The Commercialization of Management Systems, 1969–1983

Thomas J. Bergin American University

Thomas Haigh University of Wisconsin, Milwaukee

Database management systems were the most important commercial packages of the 1970s. The authors reconstruct their early history by examining the evolution of their capabilities and installed base. They also document early user experiences, including the sources from which potential users learned about these new tech- nologies, new roles such as the database administrator, and new concepts such as the data dictionary.

One approach to examining the early days of characteristics and exploring the changing commercial database management systems dynamics of the DBMS marketplace into the (DBMSs) is to study the evolution of the lead- early 1980s. This includes a brief discussion ing DBMS packages of the 1970s. Doing so, of the commercial threat posed by relational we can reconstruct the technical issues and database systems to established products and management decisions facing information efforts made to add relational capabilities to technology managers as they investigated traditional products such as IDMS. Our focus the new world of database software in the is on database management packages for early 1970s. This includes not only the IBM machines, which dominated the main- DBMS as a productivity aid for application frame market at that time. We rely here on but also the new jobs, such as contemporary market-analysis reports, partic- database administrator, created by the tech- ularly Datapro’s ‘‘Buyer’s Guide to Data Base nology and the managerial challenges of cre- Management Systems’’ starting in 1974.1 ating a shared and consistent database to be used across different departments. Indeed, Database beginnings during the 1970s, the DBMS existed both as Life in the data processing departments of a tangible technology (with considerable the 1960s revolved around applications strengths and pronounced weaknesses) and requiring the storage, updating, and retrieval as the symbol of a movement to raise the sta- of large volumes of data. The arrival of disk tus of computing within the managerial technology in the mid-1960s posed a huge world and establish the idea of data as a cor- challenge to application developers.3 Organi- porate resource. (See the ‘‘Data Management zations struggled to exploit the potential of Definitions’’ sidebar for a brief explanation random access data storage and interactive of related terminology.1) query. As historian Martin Campbell-Kelly noted, Corporate data processing budgets mush- finding good data on software adoptions and roomed in the 1960s, driven in large part by usage is difficult.2 Our research utilizes vari- the rapidly growing cost of maintaining sys- ous contemporary journal and newsletter tems and application software. Organizations articles, consulting reports, and advisory not wanting to continue to create and main- guides to document early DBMS user experi- tain large applications had two choices: hire a ences and the processes by which they eval- contractor to develop the software to the uated different packages. We also examine organization’s specifications or purchase a some of the major DBMS products of standard software package from an indepen- the 1970s, documenting their technical dent vendor (and in many cases, pay them

26 IEEE Annals of the History of Computing Published by the IEEE Computer Society 1058-6180/09/$26.00 2009 IEEE Data Management Definitions To frame our discussion on specific software pack- and centralize the data storage, the system’s principal ages, let’s review a pair of definitions from the period: intent is to perform such functions as information re- trieval, report generation, and inquiry for single DBMSs can be defined as those that are intended to applications.1 manage and maintain data in a non-redundant struc- ture for the purpose of being processed by multiple Data-management systems were also known as file- applications. A DBMS organizes data elements in management systems. Although we provide a short dis- some predefined structure and retains relationships cussion of file-management and report-generation soft- between different data elements within the database. ware, largely to clarify the continued success these A data management system, on the other hand, is in- products enjoyed throughout the 1970s, this article fo- tended primarily to permit access to and retrieval cuses on early mainframe database management soft- from pre-existing files (usually for a single applica- ware. We believe the history of data management tion). Although a data management system might systems is sufficiently important to warrant a separate provide the capability to minimize data redundancy investigation. to tailor the package to the buyer’s environ- its applications, and save money. Infosystems, ment). In the mid-1960s, however, the soft- one of the leading management-oriented ware services industry was in its infancy, data-processing publications, assured its read- and almost no commercial market existed ers in 1973 that database systems were like for standard software packages such as sort the aeronautical efforts of the Wright broth- utilities, report writers, and information- ers; although carefully planned early efforts retrieval packages. When vendors and user had ‘‘never developed much lift when ap- groups did create such software, it was plied to the practical realities of processing given away without additional charge. large files that had to be stored, indexed Two factors changed this state of affairs. and sorted with live data,’’ they were now On 7 April 1964, IBM announced the System/ poised to rise majestically into the air.9 The 360, a family of compatible computers and problem was that no one knew much about operating systems (OSs) that let users mi- them, even the organization that had grate to larger computers without throwing attempted to write generalized data access out their existing programs.4 With these, software.10 organizations relied heavily on standard Most organizations wanting to explore OSs provided by computer vendors. Then DBMSs needed training for their technical on 6 December 1968, IBM announced that staff and consultants to guide them. Initially, it was unbundling software products, systems there was not much help available, so con- engineering services, and training costs rather sulting firms and software vendors realized than supplying them free of charge to its cus- that educating potential clients was an im- tomers with the hardware lease or purchase.5 portant way to spread their message, sell Together these decisions provided the products, and build market . Technical foundation for the software products and ser- training, off-the-shelf software, and various vices industry’s growth. The creation of com- types of support services became part of the patible computer systems families gave growing . prospective software developers a large and Industry periodicals of the early 1970s growing market in which to sell their wares. contained numerous articles introducing the IBM’s unbundling meant that early software notion of DBMSs. Datamation and Computer- vendors would not have to compete with world printed tutorials, articles detailing free software from IBM.6 This environment early user experiences, and articles focusing not only motivated software entrepreneurs on specific software packages.11 Some brave to create software, it allowed the fledgling authors (and editors) published short case industry to flourish. By 1972, there were studies as well as articles comparing individ- 82 vendors offering 275 packages for the life ual software packages and introduced readers insurance industry.7 Indeed, Cincom (Total), to advanced concepts.12 We say ‘‘brave’’ be- (Mark IV), and CIM (Datacom/ cause vendors were constantly changing DB) began as software services firms.8 their offerings while working with existing DBMS software promised to reduce the customers to plan future improvements and organization’s hardware burden, speed up additions. EDP Analyzer brought a new

