), library installation that the focusing lever "had providing high-speed, real-time output. The EBR been broken off, and the person who was running unit is expected to rent for $3000 per month, with the equipment didn't know that it was necessary the additional cost per microfilm page image ex- to focus it. By this time, it was apparent that the pected to be less than two-tenths of a cent." (Van material was perfectly good . . . Here they were, Dam and Michener, 1967, p. 192.) ready to send $5,000 worth of stuff back because "3M has a new electron beam recorder (Series F) they weren't maintaining their own equipment that converts computer generated data into human properly." (C. E. Nelson in "System Magazine's readable language at speeds of 60K lines per minute Third Annual Microfilm Seminar", 1965, p. 39.) on dry silver microfilm. Film is developed by heat ". and can be viewed instantly." (Data Proc. Mag. 9, Similarly, Holmes reported: . . Our head

. . . layers of dirt and a cracked No. 6, 3 (1967).) technician found 6.5a "Graphical output has been an important lens. After these conditions were cleared up, course, the films were no longer faded." (D. C. way of presenting results to Bell Laboratories' of Holmes in "System Magazine's Third Annual computer users since the fall of 1961, when a Stromberg-Carlson 4020 microfilm printer was Microfilm Seminar", 1965, p. 39.) installed at Murray Hill. It has allowed users to 6.7 "Lockheed's experience has affirmed user replace pages of tabular data with pictures that acceptance to be broadly based and has brought are far easier to understand. Microfilm picture requests to TIC management to install a micro- generation was running recently at a rate of about film catalog in R & D oriented buildings." (Kozum- 500,000 frames per year. In addition, computer plik and Lange, 1967, p. 67).

105 "Studies at the Battelle Memorial Institute, simultaneously." (Christenson and Pinson, 1967, 1961-62, identified important factors affecting pp. 699-700). satisfaction in the reading of microtext, including 6.7b "Another scanning converter built by the especial importance of high resolution, and United Aircraft Corporation for the USAF feeds demonstrating that under proper conditions computer output directly to a local facsimile re- microreproductions can be even more readable corder or out over a network without any inter- than the originals." (Council on Library Resources, mediate hard copy." (Alden, 1964, p. 13). 10th Annual Report, 1966, p. 22). 6.7c "Remote copying and the computer have "From experience with the Library of Congress, been successfully merged at Bell Telephone Lab- the research worker prefers to work with microfilm oratories. The hnk-up . . . joins computer-gen- rather than original material." (D. C. Holmes, in erated data display with Xerox's Graphic Terminal

"System Magazine's Third Annual Microfilm Hardcopy Printer System . . . The graphic ter- Seminar", 1965, pp. 35-36.) minal has its own core memory consisting of 125,000 6.7a "Computer-stored graphic data such as bits and a drum memory of 11 milhon bits. The charts, graphs and grids can now be reconverted, latter can store 8 individual images. By using LDX [Long Distance Xerography] scanned and memory units for temporary storage of the signals transmitted to a remote printer by a device Xerox as a buffer, the terminal is able to regulate the flow calls its LDX/Computer Adapter." (Commun. of data in scan line form to the printers at a speed ACM 9, 889 (1966).) of one page every seven seconds." ("Just Looking forward to the not too distant future. .", Merged . . 1967, pp. 50-51). Dunning (1965) reports: "Long range concepts of 6.8 "General Precision, Inc., Pleasantville, hard copy include LDX (Longdistance Xerography), N.Y., has been working for a number of years on where multiple output stations will be at variable equipment designed to transmit microfilm images distances from the input. Soft copy display on by closed circuit television . . . The latest version tubes is already here with closed-circuit TV sys- shown at the 1965 Western Electronic Show in- tems. Opportunity for hard copy print-out by selec- cluded the GPL TV Printer which allows the user tive control at location is push button any out-put to take a hard copy of whatever he is viewing on the in the making". Present difficulties, of course, screen. Prints 8.5" x 11" are delivered in about include those of high costs. half a minute by rapid processing of conventional a Sched- "Facsimile equipment normally requires silver halide paper." (Veaner, 1966, p. 208). ule 2 telephotograph channel. For a small additional 6.9 "On October 2, 1957, a proposal was sub- terminal charge, the same line can be used alter- mitted to the Council on Library Resources by a natively for voice messages." (Alden, 1964, p. 11). manufacturer of closed-circuit television equip- ". . . Xerox Corporation's work in facsimile . . . ment for design and construction of a prototype, LDX, which stands for 'long distance xerogra- remotely controlled, catalog card viewing system. is phy' . . . The equipment, at the output end, The proposed 'Telereference' system would permit much hke our Xerox 914 or 813 copier. The input, a researcher to view catalog cards in a central however, may be some distance away, up to thou- catalog from a remote location by the use of closed- sands of miles away, if necessary . . . We pro- circuit television and a remotely controlled card quality, that scan pose, in order to produce high we manipulator." (Bacon et al., 1958, p. 1). the original document at approximately 200 Unes 6.10 "All the libraries in the Lake County per inch. . . . Library system share in a growing book collec- "The LDX system is presently designed for fuU- tion ... A union catalog in book form, in single size originals at the transmitting end, but it is a or multiple copies as required, is furnished to each straightforward design job to provide means to participating library; the closed circuit teletype accept the input in the form of microfilm." (Mayo, communication provides the necessary means 1964, pp. 77-78). of sharing the common collection as reflected in "The STARE hard-copy facility accommodates the union catalog." (Burns, 1964, p. 14). up to eight rapid hard-copy output stations. The 6.11 "One laboratory will have a closed-circuit system consists of multiple remote Xerox LDX TV link with the computer so that a graphical (Long Distance Xerography) Printers interfaced to presentation (plotted curves as well as alpha and a central buffer memory and control unit that is numeric information) of computer output will be located near the GE-645. The LDX printers form available at the remote terminal." (Neilsen, 1965, an image by a raster scan technique, so the data p. 635). from the central buffer unit to the printers are 6.11a "The idea behind this [video-storage dis- essentially a video signal for controlling the intensity play] device is to take advantage of the very low along each scan line. These data to the LDX Printer cost of commercial TV for the display and to are transmitted from the central buffer at about utilize reasonably inexpensive magnetic drum or a 250,000-baud date over coaxial cable links. There disk storage to record the video data for the display are eight separate buffer areas on the magnetic memory. Rapid strides in high-density magnetic drum which function independently. Therefore, recording have made this console a reality. Brook- transmission to up to eight printers can take place haven Laboratory has an operational 10 station

106 video-station of their own design [Brad]." (Stotz, which 'employs the upper- and lower-case print 1968, p. 14). chain, and ... a new technique of printing 6.11b "A new records retrieval system called boldface by overstriking the headword in each Remstar is now available from Remington Office citation'." (Stevens and Little, 1967, pp. 73, 75). Systems. With Remstar, closed circuit television 6.16 "The Potter Chain Printer, Model PS-6020, is combined with automated records retrieval units operates at up to 600 lines-per-minute and features to not only locate records, but to transmit them as interchangeable type fonts and up to 192 different weU." (Systems 7,6(1966).) characters." (Commun. ACM 9, 651 (1966).) 6.12 "AH told, twenty or more non-impact "The Model PS-6020 Off-line Printer utiHzes the devices are being, or have been, marketed. They Potter Model SC-1060 tap unit, and the Model vary widely in their capability and cost. Those PS-6030 utihzes the Model SC-1080 tape unit. which are extremely fast tend to have poor print Both models use Potter's HSP-3502 Chain Printer. quahty. Those which off^er sufficient quaUty are The chain printing mechanism permits precise Umited to one-ply, or may require additional proc- vertical registration, which is difficult to achieve essing to obtain even the single copv." (Webster, with drum printers . . . 1965. p. 43). "The HSP-3502 Chain Printer prints one to six 6.13 "The typical computer-printer or tabu- copies on forms 2.5 to 18.5 inches wide. Standard lating machine produces monospaced upper case fan-fold or roU forms, heat transfer or pressure- characters, that is, evenly spaced capital letters. sensitive papers, card stock, and multiHth masters The result is a monotonous looking product with can be printed in lengths of up to 44 inches per no typographic variation between captions and form . . . body. There is Httle reader appeal and the eye "Ten numerals, 26 upper-case and 26 lower-case must work overtime to find what it wants from the letters, and 30 symbols are available for inclusion printed page." (Lannon, 1967, p. 81). in the printing character set. Any special character "As long as Mr. Watson would only give us a or symbol can be provided optionally. Character character font in capital letters there could not be sets of any size up to 192 characters, made up of any great fervor to load [machines with biblio- any combination of numerals, letters, and special graphic information records]. If I may say so, the symbols, can be provided." (Melick, 1967, p. 46). whole picture of automation in libraries from the 6.17 "Mechanical output printers have increased thirties right down to the present date has been in speed to the point where they are now commonly controlled by that uppercase limitation." (Clapp, producing 600 to 1,000 lines per minute (2,000 cps) 1964. pp. 54^55). depending upon many factors, including the size of

6.14 "[The upper-case limitation] reduces the repertoire or character set. . . . readability so much that the page layout must be "A one-thousand-line per minute printer with 120 extremely open if the material is to be usable at print positions can create text at the rate of 2,000 all." (Kraft, 1963, p. 275). characters per second. The character set used in "A major disadvantage of computer print-out this extremely rapid process is, however, very small is its poor legibility and its low information density." and it is manipulated without any of the control (Corrado, 1965, p. 271). devices (justification, leading, etc.) used by type- "It has been reported that 20 percent of the total setting machines." (Sparks et al., 1964, p. 177). volume printed by the Government Office is data 6.17 "To achieve Thermal Printing, two novel assembled by computer. In general, such material components are utilized. The first, a print head, is has been either printed out then keyboarded again an array of resistors selectively heated electrically for typesetting or the printout has been simply to generate a thermal image. The second is a thermal photographed and reproduced. In the latter case, sensitive paper which is held in contact with the the printed material is less legible than if it were print head to provide a printed record of the thermal typeset, and 40 percent more bulky." (Webster, image. . . . 1966, p. 33). "Five wafers are mounted together to form a print 6.15 "The IBM 1403 chain printer involves use head (Figure Ic). Each print head contains ten 5x5 of a continuously moving chain consisting of 240 matrices; each matrix is capable of printing one character positions usually divided into 5 sets of character. Eight print heads are mounted side by 48 characters each, so as to minimize access time side in the printer to achieve the 80 character to the appropriate characters to be selected. How- hne. . . . ever, by redividing, a character set of 120 can be "The Thermal Printer can print data received at accommodated, although at significant loss (up to any mixed rate of input — including input from a 50 percent-60 percent) of printer speed. This tech- keyboard. The rate of input has been tried and nique is used, for example to provide upper and tested at rates as high as 240 characters/sec which lower case for the KWIC index. Chemical Biological is certainly not the upper Umit. Since we also get Activities ... In addition, the cover of the Pro- an electronic, nonimpact printer which appears to ceedings of the Literary Data Processing Con- be very reUable and can produce multiple copies on ference [Bessinger et al., 1964] illustrates an plain paper, thermal printing may have potential experimental concordance of the Gothic Bible application in many systems — both communications

107 systems and automatic data processing systems." 6.20 "However, since all photocomposers exist- (Joyce and Homa, 1967, pp. 261-262, 267). ing or under development appear to use photo- 6.18 Further, the system modifications will pro- graphic recording, we can start at the light-sensitive vide for multiple use of a battery of printers. "In a emulsion, and explore the various methods used to system of 32 printer terminals a translator clock form and position the optical images. High-speed derives a sequence control which is connected to a photocomposers generally faU into six principal 5-bit printer scanning buss. Each printer is set to categories: recognize its own scan number and will respond to a. Those which have a stationary master glass the column print pulses on the print buss only if it grid from which the symbols are selected, has been addressed and only when it is scanned." sized and positioned onto the photographic (De Paris, 1965, p. 49). emulsion by purely optical-mechanical means. 6.18a "USAECOM has also sponsored a develop- An example of this is the Mergenthaler Lino- ment of Miniaturized Techniques for Printing film System. (MINIPRINT) with NCR. This is an exploratory b. Those which have the symbols on a spinning development printer which prints at 15 characters/ master glass grid and a single stroboscopic second, can be operated from a keyboard, and was flash tube that can be flashed at the proper developed for potential use as a self-contained, instant to select anv desired character, with battery operated, tactical teleprinter. The package size control, imaging and positioning by size of the present model (cigar-box size) may also optical-mechanical means. Examples of this contain all the required electronics. The printing is are the Photon 200, 500, and 700 series immediately visible on new thermal sensitive master machines and the Harris-Intertype spinning paper. The master and first copy are made simul- disc photocomposer. taneously in this printer and at least ten legible c. which have the symbols on a stationary copies can be obtained using the master copy and Those master glass grid, with a separate flash tube plain paper in an off-line pressure device." (Joyce for each symbol, a reciprocating lens which and Homa, 1967, p. 267). aU symbols across a complete line 6.18b "New teleprinter devices have recently positions tunnel, with the various come on the market. Included are a high speed through an optical flash tubes being fired as each symbol is electrostatic printer and an ink jet, which squirts the charged ink particles onto the page, deflecting the imaged in each of its desired positions in a complete line to be composed ink jet with electrostatic plates. The added printing line, permitting machine of speed makes these devices attractive as printers; in one pass of the lens. The only is 900 (Zip or GRACE) but both units shield the printed line from the view this type the Photon the first of which was dehvered to the of the user until the line is complete which hmits machine, in 1964. their utility in an on-line terminal. Furthermore, National Library of Medicine their complexity means a higher price tag although d. Those which have the symbols on a stationary they remain alphanumeric devices." (Stotz, 1968, master glass grid, steadily illuminated and a light-sensitive p. 13). optically projected onto 6.18c Webster also reports that "the Radiation cathode of an image-dissecting tube, which 'Super-Speed' printer has a single line of styU ex- simultaneously produces scanned video tending across the continuously moving paper. These electron beams from all symbols, one of which also are pulsed several times to form characters is allowed to pass through an electrostatic from top to bottom, and by this means a speed of wire grid selection matrix to produce a video over 31,000 lines per minute is achieved." (Webster, brightening signal for a high-resolution cathode- 1966, p. 32). ray display tube, with positioning and size on 6.18d "Electronic methods for changing the display tube being controlled electronicaUy character size may be employed, for example, by The only such machine of this type is the zooming. The images produced on the CRT face Mergenthaler-CBS Linotron system being de- are focused onto a microfilm strip which then, veloped for' delivery to the Government after development and printing, becomes the hard Printing Office and the Air Force in 1967, copy output record." (Potter and Axelrod, 1968, employing the CBS "Reconotron" image p. 2). dissector-scanner, and the CBS "Microspot" 6.19 For example, as described by Barnett, high-resolution cathode-ray display tube.

Kelly, and Bailey: "In photocomposition . . . suc- e. A special cathode-ray tube employing a beam- cessive characters along a line are displayed with shaping metal plate perforated with the symbol coUinear base lines, regardless of their magnifica- shapes, with means to cause the electron beam tion (i.e.) point size. . . . The same character is to impinge on any symbol area, emerging as displayed in different point sizes, in photocomposi- a shaped electron beam, which can be deflected tion, by the use of different lenses. . . . Special to display the symbol at any position on the characters can appear anywhere in a text, and can face of the cathode-ray tube. An example is be magnified and spaced in the same ways as the Charactron tube, employed in photo- ordinary alphabetic characters." (Barnett et al., composers built by the Stromberg Carlson 1962, pp. 60-63). Division of the General Dynamics Corp., 108 such as the S-C 4020. The repertoire of line of information onto the face of a cathode-ray symbols is usually severely limited to less than tube. It does this by plotting on the tube the coor- 100. dinate points which make up the shape of an in- f. Methods employing a conventional or high- dividual letter. The result is optically projected, a resolution cathode-ray tube to display the line at a time, to a spool of moving, unexposed film." symbols as they are generated by some means (Becker, 1967, pp. 3-4). ". which will produce suitable beam deflections . . [The SC—4020] composes characters and and video brightening to form the symbols." symbols at a rate of 17,400 per second. The char- (Stevens and Little, 1967, p. 22). acters are formed by directing a beam at the desired 6.20a "The year has seen remarkable progress character; then horizontal and vertical deflection in the field of phototypesetting. Photon, Inc. added circuits deflect the selected character to the ap- to its 713 line. The model 713—5 operates at 30 propriate spot on the CRT display. This technique newspaper lines per minute. Fairchild Graphic has also been successfully combined with a dry Equipment entered the field with two machines, the process Xerographic printer which has achieved a Photo Text Setter 2000 and the Photo Text Setter composing rate of 1 million characters per minute." 8000. American Type Founders also announced an (Sparks etal., 1964, p. 180). addition to its line of phototypesetters. to be "Our investigations at RSIC [Redstone Scientific

known as the ATF Photocomp 20 . . . Information Center] have indicated that today the "Alphanumeric, Inc., has a new very high- CHARACTRON tube can give the best quahty speed phototypesetting machine that is being used within the state-of-the-art." (Burger, 1964, p. 12). in its service center at Lake Success, Long Island. 6.23 "The Linotron consists of three main units: And H arris -Intertype is producing a CRT type of Control, Character Generation, and Display. Mag- fast phototypesetting machine, the first of which netic tape is the input for this machine. This tape was scheduled to be installed at Baird-Ward Co., contains aU data characters and control codes re- Nashville. Tenn., near the end of 1967. quired to operate the photocomposer. Four grids, "K. S. Paul & Associates, London, England, also each containing 255 characters, can be loaded into has a CRT type of phototypesetting machine. It the Linotron. The character generator can position is not quite as fast in its operation as the ones the required grid in one-half second and select any mentioned above, but is less expensive. Mergen- character on the grid at electronic speeds. (There thaler Linotype Company has announced that it are no shutters or other mechanical devices involved

'j will market this machine in the United States under in character generation.) The output of the char- the of Linotron 505." (Hartsuch, 61). acter is an electric signal which as j name 1968, p. generator serves 6.21 "The A. B. Dick Videograph tube transfers an input to the display section. The beam of the j

