sign in sign up
<<
Home , Computer program, Informatics General

Software in the 1960s as Concept, Service, and Product

Thomas Haigh Colby College Packaged application established a small but important corporate niche during the 1960s. The author charts the shifting meaning of the word software, situates the first software companies within the overall services market, and probes the attractions and limitations of the first packages from the viewpoint of their potential purchasers: managers of processing.

The term software is newer than most of the ually broadened the range of system tools it things that it is today used to describe. Many shipped—from none with the first 701s, to sym- data-processing pioneers assumed that applying bolic assemblers and loaders with the 650 and their electronic to specific adminis- 704, to increasingly complex input–output (I/O) trative tasks would be straightforward. While and control programs with its second-genera- they realized that the new machines would tion transistorized machines. Cobol’s arrival require programming, many expected this to be challenged every computer manufacturer to pro- a one-time, rapid, and easily compartmentalized duce for this complex high-level lan- job, firmly subordinate to the larger task of busi- guage, an effort that frequently strained the ness analysis. They soon discovered that appli- of the art to the breaking point and beyond. cation creation was costly, difficult, and A 1962 advertisement touted the ongoing. By the mid-1950s, they had come to firm’s expertise in this new field, defining soft- care a great deal about programming.1 But only ware as “ aids that sim- around 1960, however, would a well-informed plify the task of telling the computer ‘hardware’ data-processing manager have nodded knowl- to do its job” and observing that the “three edgably if software came up in conversation. basic categories of software” were assembly sys- During the 1950s, the term was not used, tems, compilers, and operating systems.3 This although hardware was already well known as a was clearly what a Datamation editorial writer colloquial term for computer equipment. When had in mind, when observing that software did achieve currency, it was as hardware’s complement, describing everything else the the well-publicized potentials of software have computer manufacturer provided. This ensured been riddled during the past few months with a that term’s widespread, if ill-defined, use.2 barrage of generally well-aimed criticism. The effect has been one of withdrawn embarrass- Software: A fluid concept ment; a quiet revision of delivery schedules; a By the mid-1970s, accepted linguistic usage of program to check out a or two, software had shifted toward “programs and other and some judiciously phrased, -directed pleas operating information used by a computer” for compassion and patience.4 offered by today’s Concise Oxford English Dictionary. Software became a synonym for com- Likewise, when Asher Opler, then responsible puter program, excluding perhaps only a proces- for programming at Computer Usage Corpora- sor’s and the firmware burned into tion, published a 1964 article in Datamation on permanent memory. However, for much of the the “Measurement of Software Characteristics” 1960s, software was commonly understood in a he confined his attention entirely to “automatic narrower sense, as what was later called systems programming and operating systems (software).” software—programs used to construct other pro- Opler suggested that these programs so influ- grams, or operating systems to control comput- enced a computer’s real performance that it was er hardware. During the early 1960s, computer absurd to continue to evaluate a computer’s suit- manufacturers dramatically stepped up their ability for an application on the basis of its hard- efforts in this area. IBM, for example, had grad- ware capabilities alone.5

IEEE Annals of the of Computing 1058-6180/02/$17.00 © 2002 IEEE 5 1960s’ Software as Concept, Service, Product

