[3B2-9] man2011040070.3d 3/11/011 19:30 Page 70

Biographies Charles W. Bachman: Software Pioneer

Thomas Haigh University of Wisconsin, Milwaukee Editor: Thomas Haigh

Charles W. Bachman was born in position based on a straight-line extrapolation of its Manhattan, Kansas, in 1924 but did previous trajectory. This convinced him both of the im- not stay there long. His father, a portance of predictive control loops and the difficulty Hall of Fame football coach, moved of accurately hitting future targets. After the war their family several times before set- ended, he returned to Michigan State and completed tling in East Lansing, Michigan, for a his mechanical engineering degree in 1948. long spell working at what was then It was time to leave East Lansing again. An interest in Michigan State College. The youn- thermodynamics led Bachman to graduate school at ger Bachman retained a lifelong in- the University of Pennsylvania, although ‘‘it wasn’t terest in sports, playing on several until I arrived there that I had realized that the engi- teams during high school. He earned strong grades in neering graduate school at Penn was a night school’’ most subjects but decided on an engineering career at and that his classmates all had day jobs. He filled the an early age. At 11 years old, he had triumphed in a daylight hours with management courses from Penn’s local Soap Box Derby, in which children custom build Wharton School, satisfying most of the coursework and race small, unpowered cars.1 When I interviewed requirements for an MBA even though there was no for- him in 2004, he told me, ‘‘I was never interested in any- mal joint degree in engineering and business. In the thing else. I was always going to be an engineer.’’ summer of 1949, he married his college sweetheart, Connie Hadley, two weeks after her graduation from Early Education and Career Michigan State. Before Bachman could realize this dream, there was the Thenextyearhegraduated amid the uncertainty little matter of World War II, which was raging that accompanied the outbreak of the Korean War. throughout his high school years. Resolving to enter A search for work in the Great Lakes region culminated the military with a little additional preparation, he in a job with Dow Chemical in Midland, Michigan. graduated a semester early from high school to spend Bachman’s career at Dow began with a nine-month the spring and summer of 1943 satisfying freshman spell rotating through the plant’s various departments. college requirements at Michigan State University. He After that, he settled down to his role as ‘‘an engineer, spent two years in the Army Antiaircraft Artillery at a drafting table, in a room with more than one hun- Corp, serving in New Guinea, Australia, and the Philip- dred engineers.’’ The chemical industry is oriented pines. Antiaircraft guns of this era used a simple me- around networks of processes, as the products of one re- chanical computer to target a plane’s predicted future action become the inputs for another. He worked in a

Background of Charles W. Bachman Born: 1924, Manhattan, Kansas technology subsection, and manager of data manage- Education: BSc (mechanical engineering), Michigan ment software for the advanced system division), State University, 1948; MS (mechanical engineering), 1964–1970; Honeywell Information Systems, chief University of Pennsylvania, 1950. staff engineer, 1970–1980; Cullinane Database Sys- Professional Experience: Dow Chemical, assign- tems, 1980–1983; Bachman Information Systems, ments in engineering, finance, manufacturing, and sys- president and CEO, 1983–1988, and chairman, tems work, 1950–1955, manager of corporate data 1988–1996; Cayenne Software, president, 1996–1997; processing, 1956–1959; and manager of data process- Constellar, consultant, 2002–2006; InfiniteIQ, consul- ing research, 1960; General Electric, Integrated Sys- tant, 2009–2010. tems Project in Production Control Services, 1960– Honors and Awards: ACM Turing Award, 1973; 1964; Computer Department (titles include manager Distinguished Fellow of the British Computer Society, of software product planning, manager of applied 1977.

70 IEEE Annals of the History of Computing Published by the IEEE Computer Society 1058-6180/11/$26.00 c 2011 IEEE [3B2-9] man2011040070.3d 3/11/011 19:30 Page 71