October–December 2009 27 History of DBMS

management and organizational viewpoint than rely on (out-of-date) monthly or quar- to the dialogue.13 Most users, however, terly hardcopy reports? Identifying these depended on their hardware vendors. Of always-changing requirements was not a triv- course, industry periodicals, consultants, ial process and required the attention of and reports were not the only ways for poten- senior personnel. tial users to learn about advances in database After defining the organization’s require- technologies and share practical experiences ments, the selection team had to compare with its implementation. Computer user the hardware and software needed to run groups, particularly SHARE and GUIDE, con- the various DBMS packages with their cur- tinued to play an important role, with rent hardware configuration and operating many sessions organized to discuss current system and estimate the cost of additional and forthcoming software. hardware, software, media, training, and per- As emerged as a distinct research sonnel. Although all vendors were happy to area, new specialized interest groups, jour- demonstrate their products, some vendors nals, and professional meetings were created. encouraged users to benchmark their prod- For example, the ACM created a Special Inter- ucts against their competitors, and this est Group on File Description & Translation allowed many potential users to test their (SIGFIDET), which from 1969 onward pub- applications using actual DBMS software. An- lished a newsletter and hosted a series of reg- other common strategy was to visit organiza- ular conferences. In 1976, ACM’s Transactions tionsthatwereusingthehardwareand on Database Systems provided the field with a software under consideration and ask ques- prestigious academic journal. SIGFIDET and tions of their personnel. its successor, SIGMOD (Management of An organization’s strategy was Data) were more for researchers and creators therefore to (1) identify and evaluate pres- of DBMSs than for their users, but database ent hardware and software capability; (2) ex- technology was also much discussed within aminemultipleDBMSpackagesand the companion SIGBDP (Business Data Pro- identify their hardware and software cessing) and its newsletter Data Base: Advan- requirements; (3) analyze the user’s present ces in Information Systems.OutsideACM,a and future processing requirements; successful annual series of international con- (4) identify additional hardware or software ferences on very large databases was required for each potential package; (5) talk launched in 1974.14 to other users, see demonstrations, and run benchmarks; and (6) make a decision on Making the DBMS selection which database management package to To assist beginning DBMS users, Datapro buy, lease, or use via a service bureau. provided a few caveats to guide the selection Once a decision had been made, the imple- of a DBMS: ‘‘define in detail your organiza- mentation process began. This included tion’s system and data requirements ... and (7) acquiring additional hardware and soft- don’t be surprised if this definition of require- ware, (8) technical training for employees, ments takes the efforts of two or more people (9) data conversion, (10) database system for up to six months.’’15 (Italics added to em- generation, (11) various levels of testing, phasize the report’s authors lack of under- (12) creation of linkages between applica- standing of the difficulty of these new tasks.) tions, and (13) system refinement and For organizations choosing a database package maintenance. To many potential DBMS for a single application, this might be a rela- users, this was a daunting scenario. In addi- tively simple task; for organizations choosing tion to costs, users were interested in the a DBMS as the centerpiece of multiple applica- system’s efficiency, the ease of installation, tions, this would be a massive undertaking. and the quality of the vendor support and To evaluate a vendor’s offerings, the user package documentation. So much for Data- had to carefully examine existing hardware pro’s six-month estimate. capabilities—the space available on direct- access storage devices, the CPU time available The 1974 marketplace to process the application, and so on. More By the early 1970s, Codasyl’s Data Base important than the technical side of the deci- Task Group’s recommendations on database sion was defining the needs of the applica- software architecture and preliminary stan- tion’s users and managers. What were the dards for data definition and manipulation applications’ unfulfilled needs? Did users languages had begun to appear. (See Thomas want to access data on an ad-hoc basis rather Haigh’s discussion of the Codasyl’s Data Base

28 IEEE Annals of the History of Computing Task Group’s work elsewhere in this issue.) were not based directly on the Codasyl This led observers to question the need for specifications. an industry-wide standard for DBMSs. In The 1974 Datapro Buyer’s Guide con- 1972, EDP Analyzer opined that ‘‘it is gener- tained data on six packages: Total, System ally conceded that it is premature to 2000, IMS, Adabas, IDMS, and DL/1. Datapro about a standard data base management fa- selected these packages because they were cility at this time,’’ but that since ‘‘there is ‘‘the ones that generate[d] the most interest nothing comparable to the DBTG proposals’’ from ... our subscribers, and [were] the most they should be used as the basis for further widely installed at [that] time.’’1 Table 1 development of DBMSs.16 shows the range of these packages’ operating In October 1974, EDP Analyzer revisited characteristics, requirements, and capabilities. Codasyl-type database systems, reviewing IBM had settled on a single architecture implementations by the Minneapolis Star for its 360 and 370 series machines, but this and Tribune Company (DMS 1100 on Univac did not mean that software selection was 1106), Xerox Corporation (EDMS on Sigma 7), simple. Users paid close attention to the dif- Abbott Laboratories (IDMS on IBM/370), and ferent packages’ system requirements, in Northern Illinois Gas Company (DM6700 on terms of processor demands, memory a Burroughs B6700).17 Codasyl-type systems requirements, languages supported for appli- were described that ‘‘contain many deviations cation development, and OS requirements. from the CODASYL specifications.’’ Analyzer For example, the minimum memory require- concluded that ‘‘CODASYL-type DBMS repre- mentswere3to30KbytesforTotal,160 sent a worthwhile, viable effort’’ (p. 13).17 Kbytes for Adabas, and 256 Kbytes for IMS The only such package available for IBM running under the MVT ! equipment was Cullinane’s IDMS, which Independent vendors generally supported a was identified as implementing ‘‘a subset of broader range of options than did IBM, and the April 1971 DBTG report.’’ B.F. Goodrich, this explains some of their success in the under the direction of DBTG member marketplace. With respect to host languages, Richard F. Schubert, implemented a subset IBM’sIMSandDL/1packagessupported of the 1971 DBTG specifications. In 1972 Cobol, PL/1, and assembler, whereas the and 1973, Goodrich received permission to other packages also supported Fortran. Only convert GE’s Integrated Data Store to IBM Total supported IBM’s RPG II. The situation equipment.18 Goodrich named the system was more complicated for OSs. In 1970, the Integrated Database Management System IBM’s IMS did not run under its DOS operat- (IDMS, p. 6).16 In 1973, the Cullinane Corpo- ing system. Potential users had a choice: ration acquired the rights to further develop upgrade to its full OS (which was time and market IDMS.17 consuming and expensive), wait until Although IDMS did not come to market December 1973 when DL/1 became available until May 1973, the differences in the pack- under DOS/VS, or choose one of the packages ages on the market in the early 1970s high- offered by independent software vendors. lighted the need for a standard approach to Many organizations chose the latter. DBMSs: An organization’s data-processing re- quirements also influenced system selec- It is becoming increasingly apparent that standardization is the best way to ward off tion. For example, some organizations had the trauma of conversion and major hardware a need for frequent and rapid inquiries, announcements. Just as COBOL has become which meant that an inverted file system the standard for busi- would be beneficial. The availability of ness applications, virtually machine- and a report generator, data dictionary, or manufacturer-independent, many experts be- checkpoint/restart facility could influence lieve that the CODASYL Data Base Task Group the selection based on user capabilities or (DBTG) language specifications will become goals. More sophisticated users might the data base standard of tomorrow—at least want to use the package in an online envi- in basic principles.19 ronment and select a package that could Later, other vendors took notice of the be interfaced to a communications package commonality between the DBTG recom- such as CICS, Environ/1, or Intercomm. For mendations and their native implementa- other users, the availability of a report gen- tions and marketed their packages as erator such as Socrates or IDMS-Culprit ‘‘DBTG-like.’’20 The other leading DBMS might seal the deal. Although numerous packages of the mid-1970s for IBM systems software vendors offered report-generator

October–December 2009 29 History of DBMS

Table 1. Characteristics of early DBMS packages.

Total System 2000 IMS Adabas IDMS DL/1 DOS/VS Vendor Cincom MRI Systems IBM Software AG Cullinane IBM Systems First installation Early 1969 June 1970 March 1971 March 1971 May 1973 Dec 1973 Current users 750+ 70+ 400+ (est.) 70+ 30+ 100+ (est.) CPU requirements S/360/370 S/360/370 40 S/360/370 40 S/360/370 S/360/370 S/360/370 (IBM) & 145 & 145 125, 135, 145 Operating DOS, DOS/ OS, OS/VS OS/MFT, VS1, VS2 DOS, OS, DOS, DOS/VS, DOS/VS system (IBM) VS, OS, VS1, OS/VS OS, VS1, VS2 VS2, IBM System/3 Minimum 3–30 156 128 (MFT)/256 160 30 98 memory (IBM, (MVT) in Kbytes) Application Cobol, RPG Cobol, PL/1, Cobol, PL/1, Cobol, PL/1, ANS Cobol, Cobol, PL/1, language II, Fortran, Fortran, Assembler Assembler Fortran, Fortran, PL/1, Assembler PL/1, Assembler Assembler Assembler Network Partially inverted Hierarchical, Partially Hierarchical Hierarchical, hierarchical sequential, and inverted network or network sequential, and hierarchical hierarchical direct (tree) direct (tree) Checkpoint/ Yes Yes Yes Yes No No restart Online inquiry No Inquiry and IQF, GIS/VS ADASCRIPT No No update language included Report SOCRATES Self-contained GIS/VS, GIS-2 Easytrieve from IDMS-CULPRIT No generator Pansophic Systems Data dictionary Under Procedural IMS dictionary Report utility Via user defined No development language pre- system had data data definitions compiler dictionary and utility functions programs Telecom ENVIRON/1, TP2000, CICS, Included, also CICS TSO, CICS, GCI CICS interface CICS, Intercomm, TCAM Intercomm Intercomm User satisfaction 3.6 3.3 2.8 3 3.6 2.4 Response number 46 8 24 5 7 5 Price $29,500 or $40,000 or lease DB: $616/mo, $120,000 $37,500 or $330/mo lease DC: $700/mo lease

packages, there was a big advantage in get- have excerpted enough of this material ting additional software from the DBMS to provide an understanding of these early vendor rather than having to deal with packages and to provide a basis for a later dis- multiple individual vendors—especially cussion on the evolution of the DBMS when interface problems were encountered. marketplace.21