! characters by means of a set of fine wires embedded display tube is positioned to the required X- and

" in its face." (Webster, 1965, p. 43). Y-coordinates; then the signal from the character "A. B. Dick Co.'s Model 904 Videograph Page generator causes the character to be scanned out j ' Printer translates a digital coded input into a printed on the surface of the display tube. This character is output at rate 130-character lines per projected through a lens onto the film. The beam is 'l the of 7200, minute." (Data Proc. Mag. 7, No. 2, 51 (1965).) then positioned for the next character and this proc- I is until is is 1 "The Videograph electrostatic printing process ess repeated the page complete. There

1^ uses a special cathode ray tube for creating alpha- no mechanical motion of any kind while the page is numeric characters on dielectricaUy coated paper being composed. The film is advanced only after the j at rates in the order of 10,000 characters per sec- page has been completed." (Rollert, 1968, p. 7). I ond. In electronic data processing systems, the 6.24 "RCA recently unveiled a new computer- ' source data are usually provided from computer- controlled photocomposing system called VIDEO* prepared magnetic file tapes. The binary-coded COMP. With this system, it is possible to store signals from the tape are decoded in the Videograph the analog equivalent of several type fonts and system by means of an electronic character gener- styles. This enables the user, in the final printing , ator into the video signals necessary to modulate the stage, to select a variety of intermixed type fonts I printing tube and create the character shapes. The and styles with virtually no restrictions. It results principal application of Videograph printers has in page copy which resembles letter press quality been to produce address labels of the type used on in terms of page makeup precision and aesthetic

! magazines, newspapers, and direct mail." (Gerlach, appearance." (Becker, 1967, p. 3). is the generic which i 1968, p. 17). "Videocomp name RCA 6.22 "Perhaps the best quality is produced by has given to electronic composition systems and the Stromberg-Carlson 'Charactron' shaped-beam equipments which employ high-speed cathode ray tube which uses a tiny 64-character mask within the tube (or equivalent) photocomposing techniques. i- tube itself. The images displayed on the tube are These systems and equipments are capable of captured by microfilm in the S-C 4020, or printed greater speed and flexibility than first generation electrostatically in the low-cost S-C 3070." (Webster, optical-mechanical photocomposers. The 70/822 1965, p. 43). Videocomp is the output system which performs the "The SC 4020, on-Hne to the computer, uses a actual photocomposition in response to magnetic charactron tube which can produce the letters of a tape data and control input. The 70/9300 Video- 109 a

comp Electronic Composition System is an inte- A character set may be replaced with an alternate grated system which includes the 70/822 Video- set of 180 characters by transfer from auxiliary comp and executes the complete composition store in approximately one-half second." (Stevens process from manuscript to final photocomposed and Little, 1967, pp. 33-34). page." (Coleman et al., 1968, p. 30). 6.25a "Scores of printing jobs can be typeset up "A new all-electronic typesetter that can set the to six times faster than before with a computer- complete text for a magazine page in less than four controlled photocomposition unit announced by seconds or write information for microfilm storage IBM Corporation, White Plains, N.Y. The new at computer speeds has been announced by RCA, device called the IBM 2680 CRT printer, is the New York, N.Y. The Videocomp 70/830 generates first of its kind to tap the full power of a computer. characters at a rate of up to 6,000 per second— It can be used for a variety of typesetting jobs, thousand times faster than manual typesetting including books, newspapers, catalogs and direc- machines. Additionally, the Videocomp provides tories. Linked directly to an IBM System/360, full page composition through its 70 pica (12 inch) it can set graphic-quaMty type at speeds ranging line length and its ability to write in sizes from as high as 6,000 characters a second. 4 to 96 point type (approximately H inches high) "The 2680 CRT printer is operated completely with proportional reduction for 35mm microimage under the control of System/360. Draft copy is output. usually fed into the computer on paper tape or

"Videocomp can be used to set type in almost magnetic tape, but it can be entered through any every format conceivable — from a full page tabloid System/360 input device. A user can choose from newspaper composition to directories or parts a large variety of type styles and sizes — all stored manuals. Equally important, the new equipment magnetically on the disc files seen in the foreground makes possible quality typeset printout of computer of the picture. information at speeds well in excess of the 1,100 "The computer formats and justifies the copy, lines per minute typical of high-speed printers automatically hyphenating words that break now used for computer readout. This could result in between lines. It then commands the 2680 to write easier handling and r-eading of the business and the finished text electronically on a television-like scientific data now being processed by computers. cathode-ray tube (CRT)." (Computers and Auto- "In operation, original copy and composition mation 17, 58 (Jan. 1968).) instructions are fed to a computer which produces 6.25b "Consideration is given to the possibiUty an output magnetic tape or sends signals directly of providing a computer and a cathode ray printer to the 70/830. In response to taped or direct signals, with an unlimited repertory of characters. Digi- the machine generates characters on a cathode ray talizations are presented for mathematic, carto- tube. As the characters appear on the tube they graphic, and calligraphic characters. The repertory expose standard or 35 microimage film, stabiliza- is available to any computer through FORTRAN IV tion paper for quick proofing, or short run offset programming. The latest cathode ray printers plates. are almost adequate for the preparation of mathe- "Videocomp stores typefonts in electronic matical reports." (Hershey, 1967, ii, abstract). memory where they can be used instantly. This Some of the problems of character and symbol internal storage reduces load on computer memory generation are discussed in another report in this and permits use of the Videocomp off-line with series involving overall system design requirements, any computer. especially, character set considerations. "Because the entire operation is electronic, 6.26 Automatic hyphenation methods required every letter can be altered electronically to form for line justification variously involve the use of roman, oblique, extended, condensed and superior dictionary or table lookups, suffix and prefix sphtting, and inferior versions of the basic face. Every face, logic rules, and combinations of these techniques. in every variation, can be mixed at will — on the Automatic hyphenation techniques for languages same fine, in the same word — with every other face. other than Enghsh include those for Dutch, French, In addition, Videocomp can store and write any German, and Itafian as reported in the "Computer special form including scientific, mathematical and Typesetting Conference Proceedings," The Institute engineering symbols, foreign alphabets, ideographs of Printing (1965). or trademarks — literally, almost any visual image." 6.26a "The dot or raster mode of the system is (Computers and Automation 17, No. 4, 63 (Apr. used to generate half-tone pictures. The machine 1968).) can generate any desired dot density and practi- 6.25 "The Digiset developed by Dr. Rudolf cally any resolution up to a maximum of 660 dots per Hell of Kiel utilizes an 8-bit character selection inch in either field direction." (Potter and Axelrod,

code and it sets justified and hyphenated lines of 1968, p. 5). 11, 18, or 28 picas at rates of up to 400 characters 6.26b "An added feature common to the modern per second. ... In a telephone directory applica- CRT is the ported window at the rear of the tube. tion, the character set consisted of 180 alphanumeric The port is designed so that static overlays such as and other symbols. Character patterns are stored maps can be projected from a sHde or film clip onto in a core memory of 180,000-190,000 bit capacity. the face of the CRT. The overlay relieves the display

110 of the need to continuously regenerate static data." "For many situations the operator may desire (Mahan. 1968, p. 19). hard-copy text of the image on the display. Two "Static information can be superimposed on the means are currently employed. A Polaroid photo- beam written data by projecting pictures (slides) graph provides a quick copy adequate for many through an optical port in the CRT. Typical systems purposes. For complex drawings where more resolu- can select from among 25-150 slides." (Machover, tion is required, or where it is desirable to be able 1967. 157). p. to write on the copy, the . . . MODEL T [a] "So that an optically projected image can be com- device can provide signals for either a digital or bined with the electronically generated image, analog mechanical plotter." (Miller and Wine, 1968, CRT's in terminals supplied by Bunker-Ramo and p. 475). Stromberg-Carlson have a window through which "The hard copy should be of high enough quality a picture can be projected onto the screen." to reproduce legibly the information on the display, (Machover. 1967, p. 150). (although it probably need not be of record-keeping 6.26c "Techniques with the greatest potential quality). Another important consideration is that are all based on photocomposition. Photon was the the copying process should not materially delay the first to introduce a spinning-disk type of photo- user from going on with his work while printing composer. The Photon 900 or 'Zip' is the machine takes place." (Stotz, 1968, p. 16). employed in the GRACE system at the National "Hard-copy outputs can be provided, where Library of Medicine to produce Index Medicus, needed, by a variety of devices, ranging in speed described by Austin. It is a text-only machine cap- from typewriter rates of 10-15 cps up to 60,000 cps. able of operating at speeds on the order of 250 char- The high-speed end of the range tends to be im- acters per second from magnetic tape input. The practical because broad-band communication lines improved Photon 901 overcomes the problems of are needed and they are very expensive." (Berul, variable character density and variable base lineup 1968, p. 29). of characters, and also has an attachment for in- 6.29 For example, "The information system's serting graphics from projected shdes." (Amer. designer is faced with a dilemma. If he provides for unhmited hard service, a portion Institute of Physics Staff, 1967, pp. 355-356). copy major of his budget is absorbed in fulfilling hard copy requests. 6.27 Relatively early examples are discussed in If he insists that obtaining hard copy is the user's the 1961 report of Buck et al. Then we qote: (1) "The responsibility, he and the user may unwittingly tonal scale of the illustration is simulated in the be with-holding information which may be extremely printing process by making the dot pattern structure valuable. The 'missed' item in instances on the printing press plate a function of the tonal some could exceed in value the total cost of operating gradations of the original illustration. Thus the dark the information system. areas of the original illustration are represented "A third aspect of the hard copy dilemma is by a dot pattern in which the dots are relatively attempting to satisfy a need which may not be large and with little white space between the in- related to either use or value. This need will be dividual large dot areas . . . called the 'Possession is Security Syndrome' or (2) "The output graphic must be a dot pattern frequently referred to as the 'pack rat instinct'. structure in which the distribution of the dots over It is a very important aspect often encountered in the image area represents a faithful reproduction of investigations of users' needs and satisfactions — yet the continuous tone input image. It should be noted it is little understood in the documentation area." that since the dot pattern structure does not appear (Resnick, 1964, p. 317). in the input graphic, it must be created within the "The real answer to the hard-copy problem LGCP system." (Mergenthaler, 1963, 13). p. requires convincing users that they will never 6.28 [In a Lexical-Graphical Composer-Printer lose easy access to information they have once (LGCP) system at Wright-Patterson AFB], "The uncovered. This is far from a trivial problem in

I Video Graphics Copier converts illustrations to documentation and we seem no further along video tape. The Format Processor accepts full-text j towards solving it than we were a decade ago." coded mag tape, effects composition, and produces I (Ohlman, 1963, p. 194). an I instruction tape. The photocomposer accepts 6.30 "Developments include electrostatic the instruction and video tapes as input, is driven I printing, magnetic printing, smoke printing, ' by the instruction tape to select and position data thermal recording, thermoplastic recording, and

I characters onto the output and to insert the graphics photographic recording. Many of these techniques from the video tape when necessary, and creates can be applied to the making of replica copies of fully-composed pages in the form of film positives." existing documents as well as to generating the (Kehler, 1968, p. 119). initial copy of a document from an electronic or 6.28a "The need for hard copy which the user digital store." (Sparks et al., 1964, p. 181). can take to his office for contemplation still exists, 6.31 "Photographic film of the text will then be however. We currently use Polaroid photographs, provided in a format suitable for an automatic

but the need for a more convenient and larger character reader . . .

copy . . . was often expressed." (Evans and "The planned separation method is semi- Katzenelson, 1967, p. 1143). automatic in that decisions to separate text from

111 graphics are to be made by a human operator." 6.33 An example of current research in this area (Bus. Automation 12, No. 8, 66, Aug. 1965). is provided in Quarterly Progress Report No. 80, 6.31a "The area of microfiim-to-hard-copy The Research Laboratory for Electronics, M.I.T. conversion is not yet very advanced. There are a More specifically: "By a judicious choice of ele- number of reader-printers, such as the Itek 1824 ments for the lexicon and the use of a set of very or the 3M Filmac 200, which under operator simple algorithms it is possible to decompose a control can make copies from any type of microfilm. word in an orderly fashion into the stored elements The problem, however, is that the process is too from which the phonemic translation can be found. slow and the cost of the copies too high for pro- A lexicon containing approximately 32,000 elements is sufficient for duction volumes." (Berul, 1968, p. 30). the decomposition of all entries in 6.31b "The emphasis was on making it easy to the Seventh Edition of Webster's 'New Collegiate use the system to obtain information. This included Dictionary'. Additional markings' in the lexicon such things as producing printed output which provide the basis for the resolution of syntactical would not look like an accounting ledger, but would ambiguities which cannot be handled by the phonic be similar to the majority of printed documents approach." (Quarterly Progress Report No. 80, people are used to reading and which would provide Research Laboratory for Electronics, MIT, 225 visual cues for ease of scanning and interpretation." (1966).) (Ebersole, 1965, p. 172). "An efficient scheme has been developed for the automatic translation of English text letters 6.31c Further,". . . IDEA provides for displays from of specified variables. At present these include to phonemes. This process of translation is an frequency distributions and contingency tables. indispensable part of a reading machine that pro- We also intend to provide other plots, scatter dia- duces synthetic-speech output." (Quarterly Progress grams, etc., with such aids as discriminant func- Report No. 80, Research Laboratory for Electronics, tions, regression analysis, iso-probability, and iso- MIT, 225 (1966).) distant curves derived under various assumptions." "The problem of automatic translation from letter (Press and Rogers, 1967, p. 39). spelling to phonetic spelling in English text has been 6.32 "The properties and techniques by which studied. A procedure has been developed, based a high-speed printer can be made to emboss regular upon a 'dictionary' of approximately 30,000 root printing paper with braille symbols is described in words and a set of combination and generation laws, ." such that the correct pronunciation (phonetic Sterling et al. . . [Sterling, T. D., M. Lichstein, F. Scarpino, D. Stuebing and W. Stuebing, "Pro- spelling) of 250,000 English words can be auto- fessional Computer Work for the Blind", Commun. mated." (Quarterly Progress Report Research Laboratory for Electronics, No. 80, MIT, 218 (1966) ACM 7, 228 (1964)]. (Landwehr et al., 1965, p. 302). Then, "at MIT one aspect of work to aid the blind further ref. to F. Lee Ph.D. Thesis, EE. Oct. 1965). includes coupUng of a high-speed electric Braille A project at the Institut fiir Phonetik und Kom- typewriter and a small computer to achieve an munikationsfordung. University of Bonn, has both inexpensive conversion of machinable documents FAP and COMIT language programs written to into Braille code." (E. Glaser, report of panel dis- explore this possibility. cussion on Computers and Aid to the Handicapped, 6.34 Project of A. Risberg, Royal Institute of Commun. ACM 8, 638 (1965).) Kehl (1965) also Technology, Stockholm, Sweden. (Stevens, 1968, mentions the M.I.T. high-speed Braille printer. pp. 21-22). It is of particular interest to note in this connec- "A conference on compressed speech, sponsored tion that "initial experience with several trainees by the Division for the Blind . . . The method indicated that the blind programmer equipped with may be used by students as a review technique and routines which print input, output and/or memory by blind persons to increase their reading rate (L Inf. contents in braille is no more dependent on external appreciably when using recorded books." C Bull. 674 help than is his sighted colleague." (Landwehr et al., 24, (1965).) See also note 5.60, p. 98. 1965, p. 300). Nelson (1965) has also explored some of the pos- sibiUties for on-line BraiUe display.

7. System Use and Evaluation

7.1 "The [test or sample] problems were in- "The quantitative data, the analyses, and the variably simple because generally any typical conclusions about man-computer systems that one problem would require the amassing of considerable seeks in a controlled experiment will emerge only data and its manual manipulation to determine if the data-reduction effort is painstakingly planned check-points for system rating. The expense and executed." (Morey and Yntema, 1965, p. 356). involved was considerable and generally was not 7.2 "One of the most difficult problems in evalu- approved by the pertinent management groups." ating information retrieval systems is to identify (Davis, 1965, p. 82). meaningful evaluation criteria. Most of the im-

112 portant criteria can be directly related to cost, therefore, connected with user performance and and time."" (King and Terragno. 1964. capabihty which may only be assessed in a qualita- reliability' p. 393). tive manner." (Blunt. 1965, p. 9). ". . . No quick and simple evaluation procedure "The setting up of criteria of evaluation, on the which is both conceptually coherent and empirically other hand [in contrast to measures of effective- convincing is possible. There is not. in fact, wide- ness] demands user participation and provides an spread agreement or conviction as to what would indication of whether the user understands the qualify as a criterion for evaluating an information reason for the system, the role of the system and system." (Ossorio. 1965. p. 52). his responsibilities as a prospective system user.' 7.3 "The value of an information system is. (Davis, 1965, p. 82).