Software’s other implicit definition at that As real-time systems expert Robert V. Head time, broader but equally unfamiliar today, aptly observed the next year, “The term ‘soft- included not only systems tools but also any ware’... is subject to redefinition from time to combination of tools, applications, and servic- time and to varying degrees of individual inter- es purchased from an outside vendor. This pretation.” His own definition included ambiguity was captured at the 1963 meeting of Fred Gruenberger’s informal “Rand Sympo- not only the series of program instructions need- sium.” The symposium took place every year ed to direct the computer to do a particular job, immediately before the Joint Computer Con- but also ... the entire of systems analysis ference, giving an invited crowd of computing and design, programming, testing and imple- and data processing luminaries a chance to dis- mentation, as well as the documentation that cuss the issues of the day. When the possibility accompanies this process.8 of making software production an activity was raised, participants discovered that Not all “software” was programs, and not all they lacked a common definition of software. programs were “software.” According to some While one participant suggested that “many of of these definitions, at least, a given piece of us refer to software as programs to be used by code might have been considered software if ,” Frank Wagner, recently moved obtained from another firm, but simply a pro- from to fledgling gram if written in-house. Yet a contract for an software company , countered that external group to perform programming work of any variety,—even testing or documenta- I always use the term to mean anything that isn’t tion—might have been considered a software clearly hardware or engineering design.... not project. Because of these conceptual shifts, con- only applications programming, but the writing temporary readers may easily misunderstand of specification for programs, the giving of advice the intended meaning of many sources from to people who might want to use computers, the the 1960s. Early discussions of problems in the installation manuals, etc.6 “production of software,” or the “,” were invariably referring to some- Wagner was a prominent member of the south- thing other than what we would today under- ern California computing community, having stand by these terms. For example, much early played important roles in both the SHARE user discussion of a “crisis” in software or of software group and the Association for Computing development problems was actually concerned Machinery. Bob Patrick, a well-known Califor- entirely with systems software.9 Likewise, com- nia computer consultant, then recently depart- plaints about the rising proportion of computer ed from leading computer services firm CEIR costs attributed to software, or the problems of (formerly Council for Economic and Industrial software project management, generally Research), took issue with this definition. What referred to the hidden costs of systems software he challenged was not the inclusion of non- bundled with , not to the programming services but of application pro- ever-present cost of in-house application pro- grams, saying “I don’t believe that application gramming. In contrast, estimates of the soft- packages are software,” regardless of who devel- ware industry’s size and future growth usually oped them.6 referred to the entire market for separately pur- A very broad definition of software was chased computer-related services and code.10 endorsed as late as 1967 by EDP Analyzer, Richard Canning’s authoritative newsletter. Toward the packaged application When it repeated a claim by Walter F. Bauer of Those, like Bauer, who favored a broad defi- Informatics that the software market held nition of software essentially used the term to by independents (companies that did not make describe any externally provided code or services. computers) had risen from 1 percent in 1960 to 3 This conflation of independently produced pro- percent in 1965 and would reach 10 percent by grams and independently offered programming, 1970, it also quoted Bauer’s software definition: analysis, and advice made considerable sense “systems analysis and design, programming and during the early 1960s. There was, as yet, no such computer-based services, accomplished by users, thing as a shrink-wrapped application package. computer manufacturers and others.” Bauer Most application programs were written in- argued that independent producers were rapidly house from scratch, although firms increasingly shifting focus from systems software toward relied on operating systems and programming applications programs, which he estimated tools from manufacturers. Early interest in pack- would rise to about 50 percent of their revenues.7 ages stemmed from the reasonable, generally

6 IEEE Annals of the held idea that it might be more effective to take acquiring expertise, and finding themselves able an existing program and modify it than to write to complete additional projects better and cheap- an entirely new one. Computer manufacturers er. After a certain point, custom applications with supplied such “canned” applications free with recycled code gradually evolved, into something their machines. A few of these packages enjoyed more like standard packages. It was, however, the considerable success, such as IBM’s early inte- appeal of getting a truly tailored solution rather grated package ’62 CFO, which business histori- than a generic program that had steered cus- an JoAnne Yates has shown eventually found tomers away from the computer vendor toward hundreds of users among medium-sized insur- the software company in the first place. ance companies. According to Yates, many of By the mid-1960s, application packages these companies would have been unlikely to were becoming better established. EDP Analyzer order the relatively inexpensive IBM 1401 com- reported that puter it ran on had the package not lifted most of the programming burden from their data- The change in interest in application packages processing staff.11 became evident [in 1965]. While some interest Many of the earliest software firms relied on had existed previously, it appeared to be quite governmental, particularly military, contracts. local. The arrival of the general purpose inven- The importance of RAND Corporation spin-off tory forecasting packages may have been the trig- SDC to computing’s early history and to the gering influence for the growing popularity of SAGE project is well known. Martin Campbell- packages.13 Kelly has suggested that large firms producing highly complex systems of this kind satisfied a Inventory management was a most popular market almost entirely distinct from the needs application for data-processing departments of mainstream business for low-cost application looking to do more than simply automate cleri- packages. Many offered computer services as cal jobs such as payroll. The operations research part of a broader range of scientific or technical aspects of this task, however, posed considerable services. The Planning Research Corporation challenges to corporate programming staffs.13 (PRC), for example, was founded in 1953 by The first independently produced programs three scientists, but a stream of defense contracts to be licensed as standard packages apparently helped it grow to 507 professional employees by were systems software such as file management 1965, more than half of whom worked on com- and report generation utilities. These filled nich- puter-related issues. (Despite this, the firm had es left empty by the computer manufacturers’ yet to install a single computer; it rented com- own software. Perhaps the most important early puter time as needed.)12 independent software package was Informatics’ The largest software companies (in this Mark IV file management system, which like broad sense) included PRC, the Computer many early packages, evolved from work begun Sciences Corporation (CSC), CEIR, Computer as a custom development project. First offered Applications Inc., and Informatics. These firms in 1967, it helped propel Informatics to become were of very different character from a leading 1970s’ software firm. Informatics, as and other firms that spring to mind when soft- we have seen, had a much broader view of soft- ware is mentioned today. Their work was closer ware and continued to derive most of its rev- to the consulting and services tasks today enue from custom programming and consulting undertaken by accounting firms, independent work. As the 1960s closed, Bauer suggested that, specialists like AMS and EDS and computer despite earlier optimism, the total market for vendors like IBM. Then, as now, small and flex- packaged software constituted only about 10 ible companies satisfied most of the demand percent of that for contract development work.14 for computer services. Barriers to entry were In 1968, when Business Automation magazine low, and stock options lured skilled program- interviewed him, Head was already established mers to the start-up firms. A spate of initial as a leading software expert. He had worked on public offerings around 1966 led investors to the pioneering ERMA (electronic recording pour money into the field, which was soon method of accounting) bank automation and crowded by an estimated 2,000 companies. SABRE airline reservation systems in a career that Independent software companies did not gen- had already involved working for GE and IBM, erally attempt to compete head-to-head by sell- in the banking industry, and in senior IT-related ing their own skeleton application code for positions for consulting firms Touche, Ross, companies to adapt. They did, however, under- Bailey & Smart and CSC. Head had just formed take similar projects for a number of different cus- his own company, the Software Resources tomers, building a of reused routines, Corporation, which he discusses elsewhere in