Datapro evaluations As compared to IBM’s IMS, TOTAL offers a In addition to the tables on the individual more simplified and streamlined approach to packages’ technical characteristics, the Buyer’s data base management, providing a more satis- Guide on early mainframe DBMSs contained factory solution for the majority of users ... summary comments on the packages. We and using far fewer resources in terms of

30 IEEE Annals of the History of Computing personnel and hardware ... TOTAL provides a management involvement in the database simple and straightforward set of facilities for analysis and design stages. User’s reactions organizing a data base and for manipulating to IMS were more favorable than to DL/1.25 the data within an application program ... The Buyer’s Guide described Adabas as (and) provides a complete logging, backup, follows: and recovery capability.22 While ADABAS ranks as a top-notch data base As Table 1 shows, Total supported the wid- management system, few of its features offer est variety of mainframes and OSs, and Data- unique capabilities ... Rather, the system has pro concluded that, ‘‘TOTAL is one of the been put together with such thoroughness most effective high powered software systems and skill, that its most impressive features are in use today and it has been successful in dis- not directly evident to the end user and appear placing IBM’s IMS in numerous installa- in the form of reduced system overhead.26 tions.’’22 The primary reasons why users selected Total were the low memory require- Some of this was the direct result of a data- ments, the ease of use and implementation, compression algorithm that typically created and its high level of performance.22 a database that was 30% smaller than other The main feature of System 2000 was its packages. Adabas also separated the logical fast response times resulting from its partially relationships from the physical relationships, inverted file structure. The user determined which allowed more general network struc- the amount of the inversion (i.e., the number tures and provided more flexibility. Adabas of fields that could be used to order a search), was designed to work with extremely large and a well-structured database was critical for databases, accommodating up to 4.2 billion efficient usage. The more inverted fields, the records. Although designed to handle longer the processing time, and the more 255 files (or databases), a ‘‘minor internal storage required to store data and pointers. modification’’ allowed up to 65,000 files to ‘‘The system is well suited for fully integrated be used. Although Adabas did not offer a data bases as well as simple, more specific communications monitor, the Adascript applications on subsets of data ... Many of query capability used English-like statements SYSTEM 2000’s features are specifically and the Adawriter report writer was released designed to meet information processing in early 1975. Users generally liked Adabas. requirements in business and scientific com- Early DBMS users munities.’’23 For example, System 2000 had In a 1975 Datamation article, Robert M. a multithreading feature that let multiple Curtice examined ‘‘The Outlook for Data users access data in a multiprogramming en- Base Management.’’27 Curtice stated that, vironment. ‘‘System 2000 users think highly ‘‘Data base management packages have of the product.’’23 been a somewhat surprising success for the IMS was also one of the most flexible independent software houses, with total packages on the market and met a variety installations perhaps second only to their of user requirements. cousins, file management packages.’’ He also recognized that each of the existing IBM’s Information Management System (IMS) packages had developed a sizable user base is one of the most controversial and talked- about software systems on the market today. and believed that ‘‘it would take until 1980 It is also one of the most misunderstood and or beyond to convert all major applications misused of all systems, and has developed to an integrated data base ... most (users) the reputation of being everything from the will stick with their original DBMS software most sophisticated and technologically cur- until this conversion is all but complete, rent system on the market to Godzilla.24 rather than switch in midstream.’’ Adoption of database management soft- IMS was particularly useful to organizations ware proved to be a boon to application undertaking a centralized approach to the programmers. In administrative applica- issue of data management. In November tions of the kind traditionally carried out 1973, IBM released DL/1 for users running by corporate data processing departments, theDOS/VSoperatingsystem.Despiteits an enormous amount of time complexity, Datapro believed that IMS had was taken up doing the things that the potential to support management infor- DBMSs were supposed to automate. They mation system development because of its made programs cheaper to develop and centralized database and the need for much easier to maintain, and they

October–December 2009 31 History of DBMS

facilitated the integration of different busi- of the computer’s memory and used ness tasks. 10 times more processor capacity than the Yet DBMS systems were not panaceas and tape-based version. It had moved cautiously incurred significant costs. As early as 1973, a into online operation; while records were report by two Booz, Allen, & Hamilton con- retrieved using terminals, all updates were sultants suggested that the software and queued and applied at night while the system hardware needed remained immature, little was off-line in the belief that this ‘‘greatly experience existed in its use, and the general- reduces the possibility of a catastrophic loss ized features offered by DBMS brought a of data.’’32 heftyperformancepenaltyandmightwell According to a 1975 survey of large indus- trigger the purchase of more memory or a trial firms, about one third were using some new processor unit.28 Most of the true costs kind of advanced file or DBMS.33 Of that were hidden, particularly the staff require- third, approximately half were using systems ments. As they put it, intended for direct ad-hoc querying by non- programmers, such as Mark IV, and half Some DBMS’s are as complex as the operating were using DBMSs designed to integrate system which services them. Also, (the techni- with the conventional programming lan- cal systems) group must continuously apply guages such as Cobol. Only about a quarter and test new program fixes and new features of the systems were used primarily for online to keep the system ‘alive and well.’ It is not uncommon to see a small systems program- access, and only two firms claimed to have ming team double or even triple as the result implemented a database for the entire firm, of a DBMS.28 though most reported using it for multiple areas of the business. This was very slow to Later reports suggest that these problems change. Five years later, a survey of manage- continued for several years and that many ment information systems in 32 large corpo- firms installed DBMS packages because of a rations found that most of these companies technical bandwagon effect rather than a had now installed a DBMS.34 Yet when the careful and informed evaluation.29 Although researchers looked at the actual use of these the idea of executives directly accessing com- systems they found that, ‘‘The users surveyed puter data was a common part of database were only beginning to develop DBMS appli- hype during this era, no surveys of the early cations. ...This is possibly because of the diffi- 1970s were able to find any firms where man- culties involved in developing and controlling agers or analysts directly used the database.30 such activities.’’34 Even among firms acquiring the most In 1975, Elizabeth Fong and two other advanced DBMS packages, online use was employees of the Institute for Computer limited and managerial applications rare. Sciences and Technology assisted the Federal Two examples of firms using commercially Aviation Administration in assessing the supplied DBMSs in the early 1970s provide capabilities of six DBMSs.35 The purpose us some insight into these early days of was to provide an ‘‘objective overview’’ of DBMS adoption and usage. McDonnell the features of these packages. Their report Douglas, using IBM’s IMS system, claimed explained the advantages of using a database to have created a centralized database con- management package and examined the use taining all the information previously stored of Adabas, IMS, Inquire, Model 204, System in 264 files covering spare parts, production 2000, and Total in 11 federal organizations scheduling, bill of materials handling, and of varying size, technical capabilities, and inventory management. This made it much user needs. The methodology included on- easier to change the 95existingprograms site interviews with users and meetings with that relied on these files, to set up automatic vendors. cross references between different records, Their report documents that federal and it hoped, to move toward online opera- agencies and departments, along with private tion in the future.31 organizations, were successfully introducing A much smaller firm processed accounts DBMSs. From large agencies such as the receivables for doctors. It used a Codasyl- Naval Material Command’s linking of 40 sep- style DBMS on its Xerox computer to lower arate databases for 40 applications using IMS, its daily processing times, shorten updates, to the Department of Agriculture’s Statistical and design its new programs more rapidly. Reporting Service using System 2000 on a The results pleased the firm, despite the time-sharing basis to search crop and live- factthattheDBMSconsumedalargepart stock data, federal agencies were taking