113 Appendix B. Bibliography

Abramowich. J., Storage Allocation in a Certain Iterative 157, 172 p. (U.S. Govt. Printing Office, Washington, D.C.. Process, Commun. ACM 10, No. 368-370 (June 1967). Dec. 31, 1961). Adams, C. W., Responsive Time-Shared Computing in Busi- Baker, C. E. and A. D. Rugari, A Large-Screen Real-Time Display ness—Its Significance and Implications, AFIPS Proc. Fall Technique, Inf. Display 3, 37-46 (Mar./ Apr. 1966). Joint Computer Conf.. Vol. 27. Pt. 1, Las Vegas, Nev., Nov. Baker, F. T. and W. E. Triest, Advanced Computer Organization, 30-Dec. 1, 1965, pp. 483-488 (Spartan Books, Washington, Rept. No. RADAC-TR-66-148, 1 v. (Rome Air Development D.C.. 1965). Center, Griffiss Air Force Base, N.Y., May 1966). Adams, S., Information — A National Resource, Am. Doc. 7, Barnett. M. P., K. L. Kelley and M. J. Bailey, Computer Genera- No. 2, 71-75 (April 1956). tion of Photocomposing Control Tapes, Part 1, Preparation of Alden, W. L.. Cutting Communication Costs with Facsimile, Flexowriter Source Material, Am. Doc. 13, No. 1, 58-65 Data Proc. Mag. 6, 11-14 (Sept. 1964). (Jan. 1962). Allen, J., Machine-to-Man Communication by Speech Part III: Barnett, M. P.. D. J. Moss, D. A. Luce and K. L. Kelley, Computer Synthesis of Prosodic Features of Speech by Rule, AFIPS Controlled Printing, AFIPS Proc. Spring Joint Computer Conf., Proc. Spring Joint Computer Conf., Vo.l 32, Atlantic City, Vol 23, Detroit, Mich., May 1963, pp. 263-288 (Spartan Books, N.J., Apr. 30-May 2. 1968, pp. 339-344 (Thompson Book Co.. Bahimore, Md.. 1963). Washington, D.C.. 1968). Barnett, M. T., Computer Typesetting: Experiments and Pros- Amdahl, G. M., Multi-Computers Applied to On-Line Systems, pects, 245 p. (M.I.T. Press, Cambridge, Mass., 1966). in On-Line Computing Systems, Proc. Symp. sponsored by Barton, R. S., A Critical Review of the State of the Programming the Univ. of California, Los Angeles, and Informatics, Inc., Art, AFIPS Proc. Spring Joint Computer Conf., Vol. 23, Los Angeles, Calif., Feb. 2-4, 1965, Ed. E. Burgess, pp. 38-42 Detroit, Mich., May 1963, pp. 169-177 (Spartan Books, Bahi- (American Data Processing, Inc., Detroit, Mich., 1965). more, Md., 1963). American Institute of Physics Staff, Techniques for Publication Baruch, J. J., A Medical Information System: Some General and Distribution of Information, in Annual Review of Informa- Observations, in Information System Science and Technology, tion Science and Technology, Vol. 2, Ed. C. A. Cuadra, pp. papers prepared for the Third Cong., scheduled for Nov. 21—22, 339-384 (John Wiley & Sons, Inc., New York, 1967). 1966, Ed. D. E. Walker, pp. 145-150 (Thompson Book Co., Anderson, H. E., Automated Plotting of Flow-Charts on a Small Washington. D.C., 1967). Computer, Commun. ACM 8, 38-39 (Jan. 1965). Bauer, W. F., On-Line Systems — Their Characteristics and Moti- Aoki, M., G. Estrin and R. Mendell, A Probabilistic Analysis of vations, in On-Line Computing Systems, Proc. Symp. spon- Computing Load Assignment in a Multiprocessor Computer sored by the Univ. of California. Los Angeles, and Informatics, System, AFIPS Proc. Fall Joint Computer Conf., Vol. 24, Inc., Los Angeles, Calif., Feb. 2-4, 1965, Ed. E. Burgess, Las Vegas, Nev., Nov. 1963. pp. 147-160 (Spartan Books, pp. 14-24 (American Data Processing, Inc., Detroit, Mich., Bahimore, Md., 1963). 1965). Appel, A., Some Techniques for Shading Machine Renderings of Becker, C. H., UNICON Computer Mass Memory System, AFIPS Solids, AFIPS Proc. Spring Joint Computer Conf., Vol. 32, Proc. Fall Joint Computer Conf., Vol. 29, San Francisco, Cahf., Atlantic City, N.J.. Apr. 30-May 2, 1968, pp. 37-45 (Thompson Nov. 7-10, 1966, pp. 711-716 (Spartan Books, Washington, Book Co., Washington, D.C.. 1968). D.C., 1966). Appel, A., The Notion of Quantitative Invisibility and the Becker, J., Technology in Information Presentation, transcript Machine Rendering of Solids, AFIPS Proc. Spring Joint Com- of lecture delivered to the Institute of Management Science, puter Conf., Vol. 32, Atlantic City, N.J., Apr. 30-May 2. 1968, Washington, D.C., Chapter, Nov. 1, 1967, 12 p. pp. 387-393 (Thompson Book Co., Washington, D.C., 1968). Becker. R. and F. Poza, Natural Speech from a Computer, Applebaum, E. L., Imphcations of ihe National Register of Micro- Proc. 23rd National Conf., ACM, Las Vegas, Nev., Aug. 27-29, film Masters, as Part of a National Preservation Program, 1968, pp. 795-800 (Brandon/Systems Press. Inc., Princeton, Lib. Res. & Tech. Serv. 9, 489-494 (1965). N.J.. 1968). Arnovick, G. N., Computer-Processed Information-Recording A BeU, G. and M. W. Pirtle, Time-Sharing Bibliography, Proc. and Association System, in Statistical Association Methods For IEEE 54, 1764-1765 (Dec. 1966). Mechanized Documentation, Svmp. Proc, Washington. D.C., Benner, F. H.. On Designing Generalized File Records for Mar. 17-19. 1964, NBS Misc. Pub. 269. Ed. M. E. Stevens Management Information Systems, AFIPS Proc. Fall Joint et al., pp. 181-184 (U.S. Govt. Printing Office, Washington, Computer Conf., Vol. 31, Anaheim, Calif., Nov. 14-16, 1967, D.C., Dec. 15, 1965). pp. 291-303 (Thompson Books, Washington, D.C., 1967). Aron, J. D., Real-Time Systems in Perspective, IBM Sys. J. Bennett. E., J. Degan and J. Spiegel, Eds., Mihtary Information 6, No. 1, 49-67 (1967). Systems — The Design of Computer Aided Systems for Com- Austin, C. J., Dissemination of Information and Problems of mand, 180 p. (Frederick A. Praeger, Pub., New York, 1964). Graphic Presentation, Proc. 1965 Congress F.I.D. 31st Meeting Bennett, E., E. C. Haines and J. K. Summers, AESOP: A Proto- Congress, Vol. II, Washington, D.C., Oct. 7-16, 1965, pp. 241- type for On-Line User Control of Organizational Data: Storage, 245 (Spartan Books, Washington, D.C., 1966). Retrieval and Processing, AFIPS Proc. Fall Joint Computer Automatic Typesetting, First International Technical Information Conf.. Vol. 27, Pt. 1, Las Vegas, Nev., Nov. 30-Dec. 1. 1965, Congress, La Compaigne Francaise d" Editions. Paris, 1965, pp. 435-455 (Spartan Books, Washington, D.C., 1965). 374 p. Berul, L., Information Storage and Retrieval: A State-of-the-Art Report, Rept. No. PR-7500-145, 1 v. (Auerbach Corp., Phila- Bacon, F. R., N. C. Churchill, C. J. Lucas, D. K. Maxfield and delphia, Pa., Sept. 14.1964). C. J. Orwant, Application of a Telereference System to Divi- sional Library Card Catalogs: A Feasibility Analysis, Final Berul, L. H., Survey of IS & R Equipment, Datamation 14, No. 3, Rept., 91 p. (Engineering Research Inst., Michigan Univ., 27-32 (Mar. 1968). Ann Arbor, May 1958). Bessinger, J. B., Jr., S. M. Parrish and H. F. Arader, Eds., Literary Bagg, T. C. and M. E. Stevens, Information Systems Retrieving Data Processing Conference Proceedings, Sept. 9-11, 1964, Replica Copies: A State-of-the-Art Report, NBS Tech. Note 329 p. (IBM Corp., White Plains, N.Y., 1964).

*Additional references on page 124.

114 Bhimani. B. V.. R. D. Merrill. R. P. Mitchell and M. R. Stark, Examining Patent Offices, ICIREPAT, Washington, D.C., An Approach to Speech Synthesis and Recognition on a Digital Oct. 7-16, 1964, Ed. H. Pfeffer, pp. 362-377 (Spartan Books, Computer. Proc. 21st National Conf.. ACM, Los Angeles, Washington, D.C., (1966). Calif.. Aug. 30-Sept. 1. 1966. pp. 275-296 (Thompson Book Buck, C. P., R. F. Pray III and G. W. Walsh, Investigation and Co.. '^S'ashington. D.C., 1966). Study of Graphic-Semantic Composing Techniques, Rept. Blunt, C. R., An Information Retrieval System Model, Rept. No. No. RADC-TR-61-58. Final Rept. Contract AF 30(602)2091, 352, 14-R-l. 150 p. (HRB-Singer, Inc., State College, Pa., 1 V. (Syracuse Univ., Research Inst.. June 1961). Oct. 1965). Burchinal, L. G., The Expanding Role of Microforms in Educa- and D. L. Structure Bobrow. D. G. Murphy. of a LIST System tion, NMA J. 1, 2-6 (Fall 1967). Two-Level Storage, No. 155-159 Using Commun. ACM 10, 3, Burger, J. B.. High Speed Display of Chemical Nomenclature, (March 1967). Molecular Formula and Structural Diagram, Rept. No. C105^ D. G.. R. Y. Kain, B. Raphael and C. R. Licklider. Bobrow. J. R-4, 19 p. (General Electric Co., Huntsville, Ala., Dec. 31, A Coniputer-Program System to Facihtate the Study of Tech- 1964). nical Documents. Am. Doc. 17, 186-187 (Oct. 1966). Burgess, E.. Ed., On-Line Computing Systems, Proc. Symp. Bohnert. H. G., English-Like Systems of Mathematical Logic sponsored by the Univ. of California, Los Angeles, and Infor- for Content Retrieval, in Automation and Scientific Communi- matics, Inc., Los Angeles, Calif., Feb. 2-4, 1965, 152 p. (Amer- cation. Short Papers. Pt. 2. papers contributed to the Theme ican Data Processing, Inc., Detroit, Mich., 1965). Sessions of the 26th Annual Meeting, Am. Doc. Inst., Chicago, Burkhardt. W. H., Universal Programming Languages and 111., Oct. 6-11. 1963, Ed. H. P. Luhn. pp. 155-156 (Am. Doc. Processors: A Brief Survey and New Concepts, AFIPS Proc. Inst., Washington. D.C., 1963). Fall Joint Computer Conf., Vol. 27, Pt. 1, Las Vegas, Nev., Bohnert, H. and M. Kochen, The Automated Multilevel Ency- Nov. 30-Dec. 1, 1965, pp. 1-21 (Spartan Books, Washington. clopedia as a New Mode of Scientific Communication, in Some D.C.. 1965). Problems in Information Science, Ed. M. Kochen, pp. 156-160 Burns, L. R., Automation in the Public Libraries of Lake County, (The Scarecrow Press, Inc., New York, 1965). Indiana, in Proc. 1963 Clinic on Library Applications of Data Bones. W. L. and F. A. Kros, Nationwide Real-Time Warranty Processing, Univ. of Illinois, Apr. 28-May 1, 1963, Ed. H. System, in Data Processing, Vol. X, Proc. 1966 International Goldhor, pp. 9—17 (Graduate School of Library Science, Data Processing Conf.. Chicago. 111.. June 21-24, 1966, pp. IHinois Univ., Urbana, 1964). 56-64 (Data Processing Management Assoc.. 1966). Bushnell, D. D., Computer-Mediated Instruction — A Survey of Bonn, T. H.. Mass Storage; A Broad Review, Proc. IEEE 54, New Developments, Computers & Automation 4, 18-21 1861-1870 (Dec. 1966). (Mar. 1965). Borko. H., The Construction of an Empirically Based Mathemat- ically Derived Classification System, Rept. No. SP-585, 23 p. Cahngaert, P., System Performance Evaluation: Survey and (System Development Corp., Santa Monica, Calif., Oct. 26, Appraisal, Commun. ACM 10, 12-18 (Jan. 1967). 1961). Also in AFIPS Proc. Spring Joint Computer Conf., Carr, J. W., Ill and N. S. Prywes, Satellite Computers as Inter- Vol. 21. San Francisco, Calif., May 13, 1962. pp. 279-289 preters, Electronics 38, No. 24, 87-89 (1965). (National Press, Palo Alto. Calif., 1962). Carter, L. F., F. N. Marzocco, D. L. Drukey and C. Baum, Re- Borko. H.. The Conceptual Foundations of Information Systems, search and Technology Division Report for 1964, Rept. No. Rept. No. SP-2057, 37 Development TM-530/008/00, p. (Svstem' Corp., Santa 155 p. (System Development Corp., Santa Monica. Calif.. May 6. 1965). Monica, Calif., Jan. 1965). Borko. H.. Integrating Computers into Behavioral Science Re- Chang. W., A Queuing Model for a Simple Case of Time Sharing, search, .VFIPS Proc. Fall Joint Computer Conf., Vol. 27. Pt. IBM Sys. J. 5, No. 2, 115-125 (1966). 1. Las Vegas, Nev., Nov. 30-Dec. 1. 1965, pp. 527-.532 (Spartan Chapman, G. A. and J. J. Quann. VISTA-Computed Motion Books, Washington, D.C., 1965). Pictures for Space Research, AFIPS Proc. Fall Joint Computer Borko, H. and H. P. Burnaugh, Interactive Displays for Docu- Conf., Vol. 31, Anaheim, Calif., Nov. 14-16, 1967, pp. 59-63 ment Retrieval. Rept. No. SP-2557, 25 p. (System Development (Thompson Books, Washington, D.C.. 1967). Corp.. Santa Monica, Calif., Aug. 4. 1966). Chasen, S. H. and R. N. Seitz, On-Line Systems and Man- Bowers. D. M., Organization of Rotating Bulk Memories for Computer Graphics, Computers & Automation 16, 22-28 Minimum Random Access Time, Computer Design 5, No. 2, (Nov. 1967). 40-44 (Feb. 1966). Cheydleur, B. F., SHIEF: A Realizable Form of Associative Breuer, M. A., General Survey of Design Automation of Digital Memory, Am. Doc. 14, No. 1, 56-67 (Jan. 1963). Computers. Proc. 1708-1721 (Dec. IEEE 54, 1966). Cheydleur, B. F., Ed., Colloquium on Technical Preconditions Breuning, S. M. and T. N. Harvey, Implications of Integrated for Retrieval Center Operations, Proc. National Colloquium Transportation Engineering Design Systems for Computer on Information Retrieval, Philadelphia, Pa., April 24-25, 1964,

Software and Hardware Development, Proc. 22nd National 156 p. (Spartan Books, Washington, D.C.. 1965). Conf.. ACM, Washington, D.C., Aug. 29-31, 1967, pp. 255-262 Cheydleur, B. F., Information Science and Liberal Education, (Thompson Book Co., Washington. D.C., 1967). Am. Doc. 16, No. 3, 171-177 (July 1965). Brown, G. W., G. Miller A. J. and T. Keenan, Eds., EDUNET- Christensen, C. and E. N. Pinson, Multi-Function Graphics for a Report of the Summer Study on Information Networks Con- Large Computer System, AFIPS Proc. Fall Joint Computer ducted by the Interuniversity Communications Council Conf., Vol. 31, Anaheim, Calif., Nov. 14-16, 1967, pp. 697-711 (EDUCOM), 440 p. (Wiley. New York, 1967). (Thompson Books, Washington, D.C., 1967). Brown. R. M., An Experimental Study of an On-Line Man- Christian, W. C, VSMF Stars in Microfilm Publishing, Systems Computer System. IEEE Trans. Electron. Computers EC-14, 7, No. 10, pp. 12-15, 42, 46 (Oct. 1966). 82-85 (1965). Clapp, V. W., Libraries and the 'Uppercase Limitation', in Brown, S. C, A Bibliographic Search by Computer. Physics Libraries and Automation, Proc. Conf. held at Airlie Founda- Today 19, No. 5, .59-61, 63-64 (May 1966). tion, Warrenton, Va., May 26-30, 1963, under sponsorship of Brunelle, L. A., The Impact of Magazine Film Use in the Micro- the Library of Congress, The National Science Foundation, film Industry. Eleventh Annual Meeting and Convention, and The Council on Library Resources, Ed. B. E. Markuson, Vol. XI, 1962, Ed. V. D. Tate, pp. 8.5-92 (The National Micro- pp. 54-55 (Library of Congress, Washington, D.C., 1964). film Assoc., Annapohs, Md., 1962). Clem, P. L., Jr., AMTRAN — A Conversational-Mode Computer Bryan, G. E., JOSS: 20,000 Hours at a Console — A Statistical System for Scientists and Engineers, in Proc. IBM Scientific Summary. AFIPS Proc. Fall Joint Computer Conf., Vol. 31, Computing Symp. on Computer-Aided Experimentation, Anaheim. Calif., Nov. 14-16, 1967. pp. 769-777 (Thompson Yorktown Heights, N.Y., Oct. 11-13, 1965, pp. 115-150 Books. Washington. D.C.. 1967). (IBM Corp.. White Plains, N.Y., 1966). Bryant. E. C, Progress toward Evaluation of Information Re- Climenson, W. D., File Organization and Search Techniques, trieval Systems, in Information Retrieval Among Examining in Annual Review of Information Science and Technology, Patent Offices, Proc. Fourth Annual Meeting of the Committee Vol. 1, Ed. C. A. Cuadra, pp. 107-135 (Interscience Pub., for International Cooperation in Information Retrieval Among New York, 1966).

