Notes for Martin Campbell-Kelly From Airline Reservations to Sonic The Hedgehog: A History of the Software Industry

Key concepts: bedroom coder, computer utility, disruptive technology, flowcharting, killer app hypothesis, legacy software, network effects, systems integrator, teleprocessing, user group, waterfall technique.

Role of industry analysts in organizing and creating historical sources.

Related theorists: John Backus, Bogost, John F. Jacobs, Licklider, Montfort, Turkle, JoAnne Yates.

Strategic chooses to study three early eras and leave off speculation about really anything after the Internet. In is in this zone that we have to go. Also note this book is about the software industry moreso than software. This book provides great resources for designing a philosophy of computing course to delve into history of the software industry as one aspect.

Preface
(vii) As is true of the plumbing beneath the city streets, it takes an effort of will to bring software to conscious attention. Indeed, one of the great fascinations of software is its invisibility and intangibility.
(ix) Before there were ATMs, there were airline reservation systems. From the very beginning of the commercial use of computers, airline reservations was
the path-breaking application.

This time bound statement applies to the implied ages of the writer writing this, and measures from the milieu of future readers, experiencers of this virtual reality that includes this book or representations of it.

(x) And software, it should be noted, is ubiquitous rather than all-embracing; it is a part of life, but only a small part.


Acknowledgements


1
The Software Industry

Chapter 1 is prefaced with an American Airlines advertisement touting the SABRE airline reservation system as inaugurating modern commercial software; thus an aspect of studying software includes its advertising rhetoric along with other sources.

(1) In January 1952, half a century ago, Fortune magazine ran an article titled “Office Robots.”
(1) In November 1966 Business Week carried a report titled “Software Gap—A Growing Crisis for Computers.” . . . Not until September 1980 did Business Week carry a special report on the software business, its first in-depth look at the industry since 1966.
(2)
Business Week's next special report on the software industry appeared in February 1984, and it could not have been more different in tone from the 198 article. In the intervening 31/2 years, the rise of the personal computer had made the world at large aware of software.

Understanding the Software Industry
(3) The aim of this book is to explain the software industry by a historical account of its evolution. Because no simple one-dimensional framework is adequate for this purpose, I use three main vectors of explanation.

Periodization, Sectorization, and Capabilities

Is it highly customized and enormously expense and challenging to maintain, in the middle, or virtually self sufficient? These are the divisions of software reality Campbell-Kelly proposes, like early philosophers of computing defined reality by primary, secondary, and tertiary storage, dividing into dynamic random access memory units, and not, further dividing into deterministic (rapid, random access) or nondeterministic but potentially infinite.

After noting absence of popular press interest in the software industry before 1980, divides into three sectors: contracting, corporate, and mass-market products.

(3) The software industry can be divided into three sectors: software contracting, corporate software products, and mass-market software products.

Software Contractors

Explain why debatable whether any sense in trying to think about socialistic, government owned entities doing any valuable if not crucial work for capitalism, its putative successor. Relate to recent Deleuze reading.

(5) The defining event for the software contracting industry came in 1956, when the US-government-owned RAND Corporation created the Systems Development Corporation (SDC) to develop the computer programs for the huge SAGE air defense project.

(5) The business model consciously or unconsciously adopted by custom programming firms was that of an engineering or construction contractor.

Corporate Software Products

First software products are packaged programs 1965 Autoflow and 1967 Mark IV.

(6) Two packaged programs, Applied Data Research's Autoflow and Informatics' Mark IV (announced in 1965 and 1967, respectively), are generally agreed to be, if not the first, certainly the most influential of the early software products.
(6-7) Finally, as with all capital goods, pre- and after-sale support was needed to establish a long-term relationship with the customer. In the case of software products, this took the forms of product customization, user training, and regular upgrades. These services turned out to be unexpected sources of income for which the pioneers of the industry had not initially planned.

Mass-Market Software Products
(7) The industry really took off in 1979-80, with the arrival of mass-market software such as Software Arts' VisiCalc spreadsheet and MicroPro's WordStar word processor.

No gaming at work or anywhere but home and school.

(7) Closely related to the personal computer software industry (whose products were used in corporations and homes) was the recreational software industry (whose products were used exclusively in domestic and learning environments).

(7) The personal computer software products industry was completely disjoint from the corporate software products industry.
(8) The critical capabilities developed by the personal computer software firms included exploitation of scale, mass marketing, and ease of use.

Compare to short history given in piece about print in future software help.

(8) So that personal computer software products could be used by many thousands of customers without any after-sale support, programs had to have intuitive interfaces and had to require no customization. This again required the development of a set of skills different from that of corporate software makers, who could rely on training courses and third parties to install and customize software.