January–March 2002 7 1960s’ Software as Concept, Service, Product

this issue. He suggested that “a sort of software gies coupled with “application compilers” able explosion” had occurred in 1967, caused by a to go straight from requirements to code. shift toward more complex management infor- Within two years, he expected these systems to mation system (MIS) applications, a change that let line managers write their own applications demanded a level of programming and system as needed via teletype. This would facilitate design skills rarely present in corporate data- construction of an MIS and eliminate tradi- processing departments. Head, who, given the tional application programming, while giving breadth of his own work experience, was well data-processing specialists “a far higher status placed to know, suggested that “at this point vir- in the total managerial hierarchy.” Skepticism, tually no one is making a substantial profit on he insisted, was “no longer possible.”17 packaged software, although the potential seems There was thus no absolute line between sys- very great.”15 tems software and applications programs, and The assumption that hardware and operating neither was it clear that the former would invari- systems should be standard for all users while ably come from the hardware manufacturer and should be altered for differ- the latter from in-house or independent efforts. ent industries was not yet hard and fast. Head Nor was it clear that most companies seeking believed that hardware, programming languages, access to an externally produced program would and operating systems would prove more efficient run it on their own hardware or operate it using when adapted to industry sectors such as banking their own personnel. From the earliest days of or retailing. Odd as this seems given present expe- computing, firms such as payroll-processing rience, we must recall that the construction of giant ADP had sold bundles of services based on early real-time applications such as SABRE had proprietary software. From the mid-1960s, facil- required the building of specially designed oper- ities management operations such as Ross Perot’s ating systems—in this case, by a collaboration EDS were eager to operate computer installations between IBM and American Airlines. under contract. In some cases, this included pro- Such operating systems were closely opti- gramming services and standardized packages. mized for the programs they would run along- As interest in networking and remote access to side. The first for the IBM computers increased, many expected these mod- 701/704 was produced by the SHARE user con- els to become the norm. Time-sharing, a means sortium. Some firms tackling real-time business of giving several users real-time access to a single applications coded their own operating sys- computer, was viewed as the basis on which tems. Even when firms used vendor-supplied “computer utilities” would be built. Under this operating systems, the operating system had model, thousands of users would subscribe to often been tweaked or patched by individual giant networks, using terminals to access hard- users. Some companies even tried to write their ware and software running on remote comput- own operating systems for System/360 com- ers. Well over a hundred companies rushed to puters, often in conjunction with a plan to enter the time-sharing business.18 build an integrated, online MIS.16 Time-sharing services were initially popular The line between with scientists and engineers, who liked the con- and application software was also unclear. In a venience of interactive computers. The services 1965 address to a combined meeting of San offered libraries of useful programs and subrou- Diego data-processing societies, Paul H. tines to with calculations. As the market Rosenthal of CSC suggested that application became crowded toward the end of the 1960s, packages were a blind alley: and time-sharing companies targeted the poten- tially larger administrative computing market, There is actually no extensive production being they sought to differentiate themselves by offer- done today utilizing application packages .… ing applications packages rather than just com- These types of packages have not been widely puter time. A firm would effectively lease accepted, and outside of some very selected applications as a package, with computer hard- industries or functional areas are not considered ware, software, and services thrown in. GE (the by most people to be the answer to reducing the market leader) offered a package to banks; BBN cost of applications programming.17 (a computer services firm now best known for its contract work on Arpanet), a package for archi- Citing the success of generalized man- tects; and Univac had ambitious plans to build agement systems and specialized language com- nationwide networks to serve specific industries. pilers in the scientific field, he suggested that Its integrated package of billing, accounting, the long-term answer was not “packages as they inventory, and other operations was offered ini- are currently constructed” but new methodolo- tially to the wholesale wine and liquor industry.