32 IEEE Annals of the History of Computing advantage of the growing offerings of data- an abstraction into a solid organizational base vendors. At this early time, the Depart- power base. This dream was enshrined in a ment of Agriculture had already made new figure, the database administrator. System 2000 a departmental standard.36 According to one of the earliest descriptions, Ian . Palmer provided documented case the DBA must ‘‘at once be technically quali- studies from the commercial sector, includ- fied, if not inventive ... he must encourage ing one which was tailor-made for a facilities the users to work with him willingly and management company.37 In contrast to the yet he will be forced to rule against their Fong report, which contained systems used pet projects; he must represent both manage- by small organizations, Palmer’s case studies ment and the users simultaneously; he must seem to concentrate on ‘‘major’’ corporations be all things to all people at all times.’’39 and show an abundance of firms in the finan- This role did ‘‘not exist as a formally estab- cial sector, such as banks and insurance com- lished function in today’s business,’’ but its panies. Shorter case studies were published in emergence was considered imminent.39 Datamation, Computerworld, EDP Analyzer, Harvard computer management authority and other publications during this period. Richard Nolan was still bolder; he believed Palmer’s publisher, QED, also published that the DBA would be responsible for Data Base Management Systems by Leo J. ‘‘data as a resource ... much broader than Cohen, which became the bible for organiza- just computer-readable data,’’ once the tions seeking to take advantage of DBMS.38 ‘‘data resource function [had been] carved Cohen provided detailed strategies for evalu- out of the general management function.’’40 ating and selecting a DBMS and technical Another observer wrote that the DBA needed summaries of the leading packages. to be ‘‘something of a superstar.’’41 Discussion of the DBA makes apparent Data administration a rift between managerially oriented uto- To the computer technician, a DBMS rep- pians and programmer-oriented pragma- resented an improved mechanism to store tists. Richard F. Schubert, who had overseen and retrieve data; to the application user, the remarkably ambitious B.F. Goodrich in- the DBMS represented a new way of using house DBMS development project that gave the information collected as part of auto- rise to IDMS, noted simply: mated applications. One outgrowth of the DBMS environment was an understanding Data base administration is accomplished by of the actual cost of data storage and, most one or more technical experts who are knowl- importantly, the cost of storing redundant edgeable in data base design and creation, op- data. If an organization has an accounts re- eration of the data base management system, and the use of one or more data manipulation ceivable file with customer name, address, languages. The data base administrator must and contact as well as an accounts payable also be capable of working well with systems application with essentially the same infor- analysts, programmers, and computer opera- mation, a DBMS could let the user store the tions personnel.42 common data once and share it between the applications. If other applications use It seems likely that this reflected practice in the same or similar data, then the efficiencies those firms actually using the technology multiply and the case for a shared database rather than just talking about it. Certainly, environment becomes overwhelming. How- by the time DBMS technology became ubiqui- ever, the potential for data sharing raises po- tous in the 1980s, the DBA was a technical spe- litical problems: cialist rather than an information executive. One IBM advocate of the ‘‘data dictionary Who owns the data? approach’’ likened data to money: ‘‘Once What form should the data be stored in? management realizes the relationship of re- Who should be allowed to access the data? liable data to corporate well-being, they will treat their data with the same care A whole new set of problems arises when used to handle their cash.’’43 Nolan made a multiple organizations need to modify and similar pitch in his book Managing the Data maintain shared data. Resource Function, the title of which sug- These quibbles did not stop hopeful ac- gested that information, like people and counts of the database as a technological money, is a vital resource of business and marvel that would finally centralize and con- therefore deserved similar managerial atten- trol information of all kinds, turning it from tion.40 Indeed, Nolan’s claims that the DBA

October–December 2009 33 History of DBMS

would be charged with overall responsibility Eastern Airline representative, although for all corporate information, using com- most of its 200 users were ‘‘a complete new puter technology where appropriate but ul- breed of coders ... non-programmers (with) timately claiming managerial rather than little or no data processing background,’’ technical authority, directly prefigure those attempts to train them in information re- made more generally for the new chief infor- trieval techniques without giving an under- mation officer (CIO) position in the 1980s.44 standing of what went on in the computer’s ‘‘mysterious black box’’ had failed. Contrary 1970s data-management products to their expectations, ‘‘The only users able The rapid expansion of DBMS use did not to move into extended capabilities with any kill demand for simpler file-management and degree of success were those with some data report-generation products. Even products processing background.’’46 designed explicitly for use by non-specialists IBM’s RPG family of report generators found their main markets to be among enjoyed similar popularity. Although there data-processing specialists. Because they cost was a substantial demand for products that less and could run on more modest hard- would let non-specialists produce computer- ware, file-management systems remained ized reports without programmer assistance, more widely used than full DBMSs. A 1975 the leading DBMSs did not do a good job survey found that 41% of DBMS-using firms of meeting this need. In fact, DBMSs actually reported that information could only be grew the market for report generators. One of retrieved with the aid of a programmer.33 Un- the most successful software products of the like DBMSs used primarily by application 1970s, Pansophic’s Easytrieve, was an easy programmers to manipulate data on disk, to use report-generation system designed to file-management systems were primarily extract information from files and databases. used (in 77% of firms) with files stored on Easytrieve thrived in competition with more tape rather than on disk. These systems still complex DBMS and file-management soft- worked with individual files rather than ware, and many firms purchased the optional vast integrated databases—indeed, 55% of modules needed to use it in conjunction with their users had not even begun to integrate the most powerful DBMSs.47 files to remove redundant information. During the 1970s, the Mark IV file- Market developments: 1974–1979 management and report-generation system The DBMS market too was growing became the most successful single product strongly. By 1974, more than 1,400 installa- in the admittedly short history of the indus- tions were running one of the six database try; it was the first to reach the milestones packages listed in Table 1. The growing mar- of $1 million, $10 million, and $100 million ket soon attracted additional companies and in cumulative sales. Compared to the Coda- additional packages. The Buyer’s Guides for syl specifications, its capabilities were mod- 1976, 1977, and 1979 included 13, 15, and est. Its initial appeal was straightforward: 18 packages, respectively. Despite this prolif- first it was highly efficient in batch operation, eration, the market in 1974 was dominated and second it had been designed for use by by two packages: Cincom’s Total with 750 nonprogrammers. Requests for data were installations (53% market share) and IBM’s entered onto one of four simple paper IMS with 400 users (28% market share). forms, and then keypunched for later pro- Cincom, as the first vendor, enjoyed a cessing. But even Mark IV found its main au- head start on the rest of the field having dience among programmers. As time went the earliest installations and the largest num- by, Mark IV development focused more on ber of users. However, more important was the needs of full-time programmers who Cincom management’s decision to adapt used it as a foundation for constructing com- Total to a wide range of hardware platforms. plex application programs. An official com- Indeed, Total was available for the most pany history credited this process to the mainframe configurations: IBM System/360s influx of data processing specialists into its and 370s; 200s and 2000s; Univac IV League user group, which ensured that 1106s, 1108s, and 1110s; CDC 6000s; and their opinions ‘‘overwhelmed the voices of NCR Century mainframes. Most of the the non-programming end users’’ in the com- other independent vendors also ported their pany’s planning.45 This group’s proceedings software to other mainframes, while IBM’s suggest that non-specialists found advanced IMS only ran on System/360 Model 40 and work harder than expected. According to an larger and System/370 model 145 and larger.