group responsible for the water, steam, and for the new operation. By May 1958, Bach- electrical power systems that kept these pro- man had hired 30 people and the department cesses running at its Midland plants. Bach- was beginning to plan its applications. But man’s work centered more on economic Dow ran into economic trouble and began trade-offs than purely technical designs, ana- to lay off its newest hires, slashing 10 posi- lyzing design choices within the context of tions from his group. Others on his team larger set of assumptions to optimize quit in search of more promising positions, expected returns. and Dow eventually cancelled the IBM He recalled having independently rein- order, although not before constructing an vented concepts such as discounted cash empty computer room in its new headquar- flow as part of an effort dubbed the ‘‘Equiva- ters building. lent Capital Project.’’ By 1952, this led to Ordering a 709 had qualified Dow Chem- work analyzing the capital cost of building ical for membership in Share, the group for all Dow manufacturing plants, for which users of large IBM scientific computers. At a Bachman commandeered time with the me- time when IBM itself provided only a limited chanical punched-card machines in the Ac- selection of system programs for its machines counting Department to iteratively solve and there was no independent software in- networks of simultaneous equations repre- dustry, Share played a vital role in establish- senting data collected from several Dow ing a common library of programs and plants. He wired plug boards and operated subroutines between sites. Its committees the machines himself. In 1954, he worked sponsored the joint development of new in the Pricing Department and was responsi- projects, including the ambitious SOS project ble for the pricing of agricultural chemicals. to produce a standard and powerful operat- In 1955, he shifted from analysis to pro- ing system for the 709. duction for two years, working first as a pro- Bachman quickly became a leader among cess engineer at the Saran plant and then as the smaller community of computing cen- the number two manager of the Styron 475 ters intending to use the 709 primarily for plant. These experiences exposed him to business applications. At his second Share many different aspects of business, reinforc- meeting, he met Harry Tellier of General ing his tendency to see activities as part of a Electric, manager of its computer team unified system. working in the Hanford Atomic Reserva- tion. Tellier’s group had created an inge- Share and 9PAC nious system for the earlier IBM 702 In 1957 Dow Chemical decided to join the computer that generalized the laborious computer age. On the recommendation of programming work involved in maintain- its treasurer, with whom Bachman had ing master data files on tape and generating worked closely on his capital accounting reports based on the data contained in project, Bachman was chosen to head its them. This inspired many subsequent new corporate Data Processing Department. systems, including IBM’s own widely used His first job was to recommend a machine Report Program Generator (RPG) family.3 to purchase. As he recalled, Bachman quickly recognized the value of such a system to Dow, and Tellier and he be- I don’t think we ever did a feasibility study. came founding members of the Share Data The idea was that ‘‘Everyone else is doing it. Processing Committee. The committee spon- We should be doing it.’’ Which one should sored a collaborative project to create what we buy was the question. ... It wasn’t clear became known as 9PAC, an updated version what was going to happen with these com- of the Hanford report generator package for 2 puters, what they were good for. the 709. (I have previously discussed the de- velopment of 9PAC and its importance to He was courted by IBM, with a tour of its the development of data management soft- Poughkeepsie factory, and by Univac which ware,soIwillnotdosoindetailhere.4 It provided a two-week programming course. entered use in 1959 and was widely Impressed by its ability to handle both busi- deployed. Among its most important innova- ness data processing and technical comput- tions, which Bachman credits to discussion ing, he recommended renting an IBM 709. between himself and Russ McGee of GE’s Computer customers generally waited a Hanford group, was support for hierarchical year or two for delivery, during which time parent-child relationships within a magnetic- they hired and trained data processing staff tape file—for example, following each

October–December 2011 71 [3B2-9] man2011040070.3d 3/11/011 19:30 Page 72

Biographies

customer record with records for the orders General Electric was therefore investing placed by that customer. heavily in the latest management methods It is a tribute both to the practical flexibil- and technologies. One of the responsibilities ity of Bachman’s intelligence and Share’s ef- of its corporate-level teams was to research fectiveness in allowing participants to learn areas such as operations research, simulation, from the experiences of others that he forecasting, and automation so that effective could play a significant role in the 9PAC de- new practices could be spread across its di- sign without having written a data process- verse empire. Bachman was recruited to its ing application or ever actually taking prosaically named Production Control Ser- delivery of a computer. Share also established vices Group, headed by H. Ford Dickie, who a Committee on Theory Information Han- was an early and enthusiastic advocate for dling, with Bachman as an active member. the ‘‘total systems approach’’ of using com- The committee tried to abstract a concept puters and cybernetics to revolutionize the of data away from the constraints of its phys- control of industrial production.9 Bachman’s ical representation in tape or card files. It experience with data processing, engineer- formulatedtheideaofdataasasetof ing, management, and cost accounting ‘‘triples’’—so called because each was defined made him a perfect fit. by the unique combination of a single data In a 1962 presentation prepared for GE’s value, an identifying name, and a designated Information Systems Task Force, Bachman owner.5 He also worked on the design of wrote that ‘‘the introduction of computers a still more powerful successor, Ingen, that into our business as a new resource’’ had would have provided more flexibility to ‘‘created an atmosphere in which the entire restructuring data hierarchies as needed for outlook of information systems is changing.’’ particular operations.6 Bachman’s participa- He went on: tion in Share was cut short by Dow’s decision to cancel its order with IBM, but these efforts Many of the old equilibriums between mar- clearly influenced his subsequent work on keting, manufacturing, finance and engineer- database management. ing are being upset by brash young men who think they know all about everybody else’s GE and the Integrated Data Store business and who propose to change it. The Bachman left Dow Chemical in 1960, mov- orderlyabsorptionofthesemenwiththeir ing to New York City for a corporate-level tools is today’s order of business. staff position with General Electric. GE was admired throughout corporate America both By analogy with his experience with the sepa- for its leading position in high-technology ration of design and production within industries and its equally innovative manage- engineering, Bachman argued that a new rial culture. It was the leading exemplar of ‘‘information system designer’’ role should what business historian Alfred Chandler was be established to oversee the system as a about to dub the ‘‘M-form’’ corporation, whole. He identified the ‘‘information inven- with a small headquarters staff governing a tory’’ as a crucial subsystem in this new vision sprawling range of different business divi- of computerized management, writing that sions manufacturing products from light ‘‘the maintenance of information inventories bulbs to nuclear-powered steam turbines.7 is one of the largest work assignments given to Responsibility for operational decisions was computers today. It is also a thing that cur- decentralized to these divisions, or as they rently they are doing very poorly.’’10 were known within the company, ‘‘depart- The group was using a factory in Philadel- ments.’’ In 1956, the firm had opened an in- phia as a testbed for its Integrated Systems ternal business school in Crotonville, New Project #2, which eventually produced an York, to educate a new cohort of managers integrated system known as MIACS to auto- with the skills needed to run a decentralized mate its planning, scheduling, operational firm. Successful managers were moved fre- control, and inventory control. Running on quently between different parts of the com- a GE 225 computer, it integrated capabilities pany, and they were held accountable for such as production scheduling, economic their performance on a set of standard order calculations, order dispatching, receiv- financially based metrics. This reflected a ing, inventory management, and parts explo- conviction that management was a general- sion. Bachman claims it was the first ized skill that could be separated from integrated manufacturing control system, expertise in any particular industry.8 and the direct ancestor of the programs