8 IEEE Annals of the History of Computing While this was ultimately not a sizable market, corporate application programming as then continued attention to the concept left managers known. (Earlier technologies of this kind includ- uncertain whether the application packages of ed “automatic coding,” high-level languages, the 1970s would be installed on their own com- and decision tables). The data-processing depart- puters or rented through time-sharing.19 ment’s entire analysis, programming, and main- tenance could, he claimed, be replaced User views on software packages with three people: a data-processing coordina- How did the corporate data-processing man- tor, a software/applications analyst to choose agers of the 1960s feel about packaged applica- packages, and a single package-modification tion software? A survey of the magazines and to configure them. This enthusi- journals most closely associated with this com- asm apparently reflected a utopian dream rather munity (Business Automation, Datamation, and than hard-won experience.21 EDP Analyzer) offers hints, and several impor- The programming of administrative appli- tant articles from these sources are discussed cations remained an activity performed largely below. From 1967 onward, a small but growing by a company’s own data-processing staff. In number of subscribers could turn to the 1969, the Wall Street Journal surveyed almost International directories 800 executives with responsibility for comput- (discussed elsewhere in this issue). But the most er procurement and found that 44 percent important evidence is negative—akin to claimed all programs used were written in- Sherlock Holmes’ realization that a dog had house; another 38 percent said that many were. failed to bark during the night. While a good Even among the largest companies (the sector number of articles discussed the software indus- identified by the survey as the primary users of try’s boom, and time-sharing’s revolutionary what it called “Programming [Software] potential, there were few nuts-and-bolts treat- Services”), just 45 percent of respondents to ments of how to purchase or evaluate applica- this question reported any purchase of outside tion software packages. Throughout the decade, programming (a category including both pack- packages remained relatively unimportant as a ages and services). These same companies relied source of application programs, or even as part more heavily on bundled programs supplied by of the overall software and services market. the manufacturers of their computers, utilized Like anyone keeping one eye on the busi- by 80 percent of responding firms.22 ness press and one on the stock market, data- Any software acquisition—packaged or cus- processing managers were exposed to an tom; system or application—ultimately became a enthusiasm for the software industry, which set of make-or-buy questions. No package entire- had quite outstripped its accomplishments. As ly removed the need to “make” some elements Head observed in 1970, at the end of the of the system, but it was sometimes possible to first market bubble in software stocks, reduce this burden. Unlike the hopeful schedules and fuzzy definitions associated with in-house For a long time now, the software industry has development, packaged software was a known been a prime concept of interest on the part of quantity. Its costs and capabilities could be meas- investors, with resultant enormously inflated stock ured. In addition, many companies found their values. Price-earnings ratios of eighty to one seem programming and analysis teams chronically normative, and infinity to one not unusual, in the overworked, with a large backlog of urgent tasks. case of numerous companies that have never Packaged software promised to alleviate this. turned a profit. There has been almost a compul- During the second half of the 1960s, most sion to ‘go public’ on the part of many marginal data-processing departments were engaged in firms, and millions of dollars have been obtained a transition from second-generation machines from the public through such offerings.20 (such as the IBM 1401 or 7040 series) to third- generation machines such as those in the One remarkably enthusiastic statement came System/360 range. In the process, they used in 1970, from Pabst Brewing’s data-processing complex operating systems, large disk drives, director. The author claimed that the new gen- and terminals for the first time. Since it would eration of packaged applications was much take a major programming effort to reimple- superior to the old, manufacturer-supplied pack- ment their existing systems for this new envi- ages. As a result, “Low supply and high demand ronment, the transition to third-generation for in-house systems and programming talent systems was an obvious moment at which to may not be as long lived as many think.” He consider the shift to packaged applications. seized on packaged application software as one This approach held the additional promise in a long line of technologies destined to end that different applications from the same vendor