115 366-108 O - 70 - 8 Clippinger, R. F., Programming Implications of Hardware Corrado, V. M., The Integration of Computers and High Speed Trends, IFIP Congress 65, Vol. 1, New York, N.Y., May 24-29, Typesetting Into the Composing Room, in Automatic Type- 1965. Ed. W. A. Kalenich, pp. 207-212 (Spartan Books, Wash- setting, First International Technical Information Congress, ington. D.C., 1965). La Compagine Francaise d" Editions, Paris, 1965, pp. 271-306. Coffman. E., Stochastic Models of Multiple and Time-Shared Council on Library Resources, Inc., Tenth Annual Report for Computer Operations, Ph. D. Dissertation, Dept. of Engineer- the period ending June 30, 1966, Washington, D.C., 1966, 128 p. ing, Univ. of California, Los Angeles, 1966. Critchlow, A. J., Generalized Multiprocessing and Multiprogram- Coffman, E. G. and B. Krishnamoorthi, Preliminary Analyses of ming, Proc. Fall Joint Computer Conf., Vol. 24, Las Vegas, Time-Shared Computer Operation, Rept. No. SP-1719 Nev., Nov. 1963, pp. 107-126 (Spartan Books, Baltimore, Md., (System Development Corp., Santa Monica, Cahf., Aug. 1964). 1963). Coffman. E. G., Jr. R. C. and Wood, International Statistics for Crowell, F. A. and S. C. Traegde, The Role of Computers in Time Sharing Systems, Commun. ACM 9, No. 7, 500-503 Instructional Systems: Past and Future, Proc. 22nd Nation2J (July 1966). Conf., ACM, Washington, D.C., Aug. 29-31, 1967, pp. 417-425 Coffman. E. G., L. Jr. and Kleinrock, Computer ScheduUng (Thompson Book Co., Washington, D.C., 1967). Methods and Their Countermeasures, AFIPS Proc. Spring Cuadra, C. A., A Feasibihty Study for Automated Fingerprint Joint Computer Conf., Vol. 32, Atlantic City, N.J., Apr. 30- Identification, Tech. Memo, TN-3007, 64 p. (System Develop- 11-21 May 2, 1968, pp. (Thompson Book Co., Washington, ment Corp., Santa Monica, Calif., 1966). D.C.,1968). Cuadra, C. A., Ed., Annual Review of Information Science and Coffman, E. G. and L. C. Varian, Further Experimental Data Technology, Vol. 1, 389 p. (Interscience Pub., New York, 1966). on the Behavior of Programs in a Paging Environment, Cuadra, C. A., Ed., Annual Review of Information Science and Commun. ACM 1 1, 471-474 (July 1968). Technology, Vol. 2, 484 p. (Interscience Pub., New York, 1967). Coggan, B. B., The Design of a Graphic Display System, Rept. No. 67-36, 185 (Dept. of Engineering, Univ. of California, p. David, E. E., Jr., Sharing a Computer, Int. Sci. & Tech. 54, Los Angeles. Aug. 1967). 38-47(1966). Cohen, J., A Use of Fast and Slow Memories in List-Processing Davis, R. M., MiUtary Information Systems Design Techniques, Languages. Commun. ACM No. 2, 82-86 (Feb. 1967). 10, in Military Information Systems — The Design of Computer- Cohen, J. B., G. O. Gardner and B. W. Romberg. Design Automa- Aided Systems for Command, Ed. E. Bennett et al., pp. 19-28 tion in Ship Detaihng, Proc. 22nd National Conf.. ACM, Wash- (Frederick A. Praeger, Pub., New York, 1964). ington, D.C., Aug. 29-31, 1967, pp. 341-353 (Thompson Book Davis, R. M., Information Control in Command-Control Systems, Co., Washington, D.C., 1967). in New Perspectives in Organization Research, Ed. W. W. Coleman, A. H., R. F. Day and D. G. Gerlich, A Videocomp Sys- Cooper et al., pp. 464-478 (Wiley, New York, 1964). tems Approach, in Electronic Composition in Printing, Proc. Davis, R. M., Classification and Evaluation of Information System Symp. Gaithersburg, Md.. June 15-16, 1967. NBS Special Design Techniques, in Second Cong, on the Information Pub. 295, Ed. R. W. Lee and R. W. Worral, pp. 29-36 (U.S. System Sciences, held at The Homestead, Hot Springs, Va., Govt. Printing Office, Washington, D.C., Feb. 1968). Nov. 1964, Ed. J. Spiegel and D. E. Walker, pp. 77-83 (Spartan Collila, R. A.. Time-Sharing and Multiprocessing Terminology, Books, Washington, D.C., 1965). Datamation 12, No. 3, 49-51 (March 1966). Davis, R. M., Man-Machine Communication, in Annual Review Comfort, W. T., A Computing System Design for User Service, of Information Science and Technology, Vol. 1, Ed. C. A. AFIPS Proc. Fall Joint Computer Conf., Vol. 27, Pt. 1, Las Cuadra, pp. 221-254 (Interscience Pub., New York, 1966). Vegas, Nev., Nov. 30-Dec. 1, 1965, pp. 619-626 (Spartan Day, M. S., The Scientific and Technical Information Program Books, Washington, D.C., 1965). of the National Aeronautics and Space Administration, in Computer Development (SEAC and DYSEAC) at the National U.S. Atomic Energy Commission, The Literature of Nuclear Bureau of Standards, NBS Circular 551, 146 p. (Washington, Science: Its Management and Use, Proc. Conf. held at Division D.C., Jan. 1955). of Technical Information Extension, Oak Ridge, Tenn., Sept. Condon, R. A., The Design of an Information Storage and Re- 11-13, 1962, pp. 361-377 (TID-7647, Oak Ridge, Tenn., 1962). trieval System : FMA's File Search, in Automation and Scientific Denil, N. J., A Language and Model for Computer-Aided Design, Communication, Short Papers, Pt. 2, papers contributed to Proc. 21st National Conf., ACM, Los Angeles, Calif., Aug. 30- the Theme Sessions of the 26th Annual Meeting, Am. Doc. Sept. 1, 1966, pp. 527-535 (Thompson Book Co., Washington, Inst., Chicago, lU., Oct. 6-11, 1963, Ed. H. P. Luhn, pp. D.C., 1966). 137-138 (Am. Doc. Inst., Washington, D.C., 1963). Dennis, J. B., Segmentation and the Design of Multiprogrammed Connors, T. L., ADAM — A Generalized Data Management Computer Systems, J. ACM 12,589-602 (Oct. 1965). System, AFIPS Proc. Spring Joint Computer Conf., Vol. 28, L. Glaser, Infor- Boston, Mass., April 1966, pp. 193-203 (Spartan Books, Dennis, J. B. and E. The Structure of On-Line Washington, D.C., 1966). mation Processing Systems, in Second Cong, on the Informa- Coons, S. A., An Outline of the Requirements for a Computer- tion System Sciences, held at The Homestead, Hot Springs, 5-14 Aided Design System, AFIPS Proc. Spring Joint Computer Va., Nov. 1964, Ed. J. Spiegel and D. E. Walker, pp. Conf., Vol. 23, Detroit, Mich., May 1963, pp. 299-304 (Spartan (Spartan Books, Washington, D.C., 1965). Books, Baltimore, Md., 1963). Dennis, J. B. and E. C. VanHorn, Programmed Semantics for Cooper, F. S., Haskins Laboratories: Research on Human Multiprogrammed Computations, Rept. No. MAC-TR-23, Communication, Marine Ecologv, and the Biochemistry of 46 p. (M.I.T., Cambridge, Mass., Dec. 1965).

Protozoa, Science 158, 1213-1215 (Dec. 1, 1967). DeParis, J. R., Random Access, Data Proc. Mag. 7, 30-31 Cooper, W. W., H. J. Leavitt and M. W. Shelly II, Eds., New (Feb. 1965). Perspectives in Organization Research (Wiley, New York, DeParis, J. R., Desk Top Teleprinter, Data Proc. Mag. 7, No. 10, 1964). 48-49 (Oct. 1965). Coordinated Science Laboratory, LIniv. of Illinois, Progress Dertouzos. M. L., CIRCAL: On-Line Circuit Design, Proc. Report for Dec. 1964, Jan. & Feb. 1965, Urbana, 111., Apr. 7, IEEE 55, 637-654 (1967). 1965,77 p. Diaz, A. J., Microreproduction Information Sources, Lib. Res. Corbato, F. J., System Requirements for Multiple Access, and Tech. Serv. 11, No. 2, 210-214 (Spring 1967). Time-Shared Computers, Rept. No. MAC-TR-3, 14 p. (M.I.T., Cambridge, Mass., n.d.). Diaz, A. J., On-Demand Publishing — The Clearinghouse Concept, Corbato. F. J., The Transition from CTSS to MULTICS. Project in Levels of Interaction Between Man and Information, Proc. MAC Progress Report III, pp. 95-101, 1967. Am. Doc. Inst. Annual Meeting, Vol 4, New York, N.Y., Corbin, H. S. and W. L. Frank, Display Oriented Computer Oct. 22-27, 1967, pp. 238-241 (Thompson Book Co., Washing-

Usage System, Proc. 21st National Conf., ACM, Los Angeles, ton, D.C., 1967). f

Calif., Aug. 30-Sept. 1, 1966, pp. 515-526 (Thompson Book Dolby, J. L., L. L. Earl and H. L. Resnikoff, The Application of Co., Washington, D.C., 1966). English- Word Morphology to Automatic Indexing and Extract- ' Cornew, R. W., A Statistical Method of Spelling Correction, ing, Rept. No. M-21-65-1, 1 v. (Lockheed Missiles and Space Inf. & Control 12, 79-93 (Feb. 1968). Co., Palo Alto, Calif., April 1965).

116 Drew. D. L.. R. K. Summit, R. I. Tanaka and R. B Whitely, An Farrell, J., TEXTIR: A Natural Language Information Retrieval On-Line Technical Library Reference Retrieval System, in System, Tech. Memo. TM-2392, 37 p. (System Development Processing Proc. Information 1965, IFIP Congress 65, Vol. 2, Copr., Santa Monica, Calif., May 5, 1965). York, N.Y., May 24-29, 1965, New Ed. W. A. Kalenich, pp. Feingold, S. L., PLANIT-A Flexible Language Designed for 341-342 (Spartan Books, Washington, D.C., 1966). Also in Computer-Human Interaction, AFIPS Proc. Fall Joint Com- Am. Doc. 17, No. 1,3-7 (Jan. 1966). puter Conf., Vol. 31, Anaheim, Calif., Nov. 14-16, 1967, pp. Bell Dudley, H. W., The Vocoder, Lab. Rec. 18, 122-126 (1936). 545-552 (Thompson Books, Washington, D.C., 1967). Dugan. R. S. Green, and E. J. J. Minker W. Shindle, A Study Fiala, F. T., Time-Sharing Operations and Management, in Data Utility of the of Associative Memory Processors, Proc. 21st Processing, Vol. X, Proc. 1966 Int. Data Processing Conf., National Conf., Los Angeles, Calif., ACM. Aug. 30-Sept. 1, Chicago, 111., June 21-24, 1966, pp. 157-166 (Data Processing 1966, pp. 347-360 (Thompson Book Co., Washington, D.C., Management Assoc., 1966). 1966). Fine, G. H., C. W. Jackson and P. V. Mclsaac, Dynamic Program A. Dumey, L, Considerations on Random and Sequential Ar- Behavior Under Paging, Proc. 21st National Conf., ACM, Los rangements of Large Numbers of Records, in Information Angeles, Calif., Aug. 30-Sept. 1, 1966, pp. 223-228 (Thompson Processing 1965, Proc. IFIP Congress 65, Vol. 1, New York, Book Co., Washington, D.C., 1966). Also in Rept. No. SP-2397, N.Y., May 24-29, 1965. Ed. A. 255-260 W. Kalenich, pp. 19 p. (System Development Corp., Santa Monica, Calif., June (Spartan Books, Washington, D.C., 1965). 16, 1966). Duncan, C. J., Look! No Hands! in The Penrose Annual, A Fisher, R. O. and C. D. Shepard, Time Sharing on a Computer Review of the Graphic Arts, Vol. 121-167 57, pp. (Hastings with a Small Memory, Commun. ACM 10, No. 2, 77-81 (Feb. House, New York, 1964). 1967). Duncan, C. J., General Comment, in Advances in Computer Five Compatible Computers from Honeywell, Data Proc. 7, Typesetting, Proc. Int. Computer Typesetting Conf., Sussex, No. 5, 278-289 (Sept.-Oct. 1965). England, July 14-18, 1966, Ed. W. P. Jaspert, pp. x-xi (The Forgie, J. W., A Time- and Memory-Sharing Executive Program Institute of Printing, London. 1967). for Quick-Response On-Line Applications, AFIPS Proc. Fall Dunning, A. W., Data Copy Dilemmas, Systems 6, No. 6, 13-18 Joint Computer Conf., VoL 27, Pt. 1, Las Vegas, Nev., Nov. 30- (1965). Dec. 1. 1965. pp. 599-609 (Spartan Books, Washington. D.C., 1965).

Ebersole. J. L.. North American Aviation's National Operating Forte, A., Music and Computing: The Present Situation, AFIPS System, in Toward a National Information System, Second Proc. Fall Joint Computer Conf., Vol. 31, Anaheim, Calif., Annual National Colloquium on Information Retrieval, Phila- Nov. 14-16, 1967, pp. 327-329 (Thompson Books, Washington, delphia, Pa., April 23-24, 1965, Ed. M. Rubinoff, pp. 169-198 D.C., 1967). (Spartan Books. Washington, D.C., 1965). Franco, A. G., N. Marchand and L. J. Saporta, Error-Control Egbert, R. L., The Computer in Education: Malefactor or Bene- Systems Get The Message Across, Electronics 38, No. 23, factor, AFIPS Proc. Fall Joint Computer Conf., Vol. 24, Las 125-136 (1965). Vegas, Nev., Nov. 1963, pp. 619-630 (Spartan Books, Bahi- Frank, W. L., On-Line CRT Displays: User Technology and more, Md., 1963). Software, in On-Line Computing Systems, Proc. Symp. sponsored by the Univ. of California, Los Angeles, and Infor- Eichelberger, E. B., W. C. Rodgers and E. W. ^tacy. Method for matics, Inc.. Feb. 50-62 Estimation and Optimization of Printer Speed Based on Char- 1965, Ed. E. Burgess, pp. (American Data Proc, Inc., Detroit, Mich., 1965). acter Usage Statistics, IBM J. Res. & Dev. 12, 130-139 Frey, H. C, C. E. Nelson and H. E. Rubin, A Line Density (Mar. 1968). Standard to Replace Background Density, Proc. Eleventh Engelman, C. MATHL.\B: A Program for On-Line Machine Annual Meeting and Convention, Vol. XI, 1962, Ed. V. D. Tate, Assistance in Symbolic Computations, AFIPS Proc. FaU pp. 103-111 (The National Microfilm Assoc., Annapolis, Md., Joint Computer Conf., VoL 27, Pt. 1, Las Vegas, Nev., Nov. 30- 1962). Dec. 1, 1965, pp. 413-421 (Spartan Books, Washington, D.C., Fubini, E. G., The Opening Address, in Second Cong, on the 1965). Information System Sciences, held at The Homestead, Hot Engvold, K. J. and J. L. Hughes, A Model for a Multifunctional Springs, Va., Nov. 1964, Ed. J. Spiegel and D. E. Walker, pp. Teaching System, Commun. ACM 10 , 339-342 (June 1967). 1-4 (Spartan Books, Washington, D.C., 1965). Estes. S. E.. H. R. Kerby, H. D. Maxey and R. M. Walker, Speech Fuller, R. H., Content-Addressable Memory Systems, Rept. Synthesis from Stored Data, IBM J. Res. & Dev. 8, 2-12 No. 63-25. 2 V. (Dept. of Engineering, Univ. of California, (Jan. 1964). Los Angeles, 1963). Estrin, G. and L. Kleinrock, Measures, Models and Measure- Fuller, R. H. and R. M. Bird, An Associative Parallel Processor ments for Time-Shared Computer Utilities, Proc. 22nd National with Application to Picture Processing, AFIPS Proc. Fall Joint Conf.. ACM, Washington, D.C., Aug. 29-31, 1967, pp. 85-96 Computer Conf., VoL 27. Pt. 1, Las Vegas, Nev., Nov. 30- (Thompson Book Co.. Washington, D.C., 1967). Dec. 1, 1965, pp. 105-116 (Spartan Books, Washington, D.C., Evans, D. C. and Y. Leclerc, Address Mapping and the J. 1965). Control of Access in an Interactive Computer, AFIPS Proc. Spring Joint Computer Conf., Vol. 30, Atlantic City, N.J., Gaines, R. S. and C. Y. Lee, An Improved Cell Memory, IEEE April 18-20, 1967, pp. 23-30 (Thompson Books, Washington, Trans. Electron. Computers EC- 14, No. 1, 72-75 (Feb. 1965). D.C.. 1967). Gallenson, L., A Graphic Tablet Display Console for Use Under Evans, D. S. and Katzenelson, Data Structure and Man- J. Time-Sharing. AFIPS Proc. Fall Joint Computer Conf., Vol. 31, Machine Communication for Network Problems, Proc. IEEE Anaheim, Calif., Nov. 14-16, 1967, pp. 689-695 (Thompson 11.35-1144 55, (July 1967). Books, Washington, D.C., 1967). Evans. T. G. and D. L. Darley, An Extension to Cur- DEBUG — Garvin, P. L., Ed., Natural Language and the Computer, 398 p. 321— rent Online Debugging Techniques, Commun. ACM 8, (McGraw-Hill, New York, 1963). 325 (May 1965). Gault, R. H. and G. W. Crane, Tactual Patterns from Certain Evans, T. G. and D. L. Darley, On-Line Debugging Techniques: Vowel Qualities Instrumentally Communicated from a Speaker Proc. Conf., Vol. 29, A Survey, AFIPS FaU Joint Computer to a Subject's Fingers, J. Gen. Psychol. 1, 353-359 (1928). Francisco. Calif.. Nov. 7-10, 37-50 (Spartan San 1966, pp. Geddes, E. W., R. L. Emrich and J. F. McMurrer, Feasibility Books, Washington, D.C., 1966). Report and Recommendations for New York State Identifica- tion and Intelligence System, Rept. No. TM-LO-1000/000/00, Fano, R. M., The MAC System: A Progress Report, Rept. No. 276 p. (System Development Corp., Santa Monica, Calif., MAC-TR-12, 24 p. (M.I.T., Cambridge, Mass., Oct. 9, 1964). Nov. 1, 1963). Fano, R. M.. The Computer Utility and the Community, 1967 Gentle, E. C, Jr., Data Communications in Business, 163 p. IEEE Int. Conv. Rec, Pt. 12, pp. 30-37. (American Telephone and Telegraph Co., New York, 1965). Fant, G.. .\coustic Analysis and Synthesis of Speech with Appli- Gerlach, G. T., The Videograph Text Editor, in Electronic Com- cations to Swedish, reprint from Ericsson Technics No. 1, position in Printing, Proc. Symp., Gaithersburg, Md., June Sweden, 19.59, 108 p. 1.5-16, 1967, NBS Special Pub. 295, Ed. R. W. Lee and R. W.