72 IEEE Annals of the History of Computing [3B2-9] man2011040070.3d 3/11/011 19:30 Page 73

later marketed as material requirement plan- qualified to tackle the job; he had the skills ning (MRP) and bill of materials processing and mindset of an engineer, combined with systems. MIACS was used internally by a years of systems work, deep immersion number of General Electric departments, and in Share’s efforts to produce reusable general- was offered as a software package to purchas- ized data management software, access to ers of the firm’s more powerful GE 600 models prototype hardware and software, and a sup- that succeeded the 225 later in the 1960s. portive corporate culture. During the 1960s, many firms attempted Today’s database designers rely on graphi- to create what were sometimes called ‘‘totally cal tools to depict the relationships between integrated management information sys- differenttablesasawebofconnections. tems,’’ but MIACS was one of the few am- These data structure diagrams are the direct bitious integrated systems projects to descendents of a technique Bachman devised succeed.11 Asmallteamproducedauseful to illustrate the complex data structures and powerful integrated system in about required to picture the adaptive manufactur- four years. In the process, they invented ing control logic of MIACS. some of the most fundamental technologies Turning IDS from a prototype to some- behind today’s large-scale administrative thing robust and flexible enough for GE to computer systems. This reflected the excep- offer to its internal customers required some tional caliber of the people involved, a testa- changes. Bachman began taking the subway ment to the scale of GE’s investment in from his office in GE’s headquarters on Lex- systems work and operations research during ington Avenue in midtown Manhattan to In- the 1950s and early 1960s. (One of Bach- ternational General Electric, located about 20 man’s predecessors in its production control blocks south, where this work was being car- services group, Harry Markowitz, later won ried out. He delivered two boxes of punched the Nobel Prize for economics).12 cards holding the system code and returned Successful as it was, MIACS is remembered to assist the team as needed. Changes primarily because of the foundation Bach- included the addition of a backup and recov- man created for it with another piece of soft- ery system, disentanglement from GE’s ware, the Integrated Data Store (IDS).13 This proprietary GECOM business application reflected Bachman’s concept of an ‘‘informa- programming language, and improved han- tion inventory.’’ IDS was the first package to dling of memory paging.14 GE publicly adopt what was later called the ‘‘network announced IDS as a supported package for data model,’’ meaning that the system its computer customers in October 1964.15 would support and enforce relationships be- tween records in a much more flexible man- The GE Computer Department ner than earlier packages such as 9PAC. This In 1964, Bachman transferred to GE’s Com- exploited the new power of disk storage and puter Department in Arizona. He was origi- was an essential capability for the storage of nally assigned to its marketing team; a letter data in a form that could support applica- sent to him summarizing his intended tions for different functional areas. IDS pro- responsibilities promised that he would assist vided application programmers with a set of with the launch of IDS as a (bundled) product powerful commands to manipulate data, an by pitching its benefits to potential custom- early expression of what would soon be ers, ‘‘communicate system design tech- called a Data Manipulation Language. It niques’’ to key technical managers in user maintained a separate data dictionary, track- firms, and assist in the preparation of techni- ing information on the different kinds of cal proposals involving IDS.16 But GE’s com- records in the system and their relationships. puter department was in constant flux, and Only the boldness of Bachman’s IDS de- over the next six years, Bachman worked on sign, and the remarkable efficiency with a number of projects within its Project Plan- which he squeezed IDS and the MIACS appli- ning Group, and eventually he found himself cations into a computer with the equivalent in the new Advanced Software Group. of 40 Kbytes of memory, made this possible. A few years after MIACS was completed, This tight coupling meant that IDS and the Bachman helped GE build a still more ambi- Problem Controller, a transaction-oriented tious management information system proj- operating system produced by the team, al- ect on the same foundation. Known as most entirely replaced General Electric’s rudi- WEYCOS 2 (for Weyerhaeuser Comprehen- mentary operating system and other systems sive Operating Supervisor), it was produced software tools. Bachman was uniquely well for Weyerhaeuser’s Wood Products division