Similar to making a big deal about a particular author or text if you are a humanities scholar.

(9) Microsoft dominates about 10 percent of the book, just as it dominates 10 percent of the software industry. If this book serves no other purpose, I hope it will at least provide a corrective to the common misconception that Microsoft is the center of the software universe around which all else revolves.

Scope: Exclusions and Limitations
(9) The most severe limitation is that the book has a strong US focus.

Predicts free, open source software will be next important scholarly topic beyond book cutoff of 1995. Scholarship bends toward FOSS on account of its epistemological transparency, and potential therefore for complete knowledge of everything within one's virtual realities.

(10-11) A second limitation of this book is the cutoff date of 1995. . . . I don't know what it is, but I bet there is something much more important going on right now than Java, Linux, or open-source software, and that it will be 2010 before it becomes fully apparent.
(11) In
Secrets of Software Success, Detley Hoch and his co-authors project the “Internet Era,” a new period in the development of the software industry, and suggest boundary dates of 1994 and 2008.

Numbers: Software Industry Statistics and Trade Associations

Rarity of public domain scientific information of the software industry; industry analysts provide most information.

(13) Table 1.1 presents data on the US software market in the period 1970-2000. I believe this is the only 30-year time series for the software industry in the public domain, and it appears here courtesy of the industry analyst INPUT.

(13) Table 1.2 gives the ADAPSO-INPUT industry statistics for the years 1966-1990. This is the only long-term time series for US software industry revenues in the public domain.
What is unknown, hidden in the reports never written that could still be recovered, recomputed? Imagine goals for many iterations of university coursework. Note Castells well articulates role of universities in primary research and thinking that goes into technology.

(17) Since 1991, ADAPSO's successor, ITAA, has no longer published annual surveys of the industry, primarily because the ITAA's board decided to suspend funding for the statistics program; once terminated, it was never resumed.
(20) In its lobbying activities, the BSA has used census data on the US software industry to demonstrate that the software products industry is a large sector of the US economy, is a major net exporter, and deserves anti-piracy legislation.
(20) Because the Census Bureau carries out much more exhaustive data collection than commercial industry analysts (who do not have the force of law to compel responses or the machinery to analyze them), the census statistics are probably more accurate, but they are not necessarily more useful in understanding the industry.

Because it is unlikely to recur now that we have moved into the Internet based industry paradigm.

(22) As table 1.5 shows dramatically, the United States dominates the world software industry, with a 70 percent market share in 1982 and 57 percent in 1990. . . . Without doubt, the packaged software industry was and remains an American phenomenon.

Sources

Because scope of software industry as a whole broader than individual firm, not heavy use of corporate archives but monographic studies, periodic literature, and analyst reports.

Note that online FOSS communities provide historically unprecedented access to the equivalent of corporate archives of a very large number of software projects. See statement below about the sole major corporate archive readily available for study and analysis.

(22) This book is perhaps the first attempt at writing a full-length history of the software industry broader than the study of an individual firm.
(23) Since the present book does not focus on any one firm, heavy use of corporate archives was not appropriate, nor indeed was it possible. Instead, it is based largely on monographic studies of the software industry, the periodical literature, and reports by industry analysts.
(23) In discussing the monographic literature, one should perhaps begin with the bad news. There are more books written about Microsoft and Bill Gates than about the rest of the industry put together.
(24) The best sources for the broader industrial scene are the reports resulting from government-sponsored software policy studies, a few academic monographs, and the publications of market-research organizations.
(25) The periodical literature of the software industry comprises, in order of usefulness, the trade press, general business periodicals, and newspapers. . . .
Datamation is the only periodical to span almost the entire history of the software industry.
(26) In the general business press, the best sources on the history of the software industry are
Business Week and Fortune.

SDC only major corporate archive available to scholars; makes interesting point that archivization usually does not consciously begin by corporations until contemplating their silver anniversary, a twist on answering question by Derrida about the time of the archive.

(26) Indeed, the only major corporate archive available to scholars is that of SDC, now housed by the Charles Babbage Institute at the University of Minnesota. The SDC Collection covers the period from 1956 to 1981, when SDC was acquired by the Burroughs Corporation. The impressive volume of information in the SDC collection hints at the riches that will become available when software companies organize their archives.
(26) The software industry came into being at about the same time as the industry analyst.

Subscriber-only reports scarce in public domain; Software History Center current locus of preservation and dissemination.