January–March 2002 9 1960s’ Software as Concept, Service, Product

might be easier to integrate than in-house pack- • documentation (separate descriptions required ages that had never been properly coordinated. on each system level: program, operations, CSC, for example, offered a number of general- and user), ized applications to banks to handle common • installation support (must be agreed in applications such as payroll, personnel records, advance; may include file conversion and stockholder records, and general ledger. Such training assistance), and applications packages were somewhat general- • maintenance (“the acquisition terms for a ized—modification and maintenance would still package should include some assurance of be necessary—but more flexible than a typical error-free operation for a reasonable period program. While CSC charged each customer of time”).25 when new options and capabilities were deliv- ered, bug fixes were free of charge and only a The list demonstrates that software acquisition modest fee was levied for the basic updates to remained complex. The importance of the pro- meet new legal regulations. As a result, it was gramming language used, program documenta- realistic to hope that use of a package would dra- tion, and program expandability illustrate that an matically cut the cost of program maintenance.23 application program was subject to extensive cus- In 1968, Head used a Datamation article to tomization. Programming techniques of the era give data-processing managers what was then made it hard to produce a program that would their most detailed, practical guide to packaged run on -powerful hardware while taking software acquisition. He identified application- advantage of larger memories, disk drives, and the package sources as manufacturers (in whose like when available. Similarly, unnecessary fea- products he had little faith), user groups (some of tures might slow down a program and make it which maintained extensive lists of user-submit- harder to maintain. As Head warned in a later ted software), and software companies. Whatever guide to the same topic, computer staff might not the source, a package would likely require sub- prove a package’s most reliable judges. While a stantial modification. A well-designed commer- manager was to some extent at the mercy of their cial package could easily pay for itself by lowering judgments, he or she must also remember that the overall project cost, especially if this also led their “professional pride” might lead them away to reduced maintenance costs. Head suggested from cost-benefit calculations toward the convic- that commercial applications generally cost tion that they could produce a better package. between $2,000 and $20,000, representing 10 or Said Head: “Experience has shown that an outside 20 percent of the original development cost. package, no matter how estimable, can be torn While acknowledging that “[t]here may still be asunder by an astute technician bent on ferreting some resistance on the part of data processing out and magnifying all possible deficiencies.”26 managers accustomed to obtaining software ‘free’ In a short 1971 book, Head surveyed the state from the computer manufacturers,” he argued of the software industry at that time. He found that such reliance might prove a false economy.24 that some packaged application programs were Head’s advice for the evaluation of software supplied in Cobol, for compilation on a range of packages was more pragmatic. He suggested that machines from different manufacturers. Others data-processing managers assign their own were more closely tied to particular machines weights to each of the following factors and score and operating systems. His analysis of the mar- packages under consideration accordingly: ket for payroll programs, a leading application of the period, was particularly informative. Most • package cost (including indirect costs for such programs cost about $20,000. Two of the modifications, training, installation, con- most successful appeared to have sold about 75 version, running, maintenance, and so on), copies each. One of these, the CSC payroll sys- • package quality, tem, was slow and required a large, 65-Kbyte • design features (file organization, control memory. It was relatively easy to use and install, and audit features, programming tech- reducing the need for skilled computer person- niques, flexibility, and so on), nel. However, this usability came at the expense • generality (“among the most important of flexibility—reports not needed for a particular package criteria”), payroll run, for example, could not be turned • expandability (although an overgeneralized off.27 Such programs continued to appeal prima- package might prove inefficient), rily to companies without the staff, money, or • operational status (how extensively used time to develop a better system of their own. and bug-free a package is), IBM’s decision, in 1969, to unbundle its soft- • equipment configuration needed to run it, ware from its hardware is usually considered • programming language it is written in, crucial to the establishment of a viable market