October–December 2011 73 [3B2-9] man2011040070.3d 3/11/011 19:30 Page 74

Biographies

in Tacoma, Washington, by a team composed its promulgation of Cobol as the industry of employees from both companies. This was standard applications programming lan- a sequel to WEYCOS, fully operational by guage. As I have discussed elsewhere,4 this 1965, which integrated the Problem Control- group’ reports popularized the concept of a ler transaction-processing system developed ‘‘data base management system’’ and estab- for IDS with a smaller computer connected lished its key features such as separate data- to teletype units across the country. Using definition and data-manipulation languages. WEYCOS, employees in remote offices could Bachman was an active member in the send messages directly to the computer to up- group’s early history, from 1966 to 1968, date records or request information. and his influence on its final reports and Weyerhaeuser credited the WEYCOS sys- draft language specification was profound. tem with dropping order-processing time In particular, the group endorsed Bachman’s from weeks to days, greatly improving its effi- network data model approach over the less ciency and competitiveness. However, the GE flexible hierarchical data model backed by 235 system was slow, relying on the ability to IBM. Bachman himself believes that the queue requests and clear the backlog of lower data-manipulation language proposed by priority transactions over nights and week- the task group had few substantive differen- ends. WEYCOS 2 was intended to serve the cesfromtheonehehaddesignedforIDS. whole company and handle a growing vol- He told me that ume of requests. One initial goal, ultimately abandoned, was for it to run on a cluster of I wrote, rewrote, and rewrote the specification four powerful GE 600 series computers with for the DBTG specifications for a database a shared IDS database and a common memory management system. The original text was bank to coordinate transactions. The team did the text for the GE IDS database management successfully adapt IDS so that a single database system. The ‘‘master’’ and ‘‘detail’’ records of IDS became the ‘‘owner’’ and ‘‘member’’ could be shared by up to eight application records of the DBTG specifications. The word, programs running simultaneously—another ‘‘set,’’ was in the [phrase owner/member set]. key step in the evolution of the database man- These were good improvements. Some of the agement system (DBMS). changes were really to satisfy some member’s Other major additions to the standard GE desire to have something that they could operating system of the era added support for point to as their contribution. My challenge automatic job scheduling, execution of pro- was to preserve the IDS functionality; not to grams from remote locations, and message preserve the original IDS data manipulation 18 transmission between locations. Develop- language in all of its details. ment of WEYCOS 2 ran from 1966 to 1968. In 1972, WEYCOS was still in use, costing Bachman’s other main project of this era around $140,000 a year to support and main- was DataBASIC, implemented by a small tain. Weyerhaeuser stated that its capabilities team working under his direction in 1967– at that time were still not matched by stan- 1968. Basic had originally been designed at dard operating systems, though the gap was Dartmouth College as an interpreted pro- closing.17 gramming language for interactive use on As a product, IDS was considerably more GE computers by people without significant flexible than IBM’s rival Information Man- computer expertise. DataBASIC extended the agement System (IMS) package and was language to include the data-manipulation released several years earlier. However, soft- capabilities needed to use it effectively to ware packages in those days, particularly create and query . This reflected those from computer manufacturers, were general faith during the late 1960s that time- tied to particular platforms. GE had less sharing technology would offer low-cost than one tenth of the market share held by computing services to professionals via tele- IBM, and its computer department suffered phone lines and cheap terminals. Bachaman from a relative lack of resources and manage- wrote that it was ‘‘designed to do for the rial stability. Although IDS was not the most IDS/COBOL world that which BASIC had widely used package of the 1960s, it was one done for the FORTRAN world’’ and ‘‘permit- of the most influential. ted business professionals who are not This influence came in large part through programmers to construct, maintain and Bachman’s participation in the Data Base operate their own private data bases.’’19 Data- Task Group organized by Codasyl, the data BASIC was based on the data triple idea Bach- processing standards group best known for man had earlier explored within Share. It was

74 IEEE Annals of the History of Computing [3B2-9] man2011040070.3d 3/11/011 19:30 Page 75

Figure 1. teaching a group of Honeywell staff, probably in Vienna, Austria in 1978. (Image courtesy of Charles Babbage Institute.)

a completely ad hoc system, with no record group in Boston (see Figure 1). The era’s surviv- types and no predefined field names. A de- ing mainframe companies were then engaged cade later, systems such as dBase II were in massive integrated projects to deliver next- being widely used on personal computers generation computer architectures and for such purposes. DataBASIC was ahead of advanced software features. Bachman was its time, but even though GE supplied the concerned with the design of data manage- software for its 400 and 600 product lines, it ment software for these systems, including was not widely used. communication facilities and the creation of Within the GE Computer Division, Bach- databases distributed across multiple com- man’s role as ‘‘Manager À Application Tech- puters. For Honeywell, as for IBM, these mas- nology Subsection’’ gave him an opportunity sive projects consumed the attention of to develop his position as one of the industry’s many top technical people for years before leading experts on data management and in- being cancelled in favor of a more evolution- formation systems. However, he was less ary approach. Bachman moved to Boston, comfortable with the managerial side of the Massachusetts, in August 1970 to assume his job. His 1969 performance appraisal, while NPL responsibilities. Bachman recalled unstinting in its praise of his technical leader- ship, noted that ‘‘Charlie needs to improve his a thing called Shangri-La, which was a big project that GE had to design their new com- communication channels with other compe- puter line. That new computer project was tent people, who do not understand or agree taken over by Honeywell in the fall of 1970, with his ideas without compromising what so my main responsibilities for a couple he believes is right. Charlie’s primary interests years after the GE/Honeywell merger had 20 are technical rather than managerial.’’ to do with that new product line (NPL) development.21 Honeywell and the 1970s In 1970 Honeywell acquired GE’s computer In 1973, Bachman received the ACM business. Bachman worked for the merged op- Turing Award, the most prestigious in the eration in Honeywell’s advanced research computing field. Its previous winners all