(27) Industry and market-research reports were produced for subscribers only (primarily software firms and large users), and few have migrated to the public domain. . . . Of this huge volume of material [catalog of INPUT reports], no more than 50 reports survive in the Library of Congress. . . . The Software History Center is dedicated to preserving the history of the software industry by ensuring that records of the companies, the individuals, and the events that shaped the industry's growth are preserved and made accessible to anyone seeking to understand how the industry evolved.


2
Origins of the Software Contractor, the 1950s

Chapter 2 begins with System Development Corporation advertisement of a Senior Computer Systems Specialist depicted as deeply contemplating computer programming. Fitting for fitting in programming practices for humans in their original and nostalgic memories being considered following consideration of Derrida. Adding on to older note about Deleuze trying to remember why it was originally made, which is fitting given recent review of Derrida, and further back Tanaka-Ishii and others puzzling about the relations between dreams of machine cognition and beliefs about human cognition, challenged further by constant potential for synaptogenesis and environmental, in the sense of everything outside transformation of living human brains, technogenesis.

(29) All the pioneers of computers, and all the firms that built the first commercial models, greatly underestimated how difficult it would be to get programs to work and how much code would be needed.

Deleuze n-1, perhaps sustaining automata (though perhaps incorrectly or suboptimally; there is debate over the soundness of Deleuzean concept of cellular automata operation).

(29) Only weeks after the first prototype laboratory computers sprang to life, it became clear that programs had a life of their own—they would take weeks or months to shake down, and they would forever need improvement and modification in response to the demands of a changing environment.

Early Sources of Software

Sharing programs was one of the original practices, so important to Stallman.

(29) In the first half of the 1950s, there were three main sources of software: users could write it for themselves, could obtain programs from a computer manufacturer, or could share programs among themselves.
(29) For example, IBM's first production computer, the 701, came with little more than a user's manual.
(30) Maintaining a large programming staff was a necessary and not disproportionate part of the cost of running a computer. The IBM 701, for example, rented at $15,000 a month, while a programmer earned $350 a month “at most.” A cohort of 30 or more programmers for a single mainframe was typical, and what users expected from the manufacturer was not so much programs as customer support and training.

IBM's Technical Computing Bureau
(30) IBM viewed the coding of programs as simply an additional step in its “application development” process. That process, which long predated IBM's entry into computers, consisted of the systematic analysis of the data processing task, the design of machine records and files, and the division of the task into machine runs.
(31) In the first year of operation, about 100 applications were developed for 71 customers. In addition to customers' application programs, the bureau developed about three dozen “utilities.” These included “bootstrap” routines for loading programs into memory from punched cards and routines for transferring data from cards and magnetic tapes and for printing.
(31) IBM's Technical Computing Bureau was a significant force in the tiny new world of programming in the United States, and its staff members were well placed to shape the nascent world of software.

SHARE

Software terminology and classification framework established by SHARE group, making user group as primary social interface.

(33) At a time when the cost of programming ran as high as $10 an instruction, dramatic savings could be achieved through cooperation. . . . In all, approximately 300 programs had been distributed to the membership. Besides organizing a library of programs, SHARE established a terminology and a classifactory framework for software, vestiges of which persist today.
(34) The user group remains the preeminent way for computer users to interact with one another and with computer manufacturers.

Programming Languages—FORTRAN and COBOL

Programming languages FORTRAN and COBOL improved productivity first by generating multiple machine instructions per line of code.

(34) The crucial breakthrough in programmer productivity was the development of programming languages in the second half of the 1950s. In a programming language, a single line of code could generate several machine instructions, improving productivity by a factor of 5 or 10.
(34) The most important early development in programming languages was FORTRAN for the IBM 704. The FORTRAN project took place in IBM's Technical Computing Bureau under the leadership of John
Backus, a 29-year-old mathematician and programmer.
(35) The responses from users was immediate and ecstatic. . . . Backus and his team refined the system in response to feedback from users and released a new version in 1959. FORTRAN II contained 50,000 lines of code and took 50 programmer-years to develop.

Network effects popularized FORTRAN more so than help from IBM.

(35) There was no conspiracy: FORTRAN was simply the first efficient and reliable programming language. What economists later called “network effectsmade FORTRAN a standard language with little or no help from IBM.

US DoD pushed industry to agree on standard business language, from which COBOL arose; FORTRAN and COBOL garnered two-thirds of application programming activity for next twenty years.

(35) In April 1959, the US Department of Defense convened a meeting of computer manufacturers and major computer users with the aim of agreeing on a standard business language.
(36) During 1962 and 1963, COBOL became the standard language for business applications, which it remained for the next quarter-century.
(36) The “twin peaks” of FORTRAN and COBOL were to account for two-thirds of the applications programming activity of the 1960s and the 1970s. . . . This was software's