34 IEEE Annals of the History of Computing Table 2. Installed base of DBMS products, 1974–1979.

First Users Market Users Market Users Market Users Market IBM systems installation in 1974 (%) in 1976 (%) in 1977 (%) in 1979 (%) Total Early 1969 750 53 1,000 48 2,000 49 2,500 43 System 2000 June 1970 70 5 100 5 150 4 700 12 IMS Mar. 1971 400 28 500 24 1,000 25 1,200 21 Adabas Mar. 1971 70 5 150 7 200 5 300 5 IDMS May 1973 30 2 120 6 300 7 400 7 DL/1 Nov. 1973 100 7 150 7 300 7 600 10 Datacom/DB 1971 * 30 1 60 1 100 2 Model 204 1973 * 22 1 41 1 41 1 Subtotal** 1,420 2,072 4,051 5,841

Non-IBM systems Codasyl Honeywell 500 52 525 38 660 33 Sperry Univac 300 31 500 37 500 25 Other 10 1 120 10 230 11 Subtotal** 810 85 1,145 84 1,390 69 Non-Codasyl 145 15 220 16 635 31 Subtotal** 955 1365 2025

* Not included in the 1974 Datapro Buyer’s Guide. ** ‘‘Subtotal’’ is used to avoid confusion with the ‘‘Total’’ DBMS.

Indeed, IBM initially only supported the OS to have been fastest around 1977, with the and OS/VS operating systems while Cincom’s estimated number of installations almost Total could operate under IBM’s DOS and doubling from the previous year. This rapid DOS/VS. growth indicates that the idea of using Table 2 shows the data from the Datapro DBMSs was gaining widespread acceptance Buyer’s Guide for 1974, 1976, 1977, and across the industry, perhaps due to favorable 1979 and is ordered by earliest installation articles in the technical press and a growing toshowthegrowthininstallationswith cadre of technical staff with DBMS experi- respect to the package’s time in the ence.49 The importance of DBMS software marketplace. to the commercial software market was Table 2 reveals that the relative positions hard to miss. The 1979 Buyer’s Guide stated, of the vendors in 1974 continued for the ‘‘data management software of all types will rest of the decade, even as the number of account for over half of the proprietary soft- organizations adopting DBMS technology ware market revenues in 1983 (estimated to rose sharply. By 1979, the vendors had a be about $2.3 billion). A sizable portion of total of 5,841 installations using their prod- this amount will be spent specifically for ucts. Cincom continued to lead all vendors data base management systems.’’50 in the number of hardware environments Although none of the other packages supported, with a 43% market share.48 At managed to challenge IMS for the number this point, Cincom had approximately 700 two spot, they did increase their market employees, more than any of its independent share during the late 1970s as well as dramat- competitors, and a growing number of offices ically enlarge their customer base. IDMS and overseas. IMS remained a strong second System 2000 both flourished, with System place, though its market share slipped to 2000’s installed base more than tripling be- 21% (even as its number of installations tween 1977 and 1979, resulting in a 12% tripled). We also included data on non-IBM market share. IDMS grew rapidly in the systems users to provide a more comprehen- mid-1970s to finish the decade with 7% of sive picture of the overall DBMS marketplace the market while Adabas retained a stable at this time. niche with a 5% market share. Each new report showed substantial Analysts often rely on guesswork and leaks growth in the DBMS market. Growth seems for many of the details in their reports, but in

October–December 2009 35 History of DBMS

Table 3. Database installations and revenues by vendor.51

Market Employees share Revenues Anticipated involved in DBMS Vendor Package Sites* (%) (millions $)* growth rate* sales/service** IBM IMS 1,300 21 42 High — DL/1 750 12 High — Cincom Total 2,700 44 30 Low 700 Cullinane/CAI IDMS 500 8 13.9 Very High — Software AG Adabas 375 6 8 High 200 Infodata Inquire 100 2 3.5 Low 75 MRI/ System 2000 280 5 10 Medium to high 275 CIM/ADR Datacom/DB 150 2 1.1 Medium 60–70 CCA Model 204 50 1 1.5 Low 15 Total 6,205 110

* Source: Frost and Sullivan report. Site data and anticipated growth rate are from Table 1, page 20; revenue data is from Table 71, page 203. ** Number of employees are from individual reports.

this case, we are fortunate to have a second and that eventually ‘‘strides in hardware source for comparison. In December 1979, price-performance will eventually make rela- Frost and Sullivan published a report that tional data bases practical and more attrac- contained financial data on DBMS vendors tive.51 Frost and Sullivan were wrong on the and made forecasts for the 1980s.51 This first prediction and right on the second. report mirrored the Datapro findings, as Table 3 shows. Cincom led in market share Market developments: 1980–1983 with 2,700 installations. This report also pro- Datapro opened its 1980 Buyer’s Guide by vided some interesting estimates of other saying, aspects of the business. According to Frost and Sullivan, IBM led in terms of database [E]xperts in the field of data base manage- revenues (with $42 million), despite its ment (predict) the growth of the industry in smaller installed base. Because DBMSs were the 1980’s will be nothing short of phenome- still used by less than 25% of the IBM market, nal in regards to DBMS acquisition ...Up to a they predicted that mainframe DBMS would few years ago, only the wealthiest and most continue to grow rapidly. sophisticated computer user could or would Although the main DBMS market was for spend any time evaluating DBMS systems. Now there is a system for almost any size packages usable on IBM mainframes, each computer and data processing environment. of IBM’s major competitors was served by The application of DBMS technology for at least one package. Datapro included anal- use with minicomputers and even microproc- ysis of systems for these other platforms, essor systems has really caught on in the primarily self-described Codasyl-type sys- marketplace.52 tems. Products for Honeywell and Sperry Univac mainframes were the most impor- Many new companies entered the market in tant, accounting for 83% of the market for the early 1980s. International Resource De- DBMSs to run on non-IBM platforms in velopment completed a report in 1986 that 1976. On these platforms, non-Codasyl provided an overview of the entire DBMSs products held only a 15% market share. By market for 1985 stating that nearly 200 US- 1979, however, that share had almost based firms sold DBMS software, tools, and doubled to 31% of the non-IBM platform related products.53 market, so the Codasyl influence was in de- Like the DBMS market itself, the product cline in what had been its strongest niche. lines of the leading suppliers expanded and This was not fully appreciated at the time; offered an array of options and niche products. Frost and Sullivan’s 1979 report suggested As early as 1974, some DBMS vendors were of- that the future will see ‘‘greater adherence fering add-ons such as report generator pack- to the CODASYL data base specification’’ ages, data dictionaries, and communications

36 IEEE Annals of the History of Computing Table 4. Installed base of DBMS products, 1979–1983.

Amount Amount Market Amount Market Amount Market IBM systems in 1979 in 1980 (%) in 1982 (%) in 1983 (%) Total 2,500 3,000 40 4,000 59 5,750 65 System 2000 700 700 9 700 10 700 8 IMS 1,200 1,500 20 * * Adabas 300 550 7 550 8 1,000 11 IDMS 400 460 6 1,100 16 1,100 12 DL/1 600 1,000 13 * * Datacom/DB 100 250 3 350 5 330 4 Model 204 41 41 1 125 2 ** Subtotal† 5,841 7,501 6,825 8,880