October–December 2011 75 [3B2-9] man2011040070.3d 3/11/011 19:30 Page 76

Biographies

had backgrounds in mathematics or physics, Thedebatewasthecenterpieceofthe rather than engineering or business. This ACM SIGFIDET Workshop on Data Descrip- was the first time the award had been pre- tion, Access, and Control in Ann Arbor, sented to someone whose work focused on Michigan, held 1–3 May 1974. By January business data processing or to someone 1974, plans for the debate were underway whose entire career was spent in industry and letters were exchanged between rather than academia. He remains one of a Codd, Date, other team members, and the handful of winners not to have a PhD.22 workshop organizers to negotiate the format. Bachman does not believe the award had A full afternoon was reserved for ‘‘Debate: much impact on his subsequent career prog- The Differences and Similarities Between ress, although it caused him to develop an in- the Data Base Set and Relational Views of terest in Alan Turing’s life and even to visit Data.’’ Discussion spilled over into the his mother, then in an English nursing home. evening. His award acceptance lecture, ‘‘The Bachman’s strategy was to argue that the Programmer as Navigator,’’ was an influential relational model was mathematically equiva- declaration of a new world in which complex lent to his network approach, making the data structures provided the structure of cor- choice a matter of implementation efficiency porate computing systems, around whose and stylistic preference. The debate was ulti- topographies individual application programs mately inconclusive, as neither side managed would navigate.23 The award cemented Bach- to convince the other. At this point, the first man’s position within the industry as a lead- practical relational system was still several ing expert on DBMSs and the most respected years away, and relational products would advocate for the network data model and the not begin to sell in significant numbers for systems influenced by the Codasyl approach another decade. The choice of approaches that were rapidly gaining market share in was therefore also one between scientific the mainframe world. As such, he stood in rigor and engineering pragmatism, and be- opposition to the ideas of Edgar F. (‘‘Ted’’) tween elegant, but unproven theory and Codd, a mathematically inclined IBM re- functioning high-performance technology.26 search scientist whose relational model for database manipulation had attracted a grow- Data Models and Schemas ing band of supporters and was beginning to While unconvinced by the relational legitimate database systems as a theoretically approach, Bachman did not view his own respectable research field within computer as the endpoint of database science. evolution. In the late 1970s, he worked on Bachman had been arguing for several what he called the ‘‘role data model,’’ years with Codd and his leading supporters, extending the network model used by IDS although he insists that their rivalry was to enhance its data-modeling capabilities. friendly, noting that despite their ‘‘strong dif- When several database records described ferences in ideas’’ they would ‘‘go out and aspects of the same real-world entity (for ex- have dinner together’’ afterwards.24 Matters ample, records for the same individual in came to a head when Bachman delivered a rolesasemployeeandcustomer),these preview of his Turing Award lecture at an- records could be tied together as members 27 other meeting. As C.J. Date, Codd’s primary of a set. supporter, told me The role data model had similar motiva- tions to the widely adopted technique of en- I could feel Ted gripping his chair, and the in- tity relationship modeling, introduced by stant that Charlie sat down, Ted leapt up ... Peter Chen in the 1970s. (Bachman himself he said of course, ‘‘I’d like to congratulate called the resemblance ‘‘superficial’’ in my Charlie on the Turing Award. That’s really oral history, although he knew Chen, having great, and he deserves it; he’s a very nice worked with him on a joint paper on data guy. ...None of that affects the fact, however, models in this era.28) However, the addi- that what he is saying in his paper is dead tional information represented in these mod- wrong.’’ That was typical Ted! Then he said, els was not usually integrated into the data- ‘‘There are so many things that need to be definition languages of DBMSs themselves, said in order to deconstruct Charlie’s as Bachman had hoped to achieve by replac- argument. ... This is not the right forum to do it. ... So instead, I would like to issue a ing the network model. challenge that Charlie. ... have a formal de- Bachman’s work on the role data model bate on these matters.’’25 was one expression of his broader interest