10 IEEE Annals of the History of Computing for packaged application software.28 While of software, based on experience of the shrink- undoubtedly important, this event appears to wrapped PC-software market of the 1980s, will have reinforced shifts already under way, and not help here. Indeed, such a conception its effects were not felt immediately. In the early might instead mislead the contemporary ana- 1970s, most application software was still pro- lyst as well as the historian. Into the 1980s, duced by user organizations; most of the pack- Informatics continued to derive more of its ages used continued to come from IBM and revenues from custom development and con- other computer manufacturers. While the sulting services than from packaged software entrenchment of the System/360 architecture sales. Even today, the shrink-wrapped model and the acceptance of standardized high-level applies only to a small part of the business soft- languages (most particularly Cobol) provided a ware world, and many analysts believe it to be larger potential market than ever before, this in terminal decline.31 Installing complex pack- market remained largely untapped. In his 1971 ages such as those offered by SAP, for example, book, Head suggested that although “the poten- requires a tremendous configuration and cus- tially great profits” were associated with a tomization effort. Moreover, the proliferation of programming tools such as , object potential market for these systems in the tens of technology, and rapid application develop- thousands ... even highly successful packages have ment systems has been responsible for the pro- at this point sold only in the neighborhood of duction of more custom application software, fifty systems with a mere handful of outstanding not less. success stories claiming sales in the hundreds.29 A better understanding of the origins of pack- aged software enriches our understanding of its In the subsequent decade, firms such as likely future in at least two ways. First, it reminds Management Science of America (MSA) and the us that different models coexist and that pack- University Computing Corporation (UCC) final- aged software, custom programming, and con- ly managed to nurture a market for packaged sulting services are complementary. Services are applications software that grew steadily and usually inseparable, most programming effort is proved profitable. (The largest and most suc- still devoted to custom applications, and the cessful firms, however, still tended to specialize transition from custom development to package in systems software.) Change in computer usage remains a continuum. Second, a better under- lagged expectations, as firms ported their appli- standing teaches us that change is invariably cations from one machine to another, using slow and incomplete and that no new model— emulation and continual patching to support be it time-sharing in the 1960s or Internet-based ancient application code on new hardware. application service providers today—can be The challenge in application program acqui- expected to fill all niches or to sweep away exist- sition has always been to optimize the fit of a sys- ing technologies overnight. Had the technolo- tem to the requirements of a particular business gy investors and entrepreneurs of the past few while minimizing the amount of specially writ- years paid more attention to history, their dol- ten code and lowering the cost of ongoing main- lars and their dreams might not have been lost tenance. Packaged applications ultimately played to it with quite such rapidity.32 an important part in reconciling these goals, as part of a broader repertoire of complementary References and notes techniques including generalized I/O routines, 1. On the early history of data processing and the high-level languages, role of programming within it, see T. Haigh, “The and design, CASE (computer-aided software engi- Chromium-Plated Tabulator: Institutionalizing an neering) and RAD (rapid application devel- Electronic Revolution, 1954–1958,” IEEE Annals of opment) tools, management systems, the History of Computing (hereafter called Annals), and . (The latter eventually gave rise to vol. 23, no. 4, Oct.–Dec. 2001, pp. 2-31. object orientation.) These techniques were sup- 2. For discussion of the earliest known use of “soft- ported by substantial improvements in operating ware” in the context of computers, see F.R. system technology and computer architecture. Shapiro, “Origin of the Term Software: Evidence Meanwhile, software packages and services from the JSTOR Electronic Archive,” Annals, vol. remained closely intertwined in how they were 22, no. 2, Apr.–June 2000, pp. 69-71. Shapiro bundled, sold, and used.30 discusses a 1958 article by mathematician John We cannot hope to understand software’s W. Tukey, who used the term to describe early history without understanding the work automatic programming aids (such as compilers done inside user organizations to adapt and and assemblers) of the type provided by comput- supplement packages. A narrow understanding er manufacturers.