Non-IBM systems Codasyl 1,330 1,430 78% 1,760 42 non-Codasyl 735 400 22% 2,450 58 Subtotal† 2,065 1,830 4,210

* Data not provided by IBM. ** No separate report included. y ‘‘Subtotal’’ is used to avoid confusion with the ‘‘Total’’ DBMS.

interfaces. However, as organizations realized consisted of separately priced components, that they could create ad-hoc reports without including Datadictionary, Dataquery, Data- writing programs and could run online inqui- design ADR/D-Net, and ADR/Ideal for applica- ries against the ‘‘corporate database,’’ the de- tions development.56 Cullinane offered it’s mand for these capabilities exploded and IDMS Data Dictionary and the Culprit report spurred the development of additional add- writer, which was later renamed EDP/Auditor. on packages and the incorporation of new Cullinane also interfaced IDMS to applica- capabilities. One result was that by the late tions software from other vendors. Not to be 1970s most vendors adopted a ‘‘total systems’’ left out, MRI Systems integrated MSA’s finan- concept, in which ‘‘a modular set of programs cial application packages and SAS Institute’s were combined to handle inquiry/response, SAS statistical analysis and graphics package accounting and billing, report writing, and with System 2000. Computer Corporation of other functions in a single integrated sys- America did something similar with Model tem.’’54 Although Datapro opined that this 204, including offering an interface to SAS.56 was ‘‘nothing more than a clever packaging Starting at the end of the 1970s, the DBMS approach to marketing a DBMS,’’ vendors con- industry grew in synch with the computer tinued to push this new modular approach. For hardware market—which now included example, Cullinane offered the following op- minicomputers. Table 4 shows how this tional packages: IDMS-DC (an integrated tele- growth continued during the early 1980s. communications monitor), Culprit (a report Datapro estimates suggest that Total was writer), Interact (an interactive program devel- stronger than ever in the early 1980s, increas- opment facility), On-Line English (a query fa- ing its market share to approximately 65% in cility), and an Integrated Data Dictionary. 1983. Adabas also strengthened its position, The trend accelerated in the 1980s. Cin- while System 2000 seemed to plateau. com announced that ‘‘TOTAL had evolved Unfortunately, IBM refused to provide instal- into a much more sophisticated system: the lation data to Datapro Research Corporation TOTAL Information System (TIS).’’55 The in 1982, and there was no separate report first TIS installation was in 1979 when Total for Model 204. This lack of data limits the was installed at more than 4,000 mainframe value of the data in Table 4 and skews the and 1,200 minicomputer sites; the 1982 market shares shown. Buyer’s Guide shows TIS with 30 installa- In other developments, Cullinane changed tions.57 ’s Datacom/DB its name to in 1983 and was the first

October–December 2009 37 History of DBMS

independent DBMS vendor to make the Acceptance of the relational model brought Datamation 100 list of the leading US data forth a number of database packages, among processing companies—at number 94 in them Oracle and . A full discussion of 1983 with revenues of $108 million.57 Ste- relational packages falls outside this article’s phen T. McClellan attributed Cullinet’s suc- scope (though a future Annals special issue cess to a superior marketing organization is planned), but we cannot consider the and referred to John Cullinane ‘‘as a master market for conventional hierarchical and marketer,’’ who convinced McClellan that network systems in the 1980s without Cullinet will be the first billion dollar soft- acknowledging the influence of the sudden ware company.58 vogue for relational technology. Existing In 1984, Cullinet rose to number 77 on vendors had three basic options when faced the Datamation 100 list based on revenues with the flood of favorable coverage of rela- of $143 million.59 Three other independent tional database technology: ignore it, add DBMS vendors also made the l984 list of lead- ‘‘relational’’ features on top of their existing ing US companies: Applied Data Research at products, or develop entirely new packages. number 83 with $128.2 million, Software IBM and Cincom developed new relational AG at number 91 with $115 million, and products, DB2 and Supra, respectively. Data- squeaked in at number 99 pro’s reports show that IBM’s relational offer- with revenues of $89 million.60 Intel (which ingsgrewrapidlyinpopularityduringthe acquired System 2000 in 1979) was at num- early 1980s and have contributed to the de- ber61withrevenuesof$180million. cline of the navigational packages. DB2 started with a single installation in 1983 Enter the relational model and then exploded to 109 sites in 1986. Relational DBMSs became an important SQL/DS, a similar product for the VM main- commercial force during the 1980s. Origi- frame, added 305 new sites during this pe- nated in the 1970s, the relational model grad- riod, for a total of 365 installations. ually gained acceptance among database Other vendors responded by developing researchers.61 The relational model was far new relational capabilities for their existing more conceptually elegant and flexible than packages, though fundamental differences be- the Codasyl endorsed, which tween relational and navigational approaches proved both restrictive (because relationships made this problematic. Cullinet offered must be specified when the database is IDMS/R, and Model 204 added some rela- designed) and insufficiently abstracted from tional features. Datacom/DB also acquired the physical storage of data (programmers relational features; it was sold as ‘‘relational- were still forced to write code to navigate like’’ as early as 1979 and, by 1982, as ‘‘fully explicitly from one record to another when relational’’ although Datapro characterized it working with linked data). Because the rela- as ‘‘an inverted file structure with relational tional model shifted the responsibility of characteristics.’’64 Computer Corporation of specifying relationships between tables from America marketed Model 204 as relational the person designing them to the person in 1979 as well.65 These hybrid products querying them, it permitted tables to be were initially more successful commercially joined in different ways for different pur- than the truly relational systems. By 1982, poses. This turned out to be necessary (if in- there were 3,135 installations claiming to be sufficient) for establishing large, general- relational or relational-like, an 81% growth purpose databases shared between different over 1980. But the same year Datapro con- departments and computer systems. The rela- cluded that ‘‘DBMSs that are truly relational tional model has also been praised for its non- have yet to proliferate (although) ... Rela- procedural nature—further separating the tional Software’s Oracle and Relational Tech- user from the physical storage mechanisms nology’s Ingres have been on the market involved.62 This simplified programming since 1979 and 1976, respectively.’’66 and insulated application code from changes Instead of claiming (sometimes on shaky in the database structure. Accepting a Turing evidence) to be Codasyl based, vendors now award in 1981 for his development of the promoted their products (sometimes on relational , Edgar F. (Ted) even shakier evidence) as being relational. Codd suggested that the Codasyl network Even in the DBMS market for non-IBM com- model had forced the programmer to become patible mainframes, previously dominated by too much of a navigator, at too low a logical Codasyl packages, the share of the installed level.63 base held by packages that claimed to be