76 IEEE Annals of the History of Computing [3B2-9] man2011040070.3d 3/11/011 19:30 Page 77

in the concept of data independence. This Computing Communications Protocols is the principle that application programs Although Bachman is known primarily for should be shielded from the data’s actual or- his database work, his contribution to the ganization and location so that changes can field of computer communications was be made to the way data is stored without equally significant. This came when he requiring reprogramming. IDS had been an served as chair of the Open Systems Inter- important but incomplete step in that direc- connection (OSI) subcommittee of the Inter- tion. Database textbooks typically introduce national Organization for Standardization the concept of data independence via (ISO) from 1978 to 1982. Data communica- the cryptically named ANSI-SPARC Three- tion was becoming increasingly important Level Architecture, distinguishing between for corporate computer users in the 1970s, internal, conceptual, and external represen- and Bachman recognized that this would tations of data. take place not just between users and com- In fact, the group’s promulgation of the puters, or even from one computer to an- three-tier model was something of a consola- other, but from one program to another. tion prize. The Systems Planning and Resour- This would require a new approach, more ces Committee (SPARC) had been established flexible than the proprietary, terminal- as a steering committee within the American oriented communications protocols then National Standards Institute (ANSI) to deal used in the mainframe industry. Bachman with computer standards. Its Study Group À had been working within his company on Database Systems was established in 1972 to something he called the Honeywell Distrib- ‘‘review the current state of development in uted System Architecture. This ultimately the database systems field with the objective became Honeywell’s submission to the to determining whether standardization American national committee, chaired by activities were appropriate.’’29 Bachman was Bachman, which in turn formed the basis vice chairman of the committee and took for the American submission to the interna- the lead in attempting to draft a suitable tional subcommittee, also chaired by Bach- standard ‘‘in such a way that the Basic Con- man. Regarding the final product, he said, cepts [behind] IDS and the DBTG are pre- ‘‘Well, I would say that 75% was based on served.’’30 The committee met frequently my work at Honeywell. It was all well docu- for years, but no consensus could be reached, mented. We worked, as you sometimes something Bachman blames particularly on have to do, to cover the traces.’’33 the influence of IBM, which at that point By 1978, Bachman’s subcommittee had al- backed both the hierarchical and relational ready produced a conceptual diagram of the approaches but not the network model. As communications process.34 Today the entire Bachman recalled this during our interview, OSI effort is remembered primarily for this ‘‘The sense of the group was that there was seven-layer model, endorsed by the ISO’s nothing worth doing in terms of standardiza- Technical Committee on Data Processing tions, so what other interesting thing could the following year.35 This was an abstracted we do.’’31 model of communication, isolating protocols A 1977 paper summarized the committee’s at each level from the details of the levels be- attempt to at least identify the different levels neath them. For example, a transport proto- andinterfacesinvolvedinaDBMS.32 The col (level 4) would not have to deal with most abstract level, the conceptual schema, network routing (level 3), physical addressing was supposed to represent real-world entities (level 2), or bit-by-bit transmission over a on a level abstracted not only from the details physical medium (level 1). As long as interfa- of physical storage but also from the design ces between levels were standardized, it constraints that the choice of either the net- would be possible to switch protocols at any work or relational data model imposed at level without disturbing the functioning of the ‘‘internal level.’’ The committee had the other levels. agreed to differ with respect to choice of By the early 1990s, this was the organizing data model, but offered hope that standard- framework of computer communications ization could eventually be accomplished at textbooks, and posters of the model were fix- this new and more abstract level. Bachman tures in the cubicles of IT specialists every- sees computerized data modeling tools, able where. OSI eventually proposed standard to generate database schemas for actual sys- protocols for particular layers, which gained tems from a higher-level design, as a realiza- considerable institutional backing but were tion of the conceptual schema concept. ultimatelysweptawaybytheunexpected

October–December 2011 77 [3B2-9] man2011040070.3d 3/11/011 19:30 Page 78

Biographies

mass adoption of Internet standards. Internet Bachman Information Systems protocols do not map exactly onto the seven- Bachman left Cullinane Database Systems layer model, but they follow its general with a ‘‘consulting agreement’’ that provided approach of layering and abstraction. a year of salary and access to computer time. On 1 April 1983, he founded Bachman Infor- Cullinane Database Systems mation Systems with his wife Connie. The In 1980, Bachman went to work for Culli- company built on his long-standing interest nane Database Systems, a mainframe soft- in data models and targeted widespread ware company. He felt underappreciated by enthusiasm for what was then known as his managers and had grown frustrated with computer-aided software engineering (CASE). Honeywell, which was unable to invest sig- An initial project with the Norwegian firm nificant resources to develop new database Nordata to build a graphical design tool for software. In 1979, he wrote to a newly in- IDMS did not go well, but by 1984, Bachman stalled manager that ‘‘We have wasted more had partnered with Phil Cooper, a software engineering resources and frustrated and industry entrepreneur and received US$2 lost more engineers by start/stop and switch million in venture capital from two pillars operation than we have applied to new prod- of that industry: Venrock and Kleiner Per- ucts during the last few years.’’36 kins, Caufield and Byers. Its first product, In contrast, Cullinane was rapidly grow- Data Analyst, was released in 1987. This was ing and located close to Bachman’s house. a PC-based graphical tool to diagram data Its main product, IDMS, followed the Coda- models as conceptual views, in a notation syl model that Bachman had helped to de- based on Bachman’s partnership data fine. Indeed, it was a direct descendent of model. It could translate these models into his IDS, having originally been created by physical designs for mainframe database sys- a team at B.F. Goodrich attempting to pro- tems such as IDMS and could import existing duce a compatible replacement for IDS metadata from databases. As a tool to support that could run on the firm’s new IBM hard- mainframes, it was priced at $15,000 for each ware. Acquired by John Cullinane, a skilled user, much higher than the typical cost of salesman, IDMS had become one of the personal computer software. leading software products of the early Data Analyst was a success. Later versions 1980s. Bachman called Cullinane and was ran on OS/2 rather than MS-DOS and added quickly hired as the vice president of ap- support for IBM DB/2 systems. This led Bach- plied research. man Information Systems into a strategic Bachman came to view this as something partnership with IBM, which was pursuing a of a figurehead position, telling me that ‘‘I grand vision of metadata repositories and en- sometimes think the role that John really terprise data models. Bachman himself had had in mind for me was for me to sit by the become chairman and was focusing on this front door and meet the visitors coming in’’ work and its technical challenges while leav- to show that ‘‘Charlie Bachman is here ing management mostly to others. Data Ana- now.’’ ‘‘I was supposed to go out and give lyst was extended to model processes and John’s speeches,’’ he continued. ‘‘The prob- data structures, and Bachman Information lem was that I would go out and give my Systems was soon listed on the NASDAQ speeches, and that wasn’t quite what John exchange. The Data Analyst product name wanted. I think I disappointed John in eventually encompassed four products: the some ways.’’37 Data Analyst, itself, the Database Administra- Bachman pursued several projects at Cull- tor, a Forward Engineering tool, and a Re- inane, including an attempt to revive the verse Engineering tool. idea of a personal DBMS, the development The IBM relationship deteriorated, how- of computer tools to support data-modeling ever, after a rival was selected as its key part- work, and an extension of his ongoing work ner in the area. Data Analyst was eventually on new data models to something he called ported to Windows, and the company itself the ‘‘Partnership Data Model.’’38 But Culli- merged with another firm to create Cayenne nane’s growth was beginning to slow as en- Software. The late move to Windows and its thusiasm switched toward relational DBMSs. high cost meant that Data Architect was ulti- Facing new pressures, the firm launched a fu- mately eclipsed by lower cost rivals, most no- tile effort to rebrand IDMS as a relational sys- tably Rational Rose. tem and, in 1983, fired Bachman as irrelevant A downward spiral began, and in 1998, to its new direction. the company was sold to Sterling Software