January–March 2002 11 1960s’ Software as Concept, Service, Product

3. Honeywell, “A Few Quick Facts on Software,” turnkey systems, and educational services as the Business Automation, vol. 7, no. 1, Jan. 1962, pp. 11 components of the “software service” market. 16-17. See F.A. Frank, “Software Services: An Outside 4. “Software on the Couch,” Datamation, vol. 7, no. Outlook,” Business Automation, vol. 16, no. 11, 11, Nov. 1961, pp. 23-24. A similar definition of Nov. 1969, pp. 55-61. software as a “programming system package” 11. J. Yates, “Application Software for Insurance in the can be seen in H.R.J. Grosch, “Software in Sick- 1960s and Early 1970s,” Business and Economic ness and Health,” Datamation, vol. 7, no. 7, July History, vol. 24, no. 1, Fall 1995, pp. 124-126. 1961, pp. 32-33. See also “The Master Plan for 12. M. Campbell-Kelly, “Development and Structure Kludge Software,” Datamation, vol. 8, no. 7, July of the International Software Industry,” Business 1962, pp. 41-42. and Economic History, vol. 24, no. 2, Winter 1995, 5. A. Opler, “Measurement of Software Characteris- pp. 73-110. On the Planning Research Corp., see tics: Evaluation Techniques,” Datamation, vol. 10, “The Changing Software Market,” EDP Analyzer, no. 7, July 1964, pp. 27-30. vol. 4, no. 7, July 1966. 6. F. Gruenberger, “Rand Symposium 6,” in Rand 13. “Application Packages: Coming into Their Own,” Symposia Collection (CBI 78), EDP Analyzer, vol. 5, no. 7, July 1967. Institute, Univ. of Minnesota, Minneapolis (here- 14. For the Mark IV, see M. Campbell-Kelly, “Devel- after, CBI), 1963. This discussion in many ways opment and Structure of the International Soft- precipitated the much better known NATO Con- ware Industry,” Business and Economic History, ference on held in 1968 vol. 24, no. 2, Winter 1995, pp. 73-110; J.A. Post- and 1969. Participants discussed parallels ley, “Mark IV: Evolution of the Software Product, between hardware and software, the difficulty of a Memoir,” Annals, vol. 20, no. 1, Jan.-Mar. managing programmers, the importance of liter- 1998, pp. 43-50; and R.L. Forman, Fulfilling the ary style in programming, and the applicability of Computer’s Promise: The History of Informatics, engineering techniques to software. Participants 1962–1982, Informatics General Corp., 1984. included Barry Gordon, Bob Patrick, Richard 15. “The Software Explosion,” Business Automation, Hamming, Francis V. Wagner, and M.D. McIllroy. vol. 15, no. 9, Sept. 1968, pp. 24-29. Head went 7. Bauer offered his definition during a Fall 1965 on to help found SMIS, the Soc. of Management address to the chapter of the ACM. Information Systems (known today as SIM), and See “Independent Software Companies,” EDP spent the latter part of his career in senior IT posi- Analyzer, vol. 5, no. 11, Nov. 1967. tions with the US federal government and as a 8. For Head’s discussion and his own definition, see consultant in the same area. R.V. Head and E.F. Linick, “Software Package 16. Head earlier detailed his concept of industry- Acquisition,” Datamation, vol. 14, no. 10, Oct. specific computing platforms in R.V. Head, “Old 1968, pp. 22-27. Myths and New Realities,” Datamation, vol. 13, 9. See particularly H.R.J. Grosch, “Software in Sick- no. 9, Sept. 1967, pp. 26-29. For an example of a ness and Health,” Datamation, vol. 7, no. 7, July homebrew 360 operating system, see L.F. Zaino, 1961, pp. 32-33, and “Software on the Couch,” “How Does an Operating System Work?,” Datamation, vol. 7, no. 11, Nov. 1961, pp. 23-24. Systems & Procedures J., vol. 19, no. 1, Jan./Feb. One of the earliest articles to imply our modern 1968, pp. 20-23. On the history of MIS and its definition of “software costs” as including in- relationship to the promotion of third-generation house development of application programs is computers, see T. Haigh, “Inventing Information H.A. Lustig, “The High Cost of Software,” Business Systems: The Systems Men and the Computer, Automation, vol. 13, no. 5, May 1966, pp. 37-38. 1950–1968,” Business History Rev., vol. 75, no. 1, Conversely, most of the crises attributed to data Spring 2001, pp. 15-61. processing in this period had little direct relation 17. P.H. Rosenthal, “Future Programming of to software—see D.R. Daniel, “Management Computer Applications,” Systems & Procedures J., Information Crisis,” Harvard Business Rev., vol. 39, vol. 18, no. 1, Jan./Feb. 1967. no. 5, Sept./Oct. 1961, pp. 111-121, or A.E. 18. Although the computer utility concept is general- Keller, “Crisis In Machine Accounting,” Manage- ly believed to have originated with M. ment and Business Automation, vol. 5, no. 6, June Greenberger, “The Computers of Tomorrow,” 1961, pp. 30-31, for examples. The Atlantic Monthly, May 1964, pp. 63-67, it 10. In perhaps an extreme example, one author pre- goes back at least five years earlier to the 1959 sented service bureaus, contract programming, conference presentation published as A.O. Mann, consulting services, personnel services, package “A Publicly Regulated System of Management suppliers, proprietary software services, facilities Control Services,” in Management Control management, “dedicated applications Systems, D.G. Malcolm and A.J. Rowe, eds., John companies,” time-sharing services, providers of Wiley & Sons, New York, 1960, pp. 245-263. For