117 Worral, pp. 17-21 (U.S. Govt. Printing Office, Washington, Hershey, A. V., Calligraphy for Computers, NWL Report No. t

D.CFeb. 1968). 2101, 1 V. (U.S. Naval Weapons Lab., Dahlgren, Va., Aug. 1, •

GiU, S., The Changing Basis of Programming, in Information 1967). 1 Processing 1965, Proc. IFIP Congress 65. Vol. 1, New York, Hillegass, J. R., NCR's New Bid: The Century Series, Data N.Y., May 24-29, 1965, Ed. W. A. Kalenich, pp. 201-206 Proc. Mag. 10, No. 46-49, 52-54 4, (Apr. 1968). |! (Spartan Books, Washington, D.C., 1965). Hoadley, H. W., A Storage-Retrieval Display for Microfiche, Gilman, W. L., Drum Scanning Techniques for Digitizing and in Progress in Information Science and Technology, Proc.

' Recording Image Data, 1966 IEEE Int. Conv. Rec., Pt. 3, Am. Doc. Inst. Annual Meeting, Bol. 3, Santa Monica, Cahf., pp. 29-38. Oct. 3-7, 1966, pp. 29-34 (Adrianne Press, 1966). Giuliano, V. E., The Interpretation of Word Associations, in Hoagland, A. S., Storing Computer Data, Int. Sci. & Tech. Statistical Association Methods for Mechanized Documen- No. 37, 52-58 (Jan. 1965). tation, Symp. Proc, Washington, D.C., March 17-19, 1964, Hobbs, L. C, The Impact of Hardware in the 1970's, Data- NBS Misc. Pub. 269, Ed. M. E. Stevens et al., pp. 25-32 mation 12, No. 3, 36-44 (Mar. 1966). (U.S. Govt. Printing Office, Washington, D.C., Dec. 15, 1965). Hobbs, L. C, Display Applications and Technology, Proc. Gluck, S. E., Impact of Scratchpads in Design: Multifunctional IEEE 54, 1870-1884 (Dec. 1966). Scratchpad Memories in the Burroughs B8500, AFIPS Proc. Hogan, D. L., Speech as Computer Input and Output, 1966 IEEE Fall Joint Computer Conf., Vol. 27, Pt. 1, Las Vegas, Nev., International Convention Record, Vol. 14, Part 3, Computers, Nov. 30-Dec. 1, 1965, pp. 661-666 (Spartan Books, Washing- presented at IEEE Int. Conv., New York, N.Y., Mar. 21-25, [ ton, D.C., 1965). 1966, pp. 91-93 (The Inst, of Electrical and Electronics Engi- Goldhor, H., Ed., Proceedings 1963 Clinic on Library Applications neers, New York, 1966). of Data Processing, Univ. of Illinois, Apr. 28-May 1, 1963, Holm, B. E., Techniques and Trends in Effective Utilization of 176 p. (Graduate School of Library Science, Illinois Univ., Engineering Information, ASLIB Proc. 17, 134-169 (1965). Urbana, 1964). A. to Control Data Character Holt, W. (assignor Corporation), ( Gomolak. L. S., Better and Faster Design by Machine, Elec- Recognition Using Curve Tracing, U.S. Patent 3,142,818, ii tronics 37, 64-71 (June 1964). issuedjuly28, 1964(filedFeb. 21,1961),9p. Greenberger, M., The Two Sides of Time-Sharing, Datamation Hormann, A. M., Introduction to ROVER, an Information 33-36 (Nov. 1965). 11, Processor, Rept. No. FN-3487, 51 p. (System Development Gross, W. A., Information Storage and Retrieval, A State-of-the Corp., Santa Monica, Calif., Apr. 25, 1960). Art Report, Ampex READOUT, special issue, 9 p. (Ampex Hornbuckle, G. D., A Multiprogramming Monitor for Small Corp., Redwood City, Calif., 1967). Machines, Commun. ACM 10, No. 5, 273-278 (May 1967). Gunderson, D.C., W. L. Heimerdinger and J. P. Francis, Asso- Howerton, P. W., The Application of Modern Lexicographic ciative Techniques for Control Functions in a Multiprocessor, Techniques to Machine Indexing, in Machine Indexing: Rept. No. RADC-TR-66-573, 74 p. (Honeywell, Inc., Griffiss Progress and Problems, Proc. Third Institute on Information Air Force Base, N.Y., Dec. 1966). Storage and Retrieval, Washington, D.C., Feb. 13-17, 1961, Gurk, H. M. and J. Minker, The Design and Simulation of an pp. 326-330 (The American Univ., Washington, D.C., 1962). Information Processing System, J. ACM 8, 260-270 (1961). IBM System/360, Data Proc. 7, No. 5,290-302 (Sept.-Oct. 1965). Institute of Printing, Computer Type-Setting Conf., Report of Hamilton, M. L. and A. D. Weiss, An Approach to Computer- 1964, 245 (Pub. London, 1965). Aided Preliminary Ship Design, Rept. No. ESL-TM-228, Proc, London Univ., July p. Israel, D. R., System Engineering Experience with Automated 65 p. (Electronic Systems Lab., M.I.T., Cambridge, Mass., Jan. 1965). Command and Control Systems, in Information System Hanlon, A. G., Content- Addressable and Associative Memory Science and Technology, papers prepared for the Third Cong., 193-213 System: A Survey, IEEE Trans. Electron. Computers EC-15, scheduled for Nov. 21-22, 1966, Ed. D. E. Walker, pp. Washington, D.C., 1967). 509-521 (Aug. 1966). (Thompson Book Co., Hanlon, A. G., W. C. Myers and C. O. Carlson, Photochromic Iverson, J. and F. Yee, IBM's Flexible Addition to System 360, Electronics No. 24, 79-82 (1965). Micro-Image Technology, A Review, paper presented at the 38, Equipment Manual Symp., jointly sponsored by the American P., Ed., in Computer Typesetting, Proc. Ordnance Association and the Army Material Command, Jaspert, W. Advances f Int. Computer Typesetting Conf., Sussex, England, July 14-18, Detroit, Mich., Nov. 30-Dec. 2, 1965, 35 p. (The National

1966, 306 (The Institute of Printing, London, 1967). ; Cash Register Co., Hawthorne, CaUf., 1965). p. Johnson, L. R., On Operand Structure Representation, Storage, fi Hargreaves, B., J. D. Joyce, G. L. Cole, E. D. Foss, R. G. Gray, and Search, Rept. No. RC-603, 79 p. (Thomas J. Watson E. M. Sharp, R. J. Sippel, T. M. Spellman and R. A. Thorpe, Research Center, Yorktown Heights, N.Y., Dec. 5, 1961). Image Processing Hardware for a Man-Machine Graphical Johnson, T. E., Sketchpad III: A Computer Program for Drawing Communication System, AFIPS Proc. Fall Joint Computer in Three Dimensions, AFIPS Proc. Spring Joint Computer i Conf., Vol. 26, San Francisco, Cahf., Oct. 1964, pp. 363-386 Conf., Vol. 23, Detroit, Mich., May 1963, pp. 347-353 (Spartan (Spartan Books, Washington, D.C., 1964). Books, Bahimore, Md., 1963). Harris, C. M., A Study of the Building Blocks in Speech, J. Joyce, R. D. and S. Homa, Jr., High Speed Thermal Printing, Acoust. Soc. Amer. 25, 962 (1953). AFIPS Proc. Fall Joint Computer Conf., Vol. 31, Anaheim, . Harrison, M. C. and T. Schwartz, SHARER, A Time-Sharing J. Calif., Nov. 14-16, 1967, pp. 261-267 (Thompson Books, for the 6600, Commun. 10, 659-665 System ' CDC ACM Washington, D.C., 1967). (Oct. 1967). Just Merged: Copier and Computer, Bus. Automation 14, Hartsuch, P. Graphics Arts Progress ... in 1967, Graphic J., No. 12,50-51 (Dec. 1967). 52-57, 61 1968). Arts Monthly 40, (Jan. ^ Haskins, S. M., Moving Toward International Cataloging Agree- Kalagher, J. J., Micromation — Its Impact on the Photocomposing ment, ALA Bull. 54, 197-201 (1960). Industry, in Electronic Composition in Printing, Proc. Symp., f Electronics Hattery, L. H. and G. P. Bush, Eds., Automation and Gaithersburg, Md., June 15-16, 1967 NBS Special Pub. 295, Publishing, derived from a Symposium sponsored by in Ed. R. W. Lee and R. W. Worral, pp. 37-39 (U.S. Govt. Printing American Univ., held at Washington, D.C., May 1965, 206 p. Office, Washington, D.C., Feb. 1968). f (Spartan Books, Washington, D.C., 1965). Kalenich, W. A., Ed., Information Processing 1965, Proc. IFIP ; Retrieval, Hayes, R. M., Mathematical Models for Information Congress 65, Vol. 1, New York, N.Y., May 24-29, 1965, 304 p. in Natural Language and the Computer, Ed. P. L. Garvin, (Spartan Books, Washington, D.C., 1965). pp. 268-309 (McGraw-Hill, New York, 1963). Kalenich, W. A., Ed., Information Processing 1965, Proc IFIP Heilprin, L. B., Communication Engineering Approach to Micro- Congress 65, Vol. 2, New York, N.Y., May 24-29, 1965, pp. 1 forms, Am. Doc. 12, No. 3, 213-218 (July 1961). 305-648 (Spartan Books, Washington, D.C., 1966). 26-37 Proc. Mag. Herbert, E., Information Transfer, Int. Sci. & Tech. 51, Kehl, W. B., Computers and Literature, Data 7, 24-26 (1966). (July 1965).

118 Kehler. V. G.. Conversion to Linotron, in Electronic Composition Lavington, S. H. and L. E. Rosenthal, Some Facihties for Speech in Printing. Proc. Svmp.. Gaithersburg. Md.i 15-16. 1967, June Processing by Computer, The Computer J. 9, 330-339 (Feb. NBS Special Pub. 295. Ed. R. W. Lee and R. W. Worral, 1967). Govt. Printing pp. 119-128 (U.S. Office. Washington. D.C., Lee, F. F.. A Study of Grapheme to Phoneme Translation of Feb. 1968). English. Ph. D. Thesis in Electrical Engineering, 141 p. Kessler. M. M.. The M.I.T. Technical Information Project I. (M.I.T., Cambridge, Mass., 1966). Svstem Description. 25 p. (The Libraries. M.LT., Cambridge, Lee, F. F., Machine-to-Man Communication by Speech Part 1: Mass.. Nov. 2.1964). Generation of Segmental Phonemes from Text, AFIPS Proc. and P. Terragno. Techniques for King, D. W. J. Some Measuring Spring Joint Computer Conf., Vol. 32, Atlantic City, N.J., System Performance, in Parameters of Information Science, Apr. 30-May 2. 1968, pp. 333-338 (Thompson Book Co.. Proc. Am. Doc. Inst.. Annual Meeting. Vol. 1, Philadelphia, Washington, D.C.. 1968). 5-8, 393-398 Pa.. Oct. 1964. pp. (Spartan Books. Washington. Lee. M. O. and G. I. Campbell. Closing Session. Summary of D.C..1964). Area Discussions, in National Academy of Sciences — National King, G. W.. The Library of Congress Project, Lib. Res. & Tech. Research Council, Proc. Int. Conf. on Scientific Information, 90-93 Serv. 9, No. 1. (Winter 1965). Vol. 2, Washington, D.C., Nov. 16-21. 1958, pp. 1549-1562 Kleinrock. L. Sequential Processing Machines (S.P.M.) Analyzed (NAS-NRC, Washington, D.C.. 1959). with a Queuing Theorv Model. J. ACM 13, No. 2. 179-193 Lee, R. W. and R. W. Worral. Eds.. Electronic Composition in (Apr. 19661. Printing, Proc. Symp.. Gaithersburg, Md.. June 15-16. 1967, Kleinrock. L. Theory of Queues and Time-Shared Computer NBS Special Pub. 295, 128 p. (U.S. Govt. Printing Office, Systems. 1966 IEEE Region Six Annual Conf. Record, Vol. II. Washington, D.C.. Feb. 1968). pp. 491-500. 1966. Leiner. A. L., S. N. Alexander and R. P. Witt, DYSEAC, in Com- Kleinrock, L.. Time-Shared Svstems: A Theoretical Treatment. puter Development (SEAC and DYSEAC) at the National J. ACM 14, No. 2,242-261 (Apr. 1967). Bureau of Standards, Washington, D.C., NBS Circular 551. Knauft, G.. H. Lamparter and W. G. Spruth, Some New Methods issued Jan. 25. 1955. pp. 39-71 (U.S. Govt. Printing Office. for Digital Encoding of Voice Signals and for Voice Code Washington, D.C.. 1955). Translation, IBM J. Res. & Dev. 10, No. 3. 244-254 (May 1966). LeVier. P. W., Computer-Aided Design of a Fhp-Flop, Computer Knowlton. K. C A Computer Technique for Producing Animated Design 4, No. 11,62-68(1965). Movies, AFIPS Proc. Spring Joint Computer Conf.. Vol. 25, Levine, L., J. B. Wiesner, O. H. Straus, B. Howland, N. Wiener Washington. D.C., April 1964. pp. 67-87 (Spartan Books. and E. E. David, Jr.. Felix (Sensory Replacement), Quarterly Washington, D.C.. 1964), Progress Reports, Research Lab. of Electronics. Mass. Inst, Kochen, M., Ed., Some Problems in Information Science, 309 p. of Technology, Jan. 1949-April 1951. (The Scarecrow Press, New York, 1965). Levinthal. C. Computer Construction and Display of Molecular Kornblum. R. D., A Macro-View of Microfilm, Bus. Automation Models, in Proc. IBM Scientific Computing Symp. on Com- 12,No. 10..32-.36(Oct. 1965). puter-Aided Experimentation, Yorktown Heights, N.Y., Oct. Kozumplik, W. A. and R. T. Lange. Computer-Produced Micro- 11-13, 1965. pp. 315-325 (IBM Corp.. White Plains, N.Y..

film Library Catalog. Am. Doc. 18, No. 2, 67-80 (April 1967). 1966) . Kraft, F. R., The Reduction in Bulk Resulting from the Type- Levinthal. C. Molecular Model-Building By Computer, Sci. setting of Documents in Modern Composition Systems, in Amer. 214, 42-52 (June 1966).

Automation and Scientific Communication, Short Papers, Lewin, M. H., H. R. BeeUtz and J. Guarracini, Fixed Resistor- Pt. 2. papers contributed to the Theme Sessions of the 26th Card Memory, IEEE Trans. Electron. Computers EC-14, Annual Meeting. .\m. Doc. Inst., Chicago, lU., Oct. 6-11. 1963. 428-434 (1965). Ed. H. P. Luhn, pp. 275-276 (Am. Doc. Inst.. Washington. Lichtenberger, W. W. and M. W. Pirtle, A Facihty for Experi- D.C., 1963). mentation in Man-Machine Interaction, AFIPS Proc. Fall Krishnamoorthi, B. and R. C. Wood. Time-Shared Computer Joint Computer Conf., Vol. 27, Pt. 1, Las Vegas, Nev., Nov. 30- Operations with Both Inter-Arrival and Service Times Ex- Dec. 1. 1965, pp. 589-598 (Spartan Books, Washington, D.C., ponential. Rept. No. SP-2090 (System Development Corp., 1965). Santa Monica, Calif., Sept. 1965). Lickhder, J. C. R., Libraries of the Future, 219 p. (M.I.T. Press, Kubert. B.. J. Szabo and S. Giulieri. The Perspective Represen- Cambridge, Mass., 1965). tation of Functions of Two Variables, J. ACM 15, 193-204 Licklider, J. C. R., Man-Computer Interaction in Information (Apr. 1968). Systems, in Toward a National Information System, Second Kuehler, J. D. and H. R. Kerby, A Photo-Digital Mass Storage Annual National Colloquium on Information Retrieval. Phila- System, AFIPS Proc. Fall Joint Computer Conf., Vol. 29, delphia, Pa., April 23-24, 1965, Ed. M. Rubinoff, pp. 63-75 San Francisco, Calif.. Nov. 7-10, 1966. pp. 735-742 (Spartan (Spartan Books, Washington, D.C., 1965). Books, Washington. D.C.. 1966). Licklider. J. C. R., Interactive Information Processing, in Com- Kump, H. J. and P. T. Chang, Thermostrictive Recording on puter and Information Sciences — II, Proc. 2nd Symp. on Com- Permalloy FUms, IBM J. Res. & Dev. 10, No. 3, 255-260 puter and Information Sciences, Columbus, O., Aug. 22-24, (May 1966). 1966, Ed. J. T. Tou, pp. 1-13 (Academic Press, New York,

1967) .