38 IEEE Annals of the History of Computing Codasyl-like dropped from 78% in 1980 to interface rather than directly manipulating 42% in 1982. Controversy raged, as relational files. These early packages laid a vital founda- database experts, most vocally Ted Codd and tion for later developments and the success C.J. Date, sought to debunk the claims of tra- of the relational systems that thrive today. ditional systems, such as IDMS, to have somehow become relational. In the end, cus- Acknowledgments tomers with a genuine need for relational capabilities adopted truly relational systems, This article was written during Thomas but the performance advantages of network Haigh’s time as a fellow of the Center for and hierarchical systems for well-optimized 21stCenturyStudiesattheUniversityof transaction processing application ensured Wisconsin, Milwaukee, and as a participant that they retained a significant niche within in the European Science Foundation ‘‘Soft- the mainframe DBMS market. ware for Europe’’ project. Portions of this ar- ticle are revised from material in an earlier- published conference paper by Thomas Epilogue 68 Of the seven main early vendors of DBMS Haigh. software for IBM mainframes we have dis- cussed here, three are still in the software References and notes marketplace: IBM, Cincom, and Software 1. Datapro Research, ‘‘A Datapro Feature Report: AG. Computer Corporation of America offers A Buyer’s Guide to Data Base Management Sys- Model 204 and System 2000 is now offered tems (ADABAS, IDMS, IMS, DL/1 DOS/VS, Sys- by SAS. IDMS and Datacom/DB are sold by tem 2000, TOTAL),’’ report 70E-010-61a, 1974. Computer Associates. The individual parts of the guide have separate Hierarchical and network systems remain a report numbers, and pages are indicated by significant presence and are still used by many lowercase letters. large organizations, such as airlines, tele- 2. M. Campbell-Kelly, ‘‘Development and Struc- phone companies, and financial firms for ture of the International Software Industry, mainframe systems processing high transac- 1950-1990,’’ Business and Economic History, tion volumes. These products still bring in sig- vol. 24, no. 2, 1995, p. 78. nificant revenues, and programmers continue 3. See also M. Campbell-Kelly, ‘‘Origins of the to develop and maintain applications built on Software Products Industry: 1965–1970’’ and them. Indeed, 12 users of Cincom’s Total paid ‘‘The Shaping of the Software Products Indus- license fees in 2008 and 10 others were using try, the 1970s,’’ From Airline Reservations to Total in compatibility mode.67 However, since Sonic the Hedgehog, MIT Press, 2004. thelate1980s,themainstreamDBMSmarket 4. E.W. Pugh, L.R. Johnson, and J.H. Palmer, IBM’s has shifted unmistakably toward relational 360 and 370 Systems, MIT Press, 1991, p. 167. systems. This transition accompanied an ar- 5. B. Grad, ‘‘A Personal Recollection: IBM’s Unbun- chitectural shift away from mainframes and dling of Software and Services,’’ IEEE Annals, toward minicomputers, Unix systems, and vol. 24, no. 1, 2002, pp. 64–71. eventually inexpensive commodity servers as 6. Although IBM only reduced its prices by a small the most common RDBMS platforms. The fraction and IBM software was ‘‘inexpensive’’ leading commercial DBMSs during the past compared to software offered by independent decade have been Oracle, IBM’s DB2, and software vendors, it was no longer free. SQL Server. 7. M. Campbell-Kelly and W. Aspray, Computer: A Although these first-generation commer- History of the Information Machine, Basic Books, cial DBMS products have been eclipsed by 1996, p. 204. newer technologies, they demand historical 8. According to Walter Bauer, software package attention for two main reasons. First, they sales in 1970 amounted to $70 million, while were the mainstays of the independent pack- the comparable figure for software contracting aged software business during the 1970s and was $650 million. M. Campbell-Kelly, ‘‘Develop- did more than any other class of product to ment and Structure of the International Software prove the viability of software product com- Industry, 1950–1990,’’ Business and Economic panies. Second, they established the DBMS History, vol. 24, no. 2, 1995, reference 1, p. 87. as the default foundation for administrative 9. B.W. Romberg, ‘‘Data Bases: There Really Is a application software. A generation of pro- Better Way to Manage Your Files,’’ Infosystems, grammers, analysts, and managers learned a vol. 20, no. 5, May 1973, pp. 56–58. new style of application development in 10. For an excellent overview of the early efforts, see which data was manipulated via a DBMS J.P. Fry and E.H. Sibley, ‘‘Evolution of Data-Base

October–December 2009 39 History of DBMS

Management Systems,’’ ACM Computing Sur- 21. This section does not include IDMS, which we veys, vol. 8, no. 1, 1976, pp. 7–42. This is discussed in the previous section. a special issue devoted to the development of 22. Datapro Research, ‘‘Buyer’s Guide,’’ report 70E- database technology containing three other 132-01a–01d, 1974. articles of interest. 23. Datapro Research, ‘‘Buyer’s Guide,’’ report 70E- 11. For example, the July 1972 Datamation issue 652-01a–01d, 1974. (vol. 18, no. 7) contained three introductory 24. Datapro Research, ‘‘Buyer’s Guide,’’ report 70E- articles: R.F. Schubert, ‘‘Basic Concepts in Data 491-01a–01g, 1974. This is the longest and Base Management Systems, pp. 42–47; A.C. most technical of the individual reports. Patterson, ‘‘Data Base Hazards,’’ pp. 48–50, 25. For additional information, see K.R. Blackman, and R.A. McLaughlin, ‘‘Building a Data Base,’’ ‘‘Technical Note—IMS celebrates thirty years as pp. 51–55. an IBM product,’’ IBM Systems J., vol. 37, no. 4, 12. The 27 Feb. 1974 issue of Computerworld con- 1998; http://www.research.ibm.com/journal/sj/ tained a special report, ‘‘Charting a course with 374/blackman.html. Data Base Management Systems,’’ that had 26. Datapro Research, ‘‘Buyer’s Guide,’’ report 70E- 12 articles, including ‘‘Data Base Systems the 757-01a–01d, 1974. Wave of the 1970s,’’ ‘‘For the Unwary, DBMS 27. R.M. Curtice, ‘‘The Outlook for Data Base Man- Can Bring Unbridled Disaster,’’ and ‘‘Gener- agement,’’ Datamation, vol. 22, no. 4, 1975, alized DBMS Not for All.’’ pp. 46–49. 13. Technical Publishing began Datamation in 28. D.E. Cuozzo and J.F. Kurtz, ‘‘Building a Base for 1958; Canning Publications began EDP Analyzer Data Base: A Management Perspective,’’ Data- in 1963, and CW Communications began pub- mation, vol. 19, no. 10, 1973, pp. 71–75. lishing Computerworld in 1967. Datamation was 29. G Schussel, ‘‘When Not To Use a Data Base,’’ an example of a controlled-circulation periodical Datamation, vol. 21, no. 11, 1975, pp. 82, 91, 98. that was mailed free to qualified individuals; 30. R.L. Nolan, ‘‘Computer Data Bases: The Future thus, most people in the ADP community sub- is Now,’’ Harvard Business Rev., vol. 51, no. 5, scribed to it. Most organizations’ technical libra- 1973, pp. 98–114. ries had paid subscriptions to Computerworld 31. R. Hollenbach, ‘‘An Application of a Data Base and EDP Analyzer. System,’’ Data Management, vol. 11, no. 9, 14. D.K. Hsiao, ‘‘Will the Real VLDB Conference 1973, pp. 68–70. Please Stand Up,’’ SIGMOD Record, vol. 12, 32. J.S. Blanchard, ‘‘We Bet Our Company on Data no. 3, Apr. 1982, pp. 8–14. Base Management,’’ Datamation, vol. 20, no. 9, 15. Datapro Research, ‘‘Buyer’s Guide,’’ report 70E- 1974, pp. 61–65, quote p. 64. 010-61b, 1974. 33. V. Powers, ‘‘Implementing Generalized Data 16. ‘‘The Debate on Data Base Management,’’ EDP Base Management Systems,’’ Data Manage- Analyzer, vol. 10, no. 3, 1972, pp. 1–16. This ment, vol. 13, no. 5, 1975, pp. 36–40. issue contains an extensive discussion of the 34. P.H. Cheney and N.R. Lyons, ‘‘MIS Update,’’ Codasyl-Guide/Share controversy including Data Management, vol. 18, no. 10, 1980, ‘‘Some Arguments against the DBTG Pro- pp. 26–32. posals,’’ ‘‘Rebuttals to Arguments,’’ and a letter 35. E. Fong, J. Collica, and B. Marron, ‘‘Six Data from Gene Altshuler, Manager GUIDE Informa- Base Management Systems: Feature Analysis tion Management Group commenting on the and User Experiences,’’ report NBS TN-887, draft of the Analyzer article. 1975. 17. ‘‘What’s Happening with CODASYL-Type DBMS?’’ 36. A ‘‘departmental standard’’ meant that the orga- EDP Analyzer, vol. 12, no. 10, 1974, p. 10. nization supported use of the package with train- 18. The Integrated Data Store (IDS) was developed ing and technical support. Although subelements at General Electric under the direction of could use another DBMS, they would have to , as outlined in Bachman’s provide compelling reasons and get formal ap- contribution to this issue. proval to use something other than the standard. 19. Datapro Research, ‘‘Buyer’s Guide,’’ report 70E- 37. I.R. Palmer, ‘‘Appendix A: Case Studies,’’ Data 272-02a–02c, 1974. Base Systems: A Practical Reference, QED Infor- 20. In 1980, a Computerworld analysis showed that mation Sciences, 1975. of the 11 systems available for IBM System/ 38. L.J. Cohen, Data Base Management Systems, 360/370 systems, only Cullinane’s IDMS and a QED Information Sciences, 1973. version of Cincom’s TOTAL were Codasyl com- 39. J.K. Lyon, ‘‘The Role of the Data Base Administra- pliant. See M. Rosenberg, ‘‘DBMS Offerings tor,’’ Data Base, vol. 3, no. 4, 1971, pp. 11–12. Defy Comparison,’’ Computerworld, vol. 14, 40. R.L. Nolan, ed., Managing the Data Resource no. 28, 14 Jul. 1980, pp. 8–11. Function, West Publishing, 1974, quote p. 40.