12 IEEE Annals of the History of Computing an early discussion of its relevance to data Shakeout: Winners, Losers, and Survivors, John processing, see R.E. Sprague, “The Information Wiley & Sons, New York, 1984, pp. 240-263. Utilities,” Business Automation, vol. 12, no. 3, 31. Some software remains in the public domain, Mar. 1965, pp. 42-47. particularly in the Linux area where some compa- 19. E.H. Menkhaus, “Time Sharing is Everybody’s nies attempt to give away the code while charg- Thing,” Business Automation, vol. 16, no. 9, Sept. ing for support and services. In the and 1969, pp. 26-35 and p. 38; and R.V. Head, A enterprise software markets, software is leased Guide to Packaged Systems, Wiley-Interscience, annually rather than purchased and is often New York, 1971. Others believed that the crucial priced together with service and support advantage of the computer utility would come contracts. Current industry opinion suggests a with the automatic interchange of orders, general move toward the use of online software payments, and other information between firms (via application service providers—the time- that would follow once different businesses used sharing systems of the 21st century) and the shared computers to hold their information. annual leasing of desktop application software. These feelings were only reinforced in 1970, as a 32. On the recent travails of the ASP industry, see . collapse in the stock market bubble for software Koch, “Boy, That Was Fast!,” CIO Magazine, 15 and time-sharing firms caused scores of start-up Nov. 2000, http://www.cio.com/archive/111500/ firms to begin unraveling. As Business Automation boy.html [current as of 16 Jan. 2002]. For a reported at the time, the future of time-sharing dismissal of the concept as “rehashed time-sharing was seen to lie not in the general-purpose com- service bureaus,” see B. Lewis, “Rather than Focus- puter utility but in the provision of “industry- ing on Best Technologies, Let’s Look at and Learn oriented software on a national basis …. Time from the Year’s Worst,” Infoworld, 26 Jan. 2000, sharing has become an information transfer busi- http://staging.infoworld.com/articles/op/xml/01/ ness and not a computation or calculation busi- 01/29/010129oplewis.xml [current as of 16 Jan. ness.” See E.J. Menkhaus, “Time Sharing: More 2002]. Glitter than Gold,” Business Automation, vol. 17, no. 11, Nov. 1970, pp. 36-42. Acknowledgments 20. R.V. Head, “Twelve Crises—Comments on the The author would like to thank Burton Grad Future of the Software Industry,” Datamation, and Robert V. Head for their encouragement vol. 16, no. 3, Mar. 1970, pp. 124-126. and contributions and the referees for their 21. J.E. Hackney and N.L. Paul, “The Wheel Exists,” J. comments on the draft version. This research Systems Management, vol. 21, no. 5, Nov. 1970, has been supported by fellowships from the pp. 40-41. Charles Babbage Institute of the University of 22. Wall Street J., “Management and the Computer: A Minnesota and the IEEE History Center. Wall Street Journal Study of the Management Men Responsible for their Companies’ Purchases of Com- Thomas Haigh is teaching at puter Equipments and Services,” Data Processing Colby College and completing a Management Assoc. Records (CBI 88), CBI, 1969. PhD in the history and sociolo- 23. “Application Packages: Coming into Their Own,” gy of science at the University of EDP Analyzer, vol. 5, no. 7, July 1967. Pennsylvania. His dissertation, 24. R.V. Head and E.F. Linick, “Software Package Technology, Information, and Acquisition,” Datamation, vol. 14, no. 10, Oct. Power: Administrative Technicians 1968, pp. 22-27. in the American Corporation, 25. Ibid., pp. 25-26. 1917–1975, is the first full-length synthetic history of 26. R.V. Head, A Guide to Packaged Systems, Wiley- the corporate use and management of information Interscience, New York, 1971. technology during that period. He holds a BSc and 27. Ibid., pp. 41-43. MEng in systems integration from the University of 28. Unbundling has been the subject of considerable Manchester, UK. Fellowships include a Fulbright historical inquiry, including several articles in this Award, the IEEE Life Member Fellowship in Electrical issue. For contemporary reaction, see A. Dratell, History, and the Tomash Fellowship from the Charles “Unbundling: The User Will Pay for the Works,” Babbage Institute. Business Automation, vol. 16, no. 8, Aug. 1969, pp. 36-41. Readers may contact Thomas Haigh at tdhaigh@ 29. R.V. Head, A Guide to Packaged Systems, Wiley colby.edu; http://www.tomandmaria.com/tom. Interscience, New York, 1971, p. 66. 30. For a critical discussion of the largest mainframe For further information on this or any other com- software companies at the start of the 1980s, see puting topic, please visit our Digital Library at S.T. McClellan, The Coming Computer Industry http://computer.org/publications/dlib.

January–March 2002 13