Lamberts. R. L. and G. C. Higgins, A System of Recording Licklider, J. C. R. and W. E. Clark, On-Line Man-Computer Digital Data on Photographic Film Using Superimposed Grating Communications, AFIPS Proc. Spring Joint Computer Conf., Patterns. AFIPS Proc. Fall Joint Computer Conf.. Vol. 29, Vol. 21, San Francisco, Cahf., May 1-3, 1962, pp. 113-128 San Francisco. Calif., Nov. 7-10, 1966, pp. 729-734 (Spartan (National Press, Palo Aho, Calif., 1962). Books, Washington, D.C., 1966). Lindgren, N., Human Factors in Engineering, Part I — Man in the Lampson, B. W.. A Scheduling Philosophy for Multiprocessing Man-Made Environment, IEEE Spectrum 3, No. 3, 132-139 Systems, Commun. ACM 11, No. 5, 347-360 (May 1968). (Mar. 1966). Landwehr. J. B.. C. McLaughlin, H. Mueller, M. Lichstein and Lock, K., Structuring Programs for Multiprogram Time-Sharing S. V. Pollack, BLNSYS- A 1401 Operating System with Braille On-Line Apphcations, AFIPS Proc. Fall Joint Computer Conf., Capabilities. Commun. ACM 8, 300-303 (May 1965). Vol. 27, Pt. 1, Las Vegas, Nev., Nov. 30-Dec. 1, 1965, pp. 457- Lang. C. \.. R. B. Polansky and D. T. Ross. Some Experiments 472 (Spartan Books, Washington, D.C., 1965). with an .\lgorithmic Graphical Language. Rept. No. ESL-TM- Loomis, H. H., Jr., Graphical Manipulation Techniques Using 220. 55 p. (M.I.T., Cambridge, Mass., Aug. 1965). the Lincoln TX-2 Computer, Rept. No. 51-G-0017, 27 p. Langefors, B., .State of the Art in Sweden, Data Proc. Mag. 7, (Lincoln Lab., M.I.T., Lexington, Mass., Nov. 10, 1960). .\o. 5,30-31 (1965). Lourie. J. R.. J. J. Lorenzo and A. Bomberault, On-Line Textile Lannon. E. R.. Computers and Composition in the U.S. Govern- Designing. Proc. 21st National Conf., ACM, Los Angeles, ment—Past. Present and Future, in Advances in Computer Cahf., Aug. 30-Sept. 1, 1966, pp. 537-544 (Thompson Book Typesetting, Proc. Int. Computer Typesetting Conf.. Sussex. Co., Washington, D.C., 1966). England, July 14-18. 1966. Ed. W. P. Jaspert, pp. 80-84 Loutrel, P., Determination of Hidden Edges in Polyhedral fThe Institute of Printing, London, 1967). Figures: Convex Case, Rept. No. TR-400-145 (Laboratory

119 for Electroscience Research, New York Univ., New York, Microfilm Experts Scan the Future, Systems 8, No. 1, 28-29 Sept. 1966). (Jan. 1967). P., Luhn, H. Ed., Automation and Scientific Communication, Mikhailov, A. I., Problems of Mechanization and Automation Short Papers, Pt. 2, papers contributed to the Theme Sessions of Information Work, Rev. Int. Doc. 29, 49-56 (1962). of the 26th Annual Meeting, Am. Doc. Inst., Chicago, III, Miller, J. C. and C. M. Wine, A Simple Display for Characters Oct. 6-11, 196.3, pp. 129-352 (Am. Doc. Inst., Washington, and Graphics, IEEE Trans. Computers C-17, 470-475 (May D.C.. 1963). 1968). Lynch, W. C, Description of a High Capacity, Fast Turnaround Miller, J. R., On-Line Analysis for Social Scientists, Rept. No. University Computing Center, Commun. ACM 9, 117-123 MAC-TR-40, 25 p. (M.I.T., Cambridge, Mass., May 1967). (Feb. 1966). Mills, R. G., Man-Computer Interaction — Present and Future, 1966 IEEE Int. Conv. Rec, Pt. 6. pp. 196-198. Machover, C, Graphic CRT Terminals — Characteristics of Mills, R. G.. Man-Machine Communication and Problem Solving, Commercially Available Equipment, AFIPS Proc. Fall Joint in Annual Review of Information Science and Technology, Computer Conf., Vol. 31, Anaheim, Calif., Nov. 14-16, 1967, Vol. 2, Ed. C. A. Cuadra, pp. 223-254 (Interscience Pub., 149-159 pp. (Thompson Books, Washington, D.C., 1967). New York, 1967). Mahan, R. E., A State of the Art Survey of the Data Display Mills, R. G. and T. H. Van Vleck, Administrative Facilities and Field, Rept. No. v. BNWL-725, 1 (AFC Research and Develop- System Management, M.I.T. Project MAC Progress Report III, ment Report, Battelle Northwest, Richland, Wash., May 1968). 1966, pp. 3-8. Manacher, G. K., Production Stabilization and of Real-Time Minker, J. and J. Sable, File Organization and Data Management, Task Schedules, J. ACM 14, No. 7, 439-465 (July 1967). in Annual Review of Information Science and Technology, Marill, T. and L. G. Roberts, Toward a Cooperative Network Vol. 2, Ed. C. A. Cuadra, pp. 123-160 (Interscience Pub., of Time-Shared Computers. AFIPS Proc. Fall Joint Computer New York, 1967). Conf.. Vol. 29. San Francisco, Calif., Nov. 7-10. 1966, pp. 425- Mooers, C. N., Information Retrieval Selection Study. Part II: 431 (Spartan Books, Washington, D.C., 1966). Seven System Models, Rept. No. ZTB-133, Part II, 39 p. Markuson. B. E., Ed., Libraries and Automation, Proc. Conf. (Zator Co., Cambridge, Mass., Aug. 1959). held at Airlie Foundation, Warrenton, Va.. May 26-30, 1963, Mooers, C. N., The Tape Typewriter Plan: A Method for Coopera- 268 p., under sponsorship of the Library of Congress, the Na- tion in Documentation, Rept. No. ZTB-137, 22 p. (Zator Co., tional Science Foundation, and the Council on Library Re- Boston, Mass., July 1960). Also in ASLIB Proc. 12, 277- sources (Library of Congress, Washington, D.C.. 1964). 291 (1960). Martin, D. and G. Estrin, Models of Computations and Systems- Mooers, C. N., An Emerging Revolution — The Reactive Type- Evaluation of Vertex Probabilities in Graph Models of Compu- writer, in Colloquium on Technical Preconditions for Retrieval tations, J. ACM 14, No. 2, 281-299 (Apr. 1967). Center Operations, Proc. National Colloquium on Information Martin, A., W. A Step-by-Step Computer Solution of Three Retrieval, Philadelphia. Pa., April 24-25, 1964, Ed. B. F. Problems in Non-Numerical Analysis, Project MAC Memo. Cheydleur, pp. 29-39 (Spartan Books. Washington, D.C., 1965). M-323 (M.I.T.. Cambridge, Mass., 1966). Moore, R. T., M. C. Stark and L. Calm, Digitizing Pictorial Infor- Mathews, M. V. and J. E. Miller, Computer Editing, Typesetting mation with a Precision Optical Scanner, Photogrammetric and Image Generation. AFIPS Proc. Fall Joint Computer Engineering 30, 923-931 (Nov. 1964). Pt. 1, 1965, Conf., Vol 27, Las Vegas, Nev., Nov. 30-Dec. 1, Morenoff, E. and J. B. McLean, A Code for Non-Numeric Infor- pp. 389-397 (Spartan Books, Washington, D.C.. 1965). mation Processing Applications in Online Systems, Commun. Mayeda, T., Methodology in Presentation, transcript of lecture ACM 10, No. 1, 19-22 (Jan. 1967). delivered for the Institute of Management Sciences, Wash- Morey, J. L. and D. B. Yntema, Experiments on Systems, in ington, D.C., Chapter, Nov. 1, 1967, 20 p. Second Cong, on the Information System Sciences, held at in Mayo, C. R., Electrostatic Printing Today and Tomorrow, The Homestead, Hot Springs, Va., Nov. 1964, Ed. J. Spiegel Research and Engineering Council of the Graphic Arts In- and D. E. Walker, pp. 349-368 (Spartan Books, Washington, dustry, Proc. 14th Annual Conf., Rochester, N.Y., May 18-20. D.C., 1965). 1964, pp. 74-79 (Washington, D.C., 1964). Morris, D., F. H. Sumner and M. T. Wyld, An Appraisal of the Mayzner, M. S. and M. E. Tresselt, Studies in Sequential Per- Atlas Supervisor, Proc. 22nd National Conf., ACM, Washing- at Another, Percep- ception: Looking One Thing But Finding ton. D.C., Aug. 29-31, 1967, pp. 67-75 (Thompson Book Co., tual and Motor Skills 23, 257-258 (1966). Washington, D.C., 1967). Mazzarese. N. J., Experimental System Provides Flexibility and Muerzeig, W., Towards More Versatile Teaching Machines, 68-70 Continuity, Data Proc. Mag. 7, No. 5, (1965). Computers & Automation 4, 22-25 (Mar. 1965). McCarn, D. B., Large-Scale System Design Techniques, in Second Cong, on the Information System Sciences, held at Neilsen, I. R.. A Simple Data Transmission System Using the The Homestead, Hot Springs, Va., Nov. 1964, Ed. J. Spiegel Office Telephone, Commun. ACM 8, 634-635 (Oct. 1965). 95-98 and D. E. Walker, pp. (Spartan Books, Washington, Nelson, C. E., in System Magazine's Third Annual Microfilm D.C., 1965). Seminar, 1965, p. 39. McCarthy, J., S. Boilen, E. Fredkin and J. C. R. Licklider, Nelson, T. H., Suggestions for an On-Line Braille Display, Proc. A Time-Sharing Debugging System for a Small Computer, SID 6th National Symp. on Information Display, New York, AFIPS Proc. Spring Joint Computer Conf., Vol. 23, Detroit, N.Y., Sept. 29-30, 1965 (Western Periodicals Co., North Holly- Mich., May 1963, pp. 51-57 (Spartan Books, Baltimore, Md., wood, Calif., 1965). Abstract in Inf. Display 2, No. 5,69 (Sept.- 1963). Oct. 1965). McCarthy, M. P., Generalized Graphic I/O: The Problem and an New Mass Core Memory, Computer Design 4, No. 6, 40-42 Approach, Data Proc. Mag. 9, 64-65 (Nov. 1967). (June 1965). McDonald, D. F., J. D. Freeman, C. B. Uzeta, and J. Dove, The New Microfiche Standards, Systems 7, No. 1, 35-36, 39 Dove Data Storage and Retrieval System, Rept. No RADC- (Jan. 1966). TDR-64^307, 19 p. (Braddock, Dunn and McDonald, Inc., Nicholson, T. A. J. and R. D. PuUen, A Permutation Procedure El Paso, Texas, Oct. 1964). for Job-Shop Scheduling, The Computer J. 11, 48-56 (May McDonald, R. F., Audio Response Unit Using a Digitally Stored 1968). Vocabulary, 1966 IEEE Int. Conv. Rec, Pt. 1, pp. 52-55. Ninke, W. H., Graphic 1 — A Remote Graphical Display Console, McGee, W. C.. On Dynamic Program Relocation, IBM Sys. J. 4, AFIPS Proc. FaU Joint Computer Conf., Vol. 27. Pt. 1, Las No. 3, 184-199(1965). Vegas, Nev., Nov. 30-Dec. 1, 1965, pp. 839-855 (Spartan Mehck. L. F.. The Easy Way Out -and In, Data Proc. Mag. 9, Books, Washington, D.C., 1965). 42-46 (Mar. 1967). Nisenoff, N., Hardware for Information Processing Systems: Mergenthaler Linotype Co., Report on the Study and Investiga- Today and in the Future, Proc. IEEE 54, 1820-1835 (Dec. tion for the Design of a Lexical-Graphical Composer-Printer 1966). System, RADC-TDR-63-63, Final Rept. on Contract AF Noll, A. M., Computer-Generated Three-Dimensional Movies, 30(602)2624, 1 v. (Brooklyn, N.Y., Jan. 31, 1963). in Data Processing, VoL X, Proc. 1966 Int. Data Processing

120 Conf.. Chicago. lU., June 21-24, 1966, pp. 143-148 (Data 22nd National Conf., ACM, Washington, D.C., Aug. 29-31, Processing Management Assoc, 1966). 1967, pp. 35-40 (Thompson Book Co., Washington, D.C., 1967). NoU. A. M., Stereographic Projections by Digital Computer, Prince. M. D., Man-Computer Graphics for Computer-Aided in Data Processing, Vol. X, Proc. 1966 Int. Data Processing Design, Proc. IEEE 54, 1698-1708 (Dec. 1966). Conf.. Chicago, 111., June 21-24. 1966, pp. 149-156 (Data Proctor, J. W., Jr., The Voice Response System, Datamation Processing Management Assoc., 1966). 12, No. 8, 43-44 (Aug. 1966). NoU, A. M., A Computer Technique for Displaying n-Dimensional Prywes, N. S.. A Storage and Retrieval System for Real-Time

Hyperobjects, Commun. ACM 10 , 469-473 (Aug. 1967). Problem Solving, Rept. No. 66-05, 47 p. (Moore School of Engineering, Univ. of Pennsylvania, Philadelphia, June 1, Ohlman, H., State-of-the-Art: Remote Interrogation of Stored 1965). Documentary Material, in Automation and Scientific Com- Prywes, N. S., Executive and Retrieval Based Extended Machine, munication. Short Papers, Pt. 2, papers contributed to the in Information Processing 1965, Proc. IFIP Congress 65, Vol. 2, Theme Sessions of the 26th Annual Meeting, Am. Doc. Inst., New York, N.Y., May 24-29. 1965, Ed. W. A. Kalenich, pp. Chicago, m., Oct. 6-11, 1963, Ed. H. P. Luhn, pp. 193-194 460-461 (Spartan Books, Washington, D.C., 1966). (Am. Doc. Inst.. Washington. D.C., 1963). Prywes. N. S. and H. S. Gray. The Organization of a Multilist- Opler. A., Dynamic Flow of Programs and Data Through Hier- Type Associative Memory, Commun. & Electronics 82, archical Storage, in Information Processing 1965, Proc. IFIP 488-491 (1963). Congress 65, Vol. 1, New York, N.Y., May 24-29, 1965, Ed. Pyke, T. N., Jr., Computer Technology: A Forward Look, NBS W. A. Kalenich, pp. 273-276 (Spartan Books, Washington, Tech. News Bull. 51, No. 8, 161-163 (Aug. 1967). Also in D.C..1965). Yale Scientific, pp. 14-15, 28 Oct. 1967. Opler, A., Requirements for Real-Time Languages, Commun. Rabiner. L.. ACM 9, No. 3, 196-199 (Mar. 1966). Speech Synthesis by Rule: An Acoustic Domain Oppenheimer, G. and N. Weizer. Resource Management for a Approach. Bell Sys. Tech. J. 47, 17-37 (Jan. 1968). Medium Scale Time-Sharing Operating System, Commun. Ramamoorthy. C. V., The Analytic Design of a Dynamic Look Ahead and Program Segmenting System for Multiprogrammed ACM 1 1 , 313-322 (May 1968). Orchard-Hays. W.. Operating Systems for Job-to-Job Running Computers, Proc. 21st National Conf., ACM, Los Angeles, and for Special Applications: Differences and Similarities, Calif., Aug. 30-Sept. 1. 1966, pp. 229-239 (Thompson Book in Information Processing 1965, Proc. IFIP Congress 65, Co., Washington, D.C., 1966). Vol. 1, New York, N.Y., May 24-29, 1965, Ed. W. A. Kalenich, Randell, B. and C. J. Kuehner, Dynamic Storage Allocation pp. 237-242 (Spartan Books, Washington, D.C., 1965). Systems, Commun. ACM 11, No. 5, 297-306 (May 1968). Ossorio, P. G.. Dissemination Research. Rept. No. RADC-TR- Reinfelds, J., L. A. Flenker, R. N. Seitz and P. L. Clem, Jr., 65-314. 75 p. (Rome Air Development Center, Griffiss Air AMTRAN, A Remote-Terminal, Conversational-Mode Com- Force Base, New York, Dec. 1965). puter System, Proc. 21st National Conf., ACM, Los Angeles, O'SuUivan, T. C, Exploiting the Time-Sharing Environment, Cahf., Aug. 30-Sept. 1, 1966, pp. 469-477 (Thompson Book Proc. 22nd National Conf., ACM, Washington, D.C., Aug. 29- Co.. Washington, D.C., 1966). 31. 1967, pp. 169-175 (Thompson Book Co., Washington, D.C., Reiter, A.. A Resource-Allocation Scheme for Multi-User On- 1967). Line Operation of a Small Computer, AFIPS Proc. Spring Joint Computer Conf., Vol. 30, Atlantic City, N.J., April 18-20, Peele. D., Bind or Film: Factors in the Decision, Lib. Res. & Tech. 1967, 1-7 (Thompson Books, Washington, D.C., 1967). Serv. 8, 168-171 (Spring 1964). pp. Resnick, A., The Information Explosion' and the User's Need Peirce, J. G. and W. B. Shannon, Interaction Between a Com- for in of Information Science, Proc. puterized Retrieval System and a Commercially Available Hard Copy, Parameters Am. Doc. Inst. Annual Meeting, Vol. 1, Philadelphia, Pa., Microfilm Cartridge File for Backup Data From Advance Oct. 5-8, 1964, pp. 315-317 (Spartan Books, Washington, Locations, in Levels of Interaction Between Man and Informa- D.C.. 1964). tion, Proc. Am. Doc. Inst. Annual Meeting, Vol. 4, New Richter, A. Z-39 Today: Its Work and Its Subcommittees, York, N.Y., Oct. 22-27. 1967. pp. 264-267 (Thompson Book J.. Spec. Lib. 107-109 Co.. Washington, D.C., 1967). 54, (1963). Riley, W. B., Time-Sharing: One Machine Serving Many Masters, Perry, M. .\.. Handling Very Large Programs, in Information Electronics No. 24, 72-78 (1965). Processing 1965, Proc. IFIP Congress 65, Vol. 1, New York, 38, Rippy, D. E., D. E. Humphries and A. Cunningham, — N.Y.. May 24-29, 1965. Ed. W. A. Kalenich, pp. 243-247 J. MAGIC (Spartan Books, Washington, D.C., 1965). A Machine for Automatic Graphics Interface to a Computer, Proc. Fall Joint Computer Conf.. Vol. 27. Pt. 1. Las Peterson, R. M., New Techniques for Data Display, in Data AFIPS Vegas, Nev., Nov. 30-Dec. 819-830 (Spartan Processing, Vol. X, Proc. 1966 Int. Data Processing Conf., 1, 1965, pp. Books, Washington, D.C., 1965). Chicago. III.. June 21-24, 1966, pp. 131-142 (Data Processing Management Assoc.. 1966). Risberg, A., Fundamental Frequency Tracking, Offprint, Proc. Pick. G. G. and D. B. Brick. A Read-Only Muhi-Megahit Parallel Int. Congress of Phonetic Sciences, 1961, pp. 228-231 (Mouton Search Associative Memory, in Automation and Scientific & Co.'s, Gravenhage. 1962). Roberts. K. V., Readability of Computer Programs, Communication. Short Papers, Pt. 2. papers contributed to The The 17-24 the Theme Sessions of the 26th Annual Meeting, Am. Doc. Computer Bull. 10, (Mar. 1967). Roberts, L. G., Graphical Communication and Control Languages, Inst., Chicago, III, Oct. 6-11. 1963. Ed. H. P. Luhn, pp. 245- in Cong, on the Information System Sciences, held at 246 (Am. Doc. Inst., Washington, D.C., 1963). Second The Homestead, Hot Springs, Va., Nov. 1964, Ed. J. Spiegel Pickett. J. M.. Transmitting Speech Sounds by a Tactual Vo- and D. E. Walker, pp. 211-217 (Spartan Books, Washington, coder and by Lip-Reading, Rept. No. 27, 35 p. (The Speech D.C., Transmission Lab., Div. of Telegraphy-Telephony, The Royal 1965). Perception Solids, Institute of Technology, Stockholm, Sweden, Mar. 1963). Roberts, L. G., Machine of Three-Dimensional Electro-Optical Information Processing, Ed. PoUand, C. B., Advanced Concepts of Utilization of Mass Storage, in Optical and al., 159-197 (M.I.T. Press, Cambridge, in Information Processing 1965, Proc. IFIP Congress 65, J. R. Tippett et pp. Vol. 1, New York, N.Y.. May 24-29, 1965, Ed. W. A. Kalenich, Mass., 1965). pp. 249-254 (Spartan Books, Washington, D.C.. 1965). Rocchio, J. J., Possible Time-Sharing Organization for a SMART Porter, J. W. and L. E. Johnson. United Air Lines' Electronic Retrieval System, in Information Storage and Retrieval, Information System (EIS), in Data Processing, Vol. X, Proc. Scientific Rept. No. lSR-7, Ed. G. Sahon, pp. XII-1 to XII-15 1966 Int. Data Processing Conf., Chicago, 111., June 21-24, (Computation Lab., Harvard Univ., Cambridge, Mass., June 1966. pp. 74-82 (Data Processing Management Assoc., 1966). 1964). Potter, R. J. and A. A. Axelrod, Optical Input/Output Systems RoUert, D. H., The Linotron System, in Electronic Composition for Computers, 1968 WESCON Technical Papers, 16/2, in Printing, Proc. Symp., Gaithersburg, Md., June 15-^16, Aug. 1968, 6 p. 1967, NBS Special Pub. 205, Ed. R. W. Lee and R. W. Worral, Press, L. I. and M. S. Rogers, IDEA — A Conversational, Heuristic pp. 7-14, (U.S. Govt. Printing Office, Washington, D.C., Feb. Program for Inductive Data Exploration and Analysis, Proc. 1968).

121 Rome Air Development Center, A Summary of the State-of-the- Schultz, C. K. and P. A. Schwartz, A Generalized Computer Art in Microfilm Document Storage and Retrieval Systems, Method for Index Production, Am. Doc. 13, No. 4, 420-432 Rept. No. RADC-TR-67-496, 114 p. (The Staff of Recon- (Oct. 1962). naissance and Intelligence Data Handling Branch, Griffiss Schwartz, J. I., E. G. Coffman and C. Weissman, Potentials of Air Force Base, N.Y., Sept. 1967). a Large-Scale Time-Sharing System, in Second Cong, on the Roos, D., An Integrated Computer System for Engineering Information System Sciences, held at The Homestead, Problem Solving, AFIPS Proc. Fall Joint Computer Conf., Hot Springs, Va., Nov. 1964, Ed. J. Spiegel and D. E. Walker, Vol. 27, Pt. 1, Las Vegas. Nev., Nov. 30-Dec. 1, 1965, pp. pp. 15-32 (Spartan Books, Washington, D.C., 1965). 423-433 (Spartan Books, Washington, D.C., 1965). Schwartz, J. I. and C. Weissman, The SDC Time-Sharing Rosa, J., Command Control Display Technology Since 1962, System Revisited, Proc. 22nd National Conf., ACM, Wash- in Second Cong, on the Information System Sciences, held at ington, D.C., Aug. 29-31, 1967, pp. 263-271 (Thompson The Homestead, Hot Springs, Va., Nov. 1964, Ed. J. Spiegel Book Co., Washington, D.C., 1967). 411-414 and D. E. Walker, pp. (Spartan Books, Washington, Schwelling, G., One View of the Standards, Systems 7, No. 1, D.C., 1965). 36 (Jan. 1966). Ross. D. T. E. Rodriguez, Theoretical and J. Foundations for the Seehof, J. M., W. O. Evans, J. W. Friedericks and J. J. Quigley, Computer-Aided Design System, AFIPS Proc. Spring Joint Automated Facilities Layout Programs, Proc. 21st National Computer Conf., Vol. Detroit, Mich., 1963, 305- 23, May pp. Conf., ACM. Los Angeles, Calif., Aug. 30-Sept. 1, 1966, pp. 322 (Spartan Books, Baltimore, Md., 1963). 191-199 (Thompson Book Co., Washington, D.C., 1966). Roudabush, G. E.. C. R. T. Bacon, R. P. Briggs, J. A. Fierst, Serlin, O., Automatic '3-D' Plotting with a Hybrid Computing D. W. Isner and H. A. Nogun, The Left Hand of Scholarship: System, Simulation 7, 285-291 (1966). Computer Experiments with Recorded Text as a Communi- Shalla, L., Automatic Analysis of Electronic Digital Circuits cation Media, AFIPS Proc. FaU Joint Computer Conf., Vol. Using List Processing, Commun. ACM 9, No. 5, 372-380 27, Pt. 1, Las Vegas, Nev., Nov. 30-Dec. 1, 1965, pp. 399-411 (May 1966).

(Spartan Books, Washington, D.C., 1965). Shaw, J. C, JOSS: A Designer's View of an Experimental On- iiubinoff, M., Ed., Toward a National Information System, Sec- Line Computing System, AFIPS Proc. Fall Joint Computer ond Annual National Colloquium on Information Retrieval, Conf., Vol. 26, San Francisco, Calif., Oct. 1964, pp. 455-464 Philadelphia, Pa., April 23-24, 1965, 242 p. (Spartan Books, (Spartan Books, Washington, D.C., 1964). Washington, D.C., 1965). Shemer, J. E., Some Mathematical Considerations of Time- A., Ruyle, J. W. Brackett and R. Kaplow, The Status of Systems Sharing ScheduUng Algorithms, J. ACM 14, No. 2, 262-272 for On-Line Mathematical Assistance, Proc. 22nd National (April 1967). Conf., ACM. Washington. D.C., Aug. 29-31. 1967. pp. 151-167 Sherman, P. M., FLOWTRACE, A Computer Program for Flow- (Thompson Book Co., Washington, D.C., 1967). charting Programs, Commun. ACM 9, No. 12, 845-854 (Dec. 1966). Shiner, G., User in Optical Sackman, H., Time-Sharing versus Batch Processing: The Requirements, Character Recognition, Ed. G. L. Fischer, al., Experimental Evidence, AFIPS Proc. Spring Joint Computer Jr., et pp. 335-338 (Spartan Books, Washington, D.C., 1962). Conf., Vol. 32, Atlantic City, N.J., Apr. 30-May 2, 1968, Silveira, H. M., Tele-Processing Terminals Offer User Wide pp. 1-10 (Thompson Book Co., Washington. D.C., 1968). Range of Applications Ability, Data Proc. Mag. 7, No. 7, 36- Salton, G., Ed., Information Storage and Retrieval, Scientific 39 (1965). Rept. No. ISR-7, 1 v. (Computation Lab., Harvard Univ., Silvern, G. M. and L. C. Silvern, Computer Assisted Instruction: Cambridge, Mass., June 1964). Specification of Attributes for CAI Programs and Program- Salton, G., An Evaluation Program for Associative Indexing, mers, Proc. 21st National Conf., ACM, Los Angeles, Calif., in Statistical Association Methods for Mechanized Documen- Aug. 30-Sept. 1, 1966, pp. 57-62-5 (Thompson Book Co., tation, Symp. Proc, Washington, D.C., March 17-19, 1964, Washington, D.C., 1966). NBS Misc. Pub. 269, Ed. M. E. Stevens et al, pp. 201-210 Silvern, G. M. and L. C. Silvern, Programmed Instruction and (U.S. Govt. Printing Office, Washington, D.C., Dec. 15, 1965). Computer-Assisted Instruction — An Overview, Proc. IEEE Salton, G., The European Computer Gap, Commun. ACM 10, 54, 1648-1655 (Dec. 1966). No. 4, 203 (April 1967). Slade, A. E. and H. O. McMahon, A Cryotron Catalogue Memory on a Multiconsole Computer, Samuel, A. L., Time-Sharing System, Proc. Eastern Joint Computer Conf., Vol. 10, Theme: (M.I.T., Project MAC, Cam- Rept. No. MAC-TR-17, 23 p. New Developments in Computers, New York, N.Y., Dec. 10- Mass., Mar. 1965). bridge, 12, 1956, pp. 115-120 (American Inst, of Electrical Engineers, Sawyer, J. and M. F. Friedell, The Interaction Screen: An New York. 1957). Operational Model for Experimentation on Interpersonal Soref, R. A. and D. H. McMahon, Bright Hopes for Display Sys- Behavior, Behav. Sci. 10, 446-460 (1965). tems: Flat Panels and Light Deflectors, Electronics 38, No. Sayer, J. S., The Economics of a National Information System, 24, 56-62 (1965). in Toward a National Information System, Second Annual Souder, J. J., W. E. Clark, J. I. Elkind and M. B. Brown, Planning National Colloquium on Information Retrieval, Philadelphia, for Hospitals, 167 p. (Amer. Hospital Assoc., Chicago, 111., Pa., April 23-24, 1965, Ed. M. Rubinoff, pp. 135-146 (Spartan 1964). Books, Washington, D.C., 1965). Sparks. D. E., L. H. Berul and D. P. Waite, Output Printing for Library Scarrott, G. G., The Efficient Use of Multilevel Storage, in Infor- Mechanization, in Libraries and Automation, Proc. of mation Processing 1965, Proc. IFIP Congress 65, Vol. 1, Conf. held at Airlie Foundation, Warrenton, Va., May 26-30, New York, N.Y., May 24-29, 1965, Ed. W. A. Kalenich, 1963, Ed. B. E. Markuson, pp. 155-190 (Library of Congress, Washington, D.C., 1964). pp. 137-141 (Spartan Books, Washington, D.C., 1965). Spencer, H., Ed., The Penrose Annual, A Review of the Graphic Schatz, S., Facsimile Transmission in Libraries: A State of the Arts, Vol. 57, 300 p. (Hastings House, New York, 1964). Art Survey, Tech. Memo. No. 1, 24 p. (Information Systems Spiegel, J. and D. E. Walker, Eds., Second Congress on the Office, Library of Congress, Washington, D.C., June 1967). Information System Sciences, held at The Homestead, Hot Schecter, G., Ed., Information Retrieval- A Critical View, Springs, Va., Nov. 1964, 525 p. (Spartan Books, Washington, 282 (Thompson Book Co., Washington, D.C., 1967). p. D.C., 1965). Computer Systems, Scherr, A. L., An Analysis of Time-Shared Stanga, D.C., Univac R 1108 Multiprocessor System, AFIPS 178 (M.I.T.. Ph. D. Dissertation, Rept. No. MAC-TR-18, p. Proc. Spring Joint Computer Conf., Vol. 30, Atlantic City, Cambridge, Mass., June 1965). N.J., April 18-20, 1967, pp. 67-74 (Thompson Books, Wash- Scholten, L. M., A Multiconsole Display Group for Use in an ington, D.C., 1967). Information Retrieval System, in Colloquium on Technical Stevens, M. E., A Machine Model of Recall, in Information Preconditions for Retrieval Center Operations, Proc. National Processing, Proc. Int. Conf. on Information Processing, Colloquium on Information Retrieval, Philadelphia. Pa., UNESCO, Paris, June 15-20, 1959, pp. 309-315 (Oldenbourg, April 24-25, 1964, Ed. B. F. Cheydleur, pp. 125-131 (Spartan Munich; Butterworths, London, 1960). Also preprint No. Books, Washington, D.C., 1965). UNESCO/NS/ICIP/J. 54, 14 p.

122 Stevens, M. E., Nonnumeric Data Processing in Europe: A Field Totschek, R. A., An Empirical Investigation into the Behavior Trip Report, August-October 1966, NBS Tech. Note 462, of the SDS Time-Sharing System, Rept. No. SP-2191 (System 63 p. (U.S. Govt. Printing Office, Washington, D.C., Nov. 1968). Development Corp., Santa Monica, Calif, Aug. 1965). Stevens, M. E., V. E. Giuliano and L. B. Heilprin, Eds., Statistical Tou, J. T., Ed., Computer and Information Sciences — II, 2nd Association Methods for Mechanized Documentation, Symp. Symp. on Computer and Information Sciences, Columbus, O., Proc, Washington, D.C., March 17-19, 1964, NBS Misc. Aug. 22-24, 1966, 368 p. (Academic Press, New York, 1967). Pub. 269, 261 p. (U.S. Govt. Printing Office, Washington, D.C., Travis, L. E., Analytic Information Retrieval, in Natural Lan- Dec. 15, 1965). guage and the Computer, Ed. P. L. Garvin, pp. 310-353 (Mc- Stevens, M.E. and J. L. Little, Automatic Typographic-Quality Graw-Hill Book Co., New York, 1963). Composition Techniques: A State-of-the-Art Report, NBS Monograph 99, 98 p. (U.S. Govt. Printing Office, Washington, Urbach, P., A Future Microsystem for the U.S. Patent Office, D.C.. Apr. 7. 1967). Proc. Eleventh Annual Meeting and Convention, Vol. Stotz, R., Man-Machine Console Facilities for Computer-Aided XI, 1962, Ed. V. D. Tate, 153-161 (The National Design, AFIPS Proc. Spring Joint Computer Conf., Vol. 23, pp. Microfilm Assoc.. Annapolis, Md., 1962). Detroit, Mich., May 1963, pp. 323-328 (Spartan Books, Balti- Urquhart, A. B., Voice Output from System/360, more, Md., 1963). IBM AFIPS Proc. Fall Joint Computer Conf., Vol. 27, Pt. Stotz, R. H., Directions in Time-Sharing Termineds, Computer 1, Las Vegas, Nev., Nov. 30-Dec. 1, 1965, 857-865 (Spartan Group News, IEEE 2, 12-18 (May 1968). pp. Books, Washington, D.C.. 1965). Stowe, A. N. and D. B. Hampton, Speech Synthesis with Pre- recorded Syllables and Words, J. Acoust. Soc. Amer. 33, 810 (1961). Van Dam, A. and J. C. Michener, Hardware Developments and Stowe. A. N., R. A. Wiesen, D. B. Yntema and J. W. Forgie, The Product Announcements, in Annual Review of Information Lincoln Reckoner: An Operation-Oriented, On-Line Facility Science and Technology, Vol. 2, Ed. C. A. Cuadra, pp. 187-222 with Distributed Control, AFIPS Proc. Fall Joint Computer (Interscience Pub., New York, 1967). Conf., Vol. 29, San Francisco, CaUf., Nov. 7-10, 1966, pp. Veaner, A. B., Developments in Copying Methods and Graphic 433-444 (Spartan Books, Washington, D.C., 1966). Communication 1965, Lib. Res. & Tech. Serv. 10, 199-210 Summit, R. K., DIALOG: An Operational On-Line Reference (Spring 1966). Retrieval System, Proc. 22nd National Conf., ACM, Washing- Ver Hoef, E. W., Design of a Multi-Level File Management ton, D.C., Aug. 29-31, 1967, pp. 51-56 (Thompson Book Co., System, Proc. 21st National Conf., ACM, Los Angeles, Calif.. Washington, D.C., 1967). Aug. 30-Sept. 1, 1966, pp. 75-86 (Thompson Book Co., Wash- ington, D.C., 1966). Sutherland, I. E., Sketchpad, A Man-Machine Graphical Com- munication System, AFIPS Proc. Spring Joint Computer Vlahos, P., The Three-Dimensional Display: Its Cues and Techniques, Inf. 10-20 Conf., Vol. 23, Detroit, Mich., May 1963, pp. 329-346 (Spartan Display 2, No. 6, (Nov.-Dec. 1965). A., Books, Baltimore, Md., 1963). Vogel, N. Walnut Document Storage and Retrieval System, Proc. Eleventh Annual Meeting and Vol. Sutherland, I. E., The Future of On-Line Systems, in On-Line Convention, XI, Computing Systems, Proc. Symp. sponsored by the Univ. of 1962, Ed. V. D. Tate, pp. 27-39 (The National Microfilm Assoc., California, Los Angeles, and Informatics, Inc., Feb. 1965, Annapolis, Md., 1962). Von Sydow, L., Computer Typesetting Ed. E. Burgess, pp. 9-13 (American Data Processing, Inc., of ALGOL, Commun. 172-174 Detroit, Mich., 1965). ACM 10, No. 3, (Mar. 1967). Swanson, R. W., Information System Networks — Let's Profit

from What We Know, in Information Retrieval— A Critical Wagner, F. V. and J. Granholm, Design of a General-Purpose View, Ed. G. Shecter, pp. 1-52 (Thompson Book Co., Wash- Scientific Computing Facility, in Information Processing 1965, ington, D.C., 1967). Proc. IFIP Congress 65, Vol. 1, New York, N.Y., May 24-29, Systems Magazine's Third Annued Microfilm Seminar, Systems 1965, Ed. W. A. Kalenich, pp. 283-289 (Spartan Books, 6, No. 8, 34-40 (Nov. 1965). Washington, D.C., 1965). Walker, D. E., Ed., Information System Science and Tech- Tate, F. A., Handling Chemical Compounds in Information Sys- nology—papers prepared for the Third Congress, scheduled tems, in Annual Review of Information Science and Tech- for Nov. 21-22, 1966, 406 p. (Thompson Book Co., Washington, nology, Vol. 2. Ed. C. A. Cuadra, pp. 285-309 (Interscience D.C., 1967). Pub.. New York. 1967). Walker, D. E., SAFARI, An On-Line Text-Processing System, Tate, V. D., Ed., Proc. Eleventh Annual Meeting and Convention, in Levels of Interaction Between Man and Information, Proc. Vol. XI, 1962. 360 p. (The National Microfilm Assoc., Annapolis, Am. Doc. Inst. Annual Meeting, Vol. 4, New York, N.Y., Md., 1962). Oct. 22-27, 1967, pp. 144-147 (Thompson Book Co., Wash- Tauber, A. S., Document Retrieval System Analysis and Design, ington, D.C., 1967). in Progress in Information Science and Technology, Proc. Wall, H. and H. Falk, Circuit Analysis by Computer, Electro- Am. Doc. Inst. Annual Meeting, Vol. 3, Santa Monica, Calif., Technology 78, No. 5, 50-56 (1966). Oct. 3-7. 1966, pp. 27.3-281 (Adrianne Press, 1966). Walter, C. J., Mathematical Techniques to Improve Hardware Tauber, A. S. and W. C. Myers, Photochromic Micro-Images: A Accuracy of Graphic Display Devices, AFlPS Proc. Spring Key to Practical Microdocument Storage and Dissemination, Joint Computer Conf., Vol. 30, Atlantic City, N.J., Apr. 18-20, Proc. Eleventh Annual Meeting and Convention, Vol. XI, 1967, pp. 107-120 (Thompson Books. Washington, D.C., 1967). 1962, Ed. V. D. Tate. pp. 257-269 (The National Microfilm Walter, E. S. and V. L. Wallace, Further Analysis of a Com- Assoc., Annapolis, Md.. 1962). Also in Am. Doc. 13, No. 4, puting Center Environment, Commun. ACM 10, No. 5, 403-409 (Oct. 1962). 266-272 (May 1967). Taylor, R. S., Professional Aspects of Information Science and Wang, W. S. Y. and G. E. Peterson, Segment Inventory for