40 IEEE Annals of the History of Computing 41. J.W. Luke, ‘‘Data Base Systems: Putting Man- 57. In the early 1980s, Datamation magazine started agement Back in the Picture,’’ CSC Report, publishing the revenues of ‘‘The Leading U.S. DP vol. 9, 1975, p. 9. Companies,’’ including total revenue, DP Revenue 42. R.F. Schubert, ‘‘Basic Concepts in Data Base Man- (present and previous year), and growth rates. agement,’’ Datamation, vol. 18, no. 7, 1972, p. 47. For 1983, a company had to have data process- 43. J.J. Cahill, ‘‘A Dictionary/Directory Method for ing revenues of at least $94.5 million dollars (up Building a Common MIS Data Base,’’ J. Systems 27% from the 1982 minimum of $74.3 million). Management, vol. 21, no. 11, 1970, p. 23. IBM headed the list with $35.6 billion in revenues 44. W.R. Synnott and W.H. Gruber, Information Re- for 1983, and the only database vendor was Culli- source Management: Opportunities and Strategies net at number 94 with revenues of $198 million. for the 1980s, John Wiley & Sons, 1981. 58. S.T. McClellan, ‘‘Cullinet: A Company that 45. R.L. Forman, ‘‘Fulfilling the Computer’s Promise: Knows How to Sell,’’ The Coming Computer The History of Informatics, 1962–1982,’’ section Industry Shakeout, John Wiley & Sons, 1984, 9.26, Corp., 1984. pp. 243–246. 46. Mark IV User Group, Proc. IX MARC IV User 59. Because of their size and diverse offerings, it is Group Meeting, appendix F, 1971, Evan Linick impossible get accurate financial data about Collection, CBI 130, Charles Babbage Inst., DBMS sales for IBM and Intel. Univ. of Minnesota, Minneapolis. 60. Although Cincom made the Datamation 100 in 47. J.A. Piscopo, ‘‘Oral History Interview by Thomas 1984 for the first time, it was recognized on the Haigh, 03 May, Washington DC,’’ 2002, OH 342, Datamation 100 1979 list for its growth rate Charles Babbage Institute, Univ. of Minnesota, and its revenues were given as $30.9 million. Minneapolis. 61. D.M. Chamberlin, ‘‘Relational Data-Base Manage- 48. Datapro Research, ‘‘Buyer’s Guide,’’ report 70E- ment Systems,’’ ACM Computing Surveys, vol. 8, 010-61l, 1979. no.1,1976,pp.43–66,andE.F.Codd,‘‘A 49. For example, Computerworld had a special re- Relational Model for Large Shared Databanks,’’ port, ‘‘Data Base Management Systems: Over- Comm. ACM, vol. 13, no. 6, 1970, pp. 277–390. sold? Underused?’’ in the 31 Oct. 1977 issue 62. A.S. Michaels, B. Mittman, and C.R. Carson, ‘‘A (vol. 11, no. 44) with 13 articles including: Comparison of the Relational and CODASYL Ap- ‘‘Growing Body of Users Stakes Hopes in proaches to Data-Base Management,’’ ACM Com- DBMS,’’ ‘‘Data Base Provides Business Model,’’ puting Surveys, vol. 8, no. 1, 1976, pp. 125–151. and ‘‘Multiple DBMS Operate with Little File 63. E.F. Codd, ‘‘Relational Database: A Practical Redundancy.’’ Foundation for Productivity,’’ Comm. ACM, 50. Figure based on the 1979 Industry Briefing by vol. 25, no. 2, 1981, pp. 109–107. Int’l Data Corporation, Datapro Research, 64. Datapro Research, ‘‘Buyer’s Guide,’’ report 70E- ‘‘Buyer’s Guide,’’ report 70E-010-61a, 1979. 052-08a, 1983. It is not the purpose of this arti- 51. Frost and Sullivan, ‘‘Data Base Management cle to determine whether a package is relational Services Software Market,’’ report #A747, 1979, or not. Usage is based on identified references. pp. 208–213. 65. Unfortunately, Model 204 did not merit a sepa- 52. Datapro Research, ‘‘Buyer’s Guide,’’ report 70E- rate report in any of the Datapro Buyer’s Guides. 010-61a, 1980. 66. Datapro Research, ‘‘Buyer’s Guide,’’ report 70E- 53. ‘‘Database Management Systems,’’ report # 010-61c, 1982. 703, Int’l Resource Development, 1986. 67. Personal communication: C. McQueen, to 54. Datapro Research, ‘‘Buyer’s Guide,’’ report T. Bergin Cincom, 3 Feb. 2009. 70E-010-61a, 1977. This revived a term com- 68. T. Haigh, ‘‘A Veritable Bucket of Facts,’’ The History mon in the early 1960s when the idea of build- and Heritage of Scientific and Technological Infor- ing complex ‘‘totally integrated’’ management mation Systems: Proc. 2002 Conf., W.B. Rayward information systems was first widely discussed. and M.E. Bowden, eds., Information Today, 2004. See also T. Haigh, ‘‘Inventing Information Systems: The Systems Men and the Computer, Contact Thomas J. (Tim) Bergin at tbergin@ 1950–1968,’’ Business History Rev., vol. 75, american.edu. See page 5 for his full biography. no. 1, 2001, pp. 15–61. 55. Cincom, ‘‘Courage, Creativity and Commit- Contact Thomas Haigh at [email protected]. See ment: 25 Years in the Pursuit of Excellence,’’ [no page 25 for his full biography. date], p. 113. According to Datapro, Total was upward compatible to TIS: Datapro Research, ‘‘Buyer’s Guide,’’ report 70E-10-61p, 1982. 56. Datapro Research, ‘‘Buyer’s Guide,’’ report 70E- 010-61p, 1982.

October–December 2009 41