78 IEEE Annals of the History of Computing [3B2-9] man2011040070.3d 3/11/011 19:30 Page 79

for a fraction of its earlier value. Bachman views this as the biggest disappointment of his career, saying

failure came about because I wasn’t the man- ager the company needed. The company didn’t learn to live within its revenue. Secondly ...our business did not understand what it was doing and how its world was changing, fast enough to survive and grow andbeasuccess.Weblewtheopportunity that Rational Software seized.39

Although by then in his mid-70s, Bach- man was not ready for full-time retirement. He began work as a consultant, helping com- panies model their data. He worked for two years with a startup called Constellar, which was trying to produce a package to help companies integrate information systems Figure 2. Charles Bachman in 2004, at home with his wife, Connie. acquired during mergers. At the time of our interview in 2004, he was engaged in an elab- orate data-modeling project for Cbr System, References and Notes operator of the Cord Blood Registry (see 1. These details are taken from C.W. Bachman, Figure 2). He retained a keen interest in ‘‘Oral History Interview by Thomas Haigh,’’ new ideas and technologies, although it was 25–26 Sept. 2004, ACM Oral History Interviews somewhat incongruous to see the man who collection. Material in this biography not explicitly invented the DBMS working with gusto to referenced should be assumed to come from this prototype a design in Microsoft Access interview, referred to hereafter as CWB TH OH. 2003. He is currently working on a book 2. CWB TH OH, p. 16. about data modeling and his enhanced 3. SHARE, 9PAC and IDS are discussed in more de- data-structure diagrams, documenting their tail in T. Haigh, ‘‘How Data Got its Base: Gener- evolution since 1964 and application to alized Information Storage Software in the today’s database systems. He considers data 1950s and 60s,’’ IEEE Annals of the History of structure diagrams to be a new algebra. Computing, vol. 31, no. 4, 2009, pp. 6–25. 4. T. Haigh, ‘‘How Data Got its Base: Generalized Summary Information Storage Software in the 1950s and Bachman’s career was long, varied, and dis- 1960s,’’ IEEE Annals of the History of Computing, tinguished. He ran a chemical plant, created vol. 31, no. 4, 2009, pp. 6–25. a capital accounting systems, headed an 5. CWB OH TH, pp. 23–24. B. Orchard-Hays, early data processing group, pioneered the ‘‘SHARE Committee on Theory of Information application of computers to manufacturing Handling: Report TIH:1,’’ Charles W. Bachman, control, led efforts to standardize database Papers, CBI 125, Charles Babbage Inst., 1958. and computer communication concepts, 6. C.W. Bachman, ‘‘INGEN Proposal,’’ Charles won the highest honor in computer science, W. Bachman, Papers, CBI 125, Charles Babbage and founded a publicly traded company. In- Inst., 1 May 1959 or 1960. novation is a collaborative process, but by 7. A.D. Chandler, Jr. Strategy and Structure: Chap- creating IDS and advocating forcefully for ters in the History of the American Industrial En- the concepts behind it Bachman did more terprise, MIT Press, 1962. than any other person to invent the DBMS 8. For a snapshot of General Electric’s managerial as we know it today. cultureintheperiodseeR.J.Cordiner,New Fron- Through all of this he retained an engi- tiers for Professional Managers, McGraw-Hill, 1956. neer’s zest for the elegant solution of difficult 9. H.F. Dickie, ‘‘Integrated Systems Planning at G.E.,’’ problems and faith in the power of careful Management Control Systems, D.G. Malcolm and analysis and the systems approach to make A.J. Rowe, eds., John Wiley & Sons, Inc., 1960, the world a better place. As he wrote in a pp. 137–155. note at the end of our oral history transcript, 10. C.W. Bachman, ‘‘Presentation Prepared for the ‘‘My work has been my play.’’ Information Systems Task Force, September 4,’’