Technology, in Annual Review of Information Science and Speech Synthesis, J. Acoust. Soc. Amer. 30 , 743 (1958). Technology. Vol. 1. Ed. C. A. Cuadra, pp. 15-40 (Interscience Wantman, M. E., CALCULAID: An On-Line System for Alge- Pub.. New York, 1966). braic Computation and Analysis, Rept. No. MAC-TR-20 Teitelman, W., PILOT: A Step Toward Man-Computer Sym- (thesis), 53 p. (M.I.T., Cambridge, Mass., Sept. 1965). biosis. Project MAC, Rept. No. MAC-TR-32, 193 p. (M.I.T., Ward, J. E., Systems Engineering Problems in Computer- Cambridge, Mass., Sept. 1966). Driven CRT Displays for Man-Machine Communication, Terlet, R. H.. The CRT Display Subsystem of the IBM 1500 IEEE Trans. Systems Sci. & Cybernetics SSC-3, 47-54 Instructional System, AFIPS Proc. Fall Joint Computer Conf., (June 1967). Vol. 31, Anaheim, Calif., Nov. 14-16, 1967, pp. 169-176 Ware, W. H., Security and Privacy: Similarities and Differences, (Thompson Books, Washington, D.C., 1967). AFIPS Proc. Spring Joint Computer Conf., Vol. 30, Atlantic Tippett, J. R., D. A. Berkowitz, L. C. Clapp, C. J. Koester and City, N.J., Apr. 18-20, 1967, pp. 287-290 (Thompson Books, A. Vanderburgh, Jr., Eds., Optical and Electro-Optical In- Washington, D.C., 1967). formation Processing, 780 p. (M.I.T. Press, Cambridge, Mass., Webster, E., The Second Generation Line Printer, Data froc.

1965). 7 , 42-43, 45 (1965).

123 Webster, E., Nonimpact Printout- What Is It? Where Is It Wilkes, M. V., The Design of Multiple-Access Computers Sys-

Going? Data Proc. Mag. 8, 32-33 (July 1966). tems, The Computer J. 10, 1-9 (May 1967). Wegner, P., Machine Organization for Multiprogramming, Proc. Woodward, P. M., The Synthesis of Music and Speech, The 22nd National Conf., ACM, Washington, D.C., Aug. 29-31, Computer J. 9, 257-262 (Nov. 1966). 1967, pp. 135-150 (Thompson Book Co., Washington, D.C., Wooster, H., Towards a Uniform Federal Report Numbering 1967). System and a Cuddly Microfiche Reader — Two Modest Pro- Weil, J. W., The Impact of Time-Sharing on Data Processing posals, Rept. No. AFOSR-68-0772, 13 p. (Air Force Office of Management, DPMA Q. 2, No. 2, 2-16 (1966). Scientific Research, Arlington, Va., Apr. 17, 1968). Weiss. R. A., Be VISION, A Package of IBM 7090 FORTRAN Wylie, C, G. Romney, D. Evans and A. Erdahl, Half-Tone Per- Programs to Draw Orthographic Views of Combinations of spective Drawings by Computer, AFIPS Proc. Fall Joint Plane and Quadric Surfaces, J. ACM 13, No. 2, 194-204 Computer Conf., Vol. 31. Anaheim, Calif., Nov. 14-16, 1967, (Apr. 1966). pp. 49-58 (Thompson Books, Washington, D.C., 1967). Whiteman, I. R., New Computer Languages, Int. Sci. & Tech. 52, 62-68 (1966). Yerkes, C. P., Microfiche, A New Information Media, in Auto- Wigington, R. L., Graphics as Computer Input and Output, 1966 mation and Scientific Communication, Short Papers, Pt. 2, IEEE Int. Conv. Rec. Pt. 3, Computers. New York. N.Y.. papers contributed to the Theme Sessions of the 26th Annual Mar. 21-25, 1966, pp. 86-90 (The Inst, of Electrical and Meeting, Am. Doc. Inst., Chicago. 111.. Oct. 6-11, 1963. Ed. Electronics Engineers, New York, 1966). H. P. Luhn, p. 129 (Am. Doc. Inst., Washington, D.C.. 1963). Wilkes, M. V., Slave Memories and Dynamic Storage Allocation, Yershov, A. P.. One View of Man-Machine Interaction, J. ACM IEEE Trans. Electron. Computers EC-14, No. 2, 270-271 12, No. 3, 315-325 (July 1965). (April 1965).

Additional References

Abrams, M. D., A Comparative Sampling of the Systems for Core Storage and TSS/360, Proc. 23d National Conf. ACM, Producing Computer-Drawn Flowcharts, Proc. 23rd National Las Vegas, Nev., Aug. 27-29, 1965, pp. 7-18 (Braudon/Systems Conf., ACM, Las Vegas. Nev.. Aug. 27-29. 1968. pp. 743-750 Press, Inc., Princeton, N.J., 1965).

(Brandon/Systems Press, Inc., Princeton, N.J., 1968). Fleisher, A. , P. PengeDy, J. Reynolds, R. Schools and G. Sincerbox, An Optically Accessed Memory Using the Lippman Process Baran, P., On Distributed Communications: IV. Priority, Prec- for Information Storage, in Optical and Electro-Optical edence, and Overload. Research Memo. RM-3638-PR, 63 p. Information Processing. Ed. J. R. Tippett et al., pp. 1-30 (The RAND Corp.. Santa Monica, Calif., Aug. 1964). (M.I.T. Press, Cambridge, Mass., 1965). Butler, E. J., Microfilm and Company Survival Plans. Proc. Eleventh Annual Meeting and Convention. Vol. XI. 1962, Ed. Glauthier, T. J., Computer Time Sharing: Its Origins and Devel- V. D. Tate. 57-68 (The National Microfilm Assoc.. Annap- pp. opment, Computers & Automation 16, 23-27 (Oct. 1967). ohs,Md., 1962). Gosden, J. A., The Operations Control Center MuH-Computer Operating System, Proc. 19th National Conf., ACM, Phil- Cantrell, H. N. and A. L. ElUson, Multiprogramming System adelphia, Pa., Aug. 25-27, 1964, pp. E2.2-1-E2.2-9 (Assoc. and Analysis, AFIPS Proc. Spring Performance Measurement For Computing Machinery, New York, 1964). Joint Computer Conf., Vol. 32, Atlantic City, N.J., Apr. 30- Greenway, R. D. and M. V. Russell, Let's Design an Information May 2. 1968, pp. 213-221 (Thompson Book Co.. Washington, System, p. 3 (preprint). D.C., 1968). Destructive-Readout Associative Memory, IEEE Chu, Y.. A Haddon, B. K. and W. M. Waite, A Compaction Procedure for Trans. Electron. Computers EC-14, No. 4, 600-605 (Aug. Variable-Length Storage Elements, The Computer J. 10, 1965). 162-165 (Aug. 1967). Harder, E. L., The Expanding World of Computers. Commun. Daley, R. C. and J. B. Dennis, Virtual Memory, Processes, and ACM 1 1 , No. 4. 231-239 (Apr. 1968). Sharing in MULTICS, Commun. ACM 11, No. 5, 306-312 Haring. D. R.. A Display Console for an Experimental Computer- (May 1968). Based Augmented Library Catalog. Proc. 23rd National Davis, R. M., Information Control In An Information System, Conf.. ACM, Las Vegas, Nev., Aug. 27-29, 1968, pp. 35-43 draft of lecture delivered to the Washington, D.C. Chapter, The (Brandon/Systems Press, Inc., Princeton, N.J., 1968). Institute of Management Sciences, Oct. 18, 1967. 49 p. Head, R. V., The Programming Gap in Real-Time Systems, Day, M. S.. Appfication of Basic Principles in the Design and Datamation 9, 39-41 (Feb. 1963). in a Operation of a Large Information System, Toward Na- Hoagland, A. S., Mass Storage Revisited, Proc. 22nd National tional Information System, Second Annual National Colloquium Conf., ACM, Washington, D.C, Aug. 29-31, 1967, pp. 255-260 on Information Retrieval, Philadelphia, Pa., Apr. 23-24, 1965. (Thompson Book Co., Washington. D.C. 1967). Ed. M. Rubinoff, pp. 123-133 (Spartan Books, Washington. Holt. A. W., File Structure and the General Problem of Repre- D.C. 1965). sentation. Proc. IFIP Congress 65. Vol. 2, New York. N.Y.. Denning, P. J.. The Working Set Model for Program Behavior. May 24-29, 1965, Ed. W. A. Kalenich, pp. 464-466 (Spartan Commun. ACM 1 1, No. 5, 32-3-333 (May 1968). Books, Washington, D.C, 1966). Howerton, P. W., The Case for Realism in Document Storage Fant, G., K. Fintoff, J. Liljencrants, B. Lindblom and J. Martony, Systems, Proc. Eleventh Annual Meeting and Convention, Formant-Amplitude Measurements, J. Acoust. Soc. Amer. 35, Vol. XI, 1962, Ed. V. D. Tate. pp. 21-25 (The National Micro- 1753-1761 (1963). film Assoc., Annapolis, Md., 1962). Feige, E. L. and H. W. Watts, Partial Aggregation of Micro- Hudson, D. M., The Applications and Implications of Large- Economic Data, 24 p. (Social Systems Research Institute, Scale Integrations, Computer Design 7, 38-42, 47-48 (June Univ. of Wisconsin, Madison, June 1967). 1968). Fife, D. W. and R. S. Rosenberg, Queueing in a Memory Shared Computer, Proc. 19th National Conf., ACM, Philadelphia, Klerer, M.. Final Report to the Advnced Research Projects Pa., Aug. 25-27, 1964, pp. Hl-1 to Hl-12 (Assoc. for Com- Agency Computer Science Program, 33 p. (Hudson Labs.. puting Machinery, New York, 1964). Columbia Univ., Dobbs Ferry. N.Y.. Dec. 1. 1966). Fikes, R. E., H. C. Lauer, and A. L. Vareha, Jr., Step Toward a Klerer. M. and J. May. Two-Dimensional Programming, AFIPS General-Purpose Time-Sharing System Using Large Capacity Proc. Fall Joint Computer Conf., Vol. 27, Pt. 1, Las Vegas.

124 Nev., Nov. 30-Dec. 1. 1965, pp. 63-75 (Spartan Books, Wash- Raphael, B., The Structure of Programming Languages, Commun. ington. D.C., 1965). ACM 9, No. 2, 67-71 (Feb. 1966). Reich, A. and G. H. Dorion, Photochromic, High-Speed, Large Kuipers, J. W., A. W. Tyler and W. L. Myers, A Minicard Optical and System for Documentary Information. Am. Doc. 8, No. 4, Capacity, Semirandom Access Memory, in 246-268 (Oct. 1957). Electro-Optical Information Processing, Ed. J. R. Tippett et al., pp. 567-580 (M.I.T. Press, Cambridge, Mass., 1965). Research on Archival Microfilm. NBS Tech. News Bull. 50, Landauer, W. I., The Balanced Tree and Its Utilization in In- formation Retrieval, IEEE Trans. Electron. Computers No. 9, 154-155 (Sept. 1966). Rosin, R. F., An Approach to Executive System Maintenance in EC-12, 863-871 (Dec. 1963). Disk-Based Systems, The Computer J. 9, 242-247 (Nov. 1966). Laski, J. G.. Segmentation and Virtual Address Topology— An Essay in Virtual Research, The Computer J. 11, 35-40 Smith. J. L., Multiprogramming under a Page on Demand Strat- (May 1968). egy, Commun. ACM 10, No. 10, 636-646 (Oct. 1967). Linde, R. R. and P. E. Chaney, Operational Management of Time-Sharing Systems. Proc. 21st National Conf., ACM, Los Taube, M. and H. Wooster, Eds., Information Storage and Angeles, Cahf.. Aug. 30-Sept. 1. 1966. pp. 149-159 (Thompson Retrieval Theory, Systems and Devices, Air Force Office of Book Co., Washington. D.C.. 1966). Scientific Research Symp., Washington, D.C., 1958, 228 p. (Columbia Univ. Press, New York, 1958). McCormick, B. H., The lUinois Pattern Recognition Computer — Taylor, R. S., The Process of Asking Questions, Am. Doc. 13, III, IEEE Trans. Electron. Computers EC-12, ILLIAC No. 4. 391-396 (Oct. 1962). 791-813 (Dec. 1963). Taylor, R. S., Question-Negotiation Seeking in Libraries, Rept. Mendelson, M. J. and A. W. England, The SDS SIGMA 7: A No. 3, Studies in the Man-System Interface in Libraries, 99 p. Proc. Fall Joint RejJ-Time Time-Sharing Computer, AFIPS (Center for the Information Sciences. Lehigh Univ.. Bethlehem. Francisco, Calif.. Nov. 7-10, Computer Conf., Vol. 29. San Pa., July 1967). 1966, pp. 51-64 (Spartan Books. Washington, D.C.. 1966). Meadow, C. T.. The Analysis of Information Systems: A Pro- Wall, E., The Grouping and Arrangement of Terms, Items, and grammer's Introduction to Information Retrieval. 301 p. Their Codes, in Information Storage and Retrieval Theory, (Wiley, New York. 1967). Systems and Devices, Air Force Office of Scientific Research Symp., Washington, D.C., 1958, Ed. M. Taube and H. Wooster, Pfeffer. H.. Ed.. Information Retrieval Among Examining pp. 170-189 (Columbia Univ. Press, New York, 1958). Patent Offices. Proc. Fourth Annual Meeting of the Com- Wente, V. A., Specificity and Accessibility in a System of In- mittee for International Cooperation in Information Retrieval formation Centers on Space and Aeronautics, in Colloquium Among Examining Patent Offices. ICIREPAT. Washington, on Technical Preconditions for Retrieval Center Operations. D.C., Oct. 7-16, 1964, 627 p. (Spartan Books, Washington, Proc. National Colloquium on Information Retrieval, Phila- D.C., 1966). delphia, Pa., Apr. 24-25, 1964, Ed. B. F. Cheydleur, pp. 55-60 Pointel, N. and D. Cohen, Computer Time Sharing- A Review, (Spartan Books, Washington, D.C., 1965). Computers & Automation 16,38-46 (Oct. 1967). Wigington, R. L., A Machine Organization for a General Purpose Prywes, N. S., Man-Computer Problem Solving with Multihst, List Processor, IEEE Trans. Electron. Computers EC-12, Proc. IEEE 54, 1788-1801 (Dec. 1966). 707-714 (Dec. 1963).

125

u s. GOVERNMENT PRINTING OFFICE : 1970 OL — 366-108 i —

NBS TECHNICAL PUBLICATIONS p 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. Building Science Series. Research results, test Comprehensive scientific papers give complete details methods, and performance criteria of building ma- of the work, including laboratory data, experimental terials, components, systems, structures. procedures, and theoretical and mathematical analy- and ses. Illustrated with photographs, drawings, and Handbooks. Recommended codes of engineering charts. and industrial practice (including safety codes) developed in cooperation with interested industries, Published in three sections, available separately: professional organizations, and regulatory bodies.

• Physics and Chemistry Special Publications. Proceedings of NBS conferences, bibliographies, annual reports, wall charts, Papers of interest 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 Series. eign, $11.75*. Standard Reference Data 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 and computer systems. Short numerical tables. product characteristics for both buyers and sellers. Issued quarterly. 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 Clearinghouse for Federal Scientific and The best single source of information concerning the Technical Information, operated by NBS, supplies Bureau's research, developmental, cooperative and unclassified information related to Government-gen- publication activities, this monthly publication is erated science and technology in defense, space, designed for the industry-oriented individual whose atomic energy, and other national programs. For daily work involves intimate contact with science and further information on Clearinghouse services, write: technology for engineers, chemists, physicists, research managers, product-development managers, and Clearinghouse company executives. Annual subscription: Domestic, U.S. Department of Commerce $3.00; foreign, $4.00*. Springfield, Virginia 22151

• Difference in price is due to extra cost of foreign BMllint;.

Order NBS publications from: Superintendent of Documents Government Printing Office Washington, D.C. 20402 U.S. DEPARTMENT OF COMMERCE WASHINGTON, D.C. 20230

OFFICIAL BUSINESS POSTAGE AND FEES PAID U.S. DEPARTMENT OF COMMERCE