October–December 2011 79 [3B2-9] man2011040070.3d 3/11/011 19:30 Page 80

Biographies

Charles W. Bachman, Papers, CBI 125, Charles correspondence. The event features promi- Babbage Inst., 1 May 1962. nently in discussion of the period by database 11. T. Haigh, ‘‘Inventing Information Systems: The Sys- researchers. tems Men and the Computer, 1950–1968,’’ Busi- 27. C.W. Bachman and M. Daya, ‘‘The Role Con- ness History Rev., vol. 75, no. 1, 2001, pp. 15–61. cept in Data Models,’’ Proc. 3rd Int’l Conf. Very 12. According to Bachman SIMSCRIPT, one of the Large Data Bases, VLDB Endowment, 1977, accomplishments for which Markowitz won the pp. 464–476. prize, was a continuation of work begun on Gen- 28. An incomplete draft of ‘‘On the Basic Similarity eral Electric’s first Integrated Systems Project. of Three Major Data Models’’ is in the Bachman 13. C.W. Bachman, ‘‘The Origin of the Integrated collection at CBI. It does not appear to have Data Store (IDS): The First Direct-Access been published. DBMS,’’ IEEE Annals of the History of Computing, 29. C.W. Bachman, ‘‘Summary of Current Work: vol. 31, no. 4, 2009, pp. 42–54. ANSI/X3/SPARC/Study Group - Database Systems, 14. CWB OH TH, pp. 71–75. revised January 29,’’ Charles W. Bachman, Papers, 15. J.R. Elliot, Jr. ‘‘More Information Please,’’ CBI 125, Charles Babbage Inst., 1 May 1974. Baron’s, 2 Nov. 1964. 30. C.W. Bachman, ‘‘Memo Subject: ANSI/X3/ 16. V.S. Cooper, ‘‘Letter to Charles Bachman, SPARC/Sudy Gorup À Database Systems Status July 14,’’ Charles W. Bachman, Papers, CBI 125, Report,’’ Charles W. Bachman Papers, CBI 125, Charles Babbage Inst., 1 May 1964. Charles Babbage Inst., 24 Sept. 1974. 17. WEYCOS is discussed in J.P. Fichten, ‘‘The Weyer- 31. CWB TH OH, p. 129. haeuser Information Systems À A Progress Re- 32. D. Tsichritzis and A.C. Klug, ‘‘The ANSI/X3/ port,’’ Proc. Fall Joint Computer Conference, SPARC DBMS Framework Report of the Study part II, Am. Federation of Information Processing Group on Database Management Systems,’’ Societies, 1972; and CWB TH OH, pp. 82–91. Information Systems, vol. 3, no. 3, 1978, 18. CWB TH OH, p. 77. pp. 173–191. 19. C.W. Bachman, ‘‘Biographical Sketch (attached 33. CWB TH OH, p. 114. to May 6, 1977 letter to Chris Bunyan),’’ 34. ISO/TC97/SC16, ‘‘Provisional Model of Open- Charles W. Bachman, Papers, CBI 125, Charles Systems Architecture,’’ Charles W. Bachman, Babbage Inst., 1 May 1977. Papers, CBI 125, Charles Babbage Inst., c. 1 20. S.B. Williams, ‘‘Performance Appraisal: C.W. May 1978. Similar five- and six-layer models ap- Bachman, Manager - Application Technology pear in the group’s internal materials for 1977. Subsection, June 1968–May 1969,’’ Charles 35. H. Zimmerman, ‘‘OSI Reference Model—The W. Bachman, Papers, CBI 125, Charles Babbage ISO Model of Architecture for Open Systems Inst., 1 May 1969. Interconnection,’’ IEEE Trans. Comm., vol. 28, 21. CWB TH OH, p. 101. no. 4, 1980, pp. 425–432. 22. Of the 55 other Turing Award winners through 36. C.W. Bachman, to R.A. Douglas, 12 Feb. 1979, 2010, only four resemble Bachman in having Charles W. Bachman Papers, CBI 125, Charles spent all or most of their careers in industry with- Babbage Inst. out earning a PhD: John Backus, Ken Thompson, 37. CWB TH OH, p. 135. Frances E. Allen, and Charles Thacker. Several 38. The partnership data model is described in US other winners were career academics without patent 4631664, in several documents included PhDs (such as Robert Floyd) or industrial in the Bachman collection at CBI, and along researchers with PhDs (such as Denis Ritchie). with other aspects of Bachman’s career in C.W. 23. C.W. Bachman, ‘‘The Programmer as Naviga- Bachman ‘‘A Personal Chronicle: Creating Better tor,’’ Comm. ACM, vol. 16, no. 11, 1973, Information Systems, with Some Guiding Princi- pp. 653–658. ples,’’ IEEE Trans. Knowledge and Data Eng., 24. CWB TH OH, p. 108. vol. 1, no. 1, 1989, pp. 17–32. 25. C.J. Date, ‘‘Oral History Interview with Thomas 39. CWB TH OH, p. 162 Haigh,’’ Computer History Museum, 13 June 2007; http://www.computerhistory.org/ Thomas Haigh is a historian of computing and an collections/accession/102658166. associate professor in the School of Information Studies 26. Papers from debate participants were published at the University of Wisconsin, Milwaukee. Contact in R. Rustin, ed., Proc. 1974 ACM SIGFIDET him at [email protected]. Workshop on Data Description, Access and Con- trols: Data Models: Data-Structure-Set versus Relational, ACM Press, 1974. The Bachman collection at CBI includes the program for the debate and a collection of related

80 IEEE Annals of the History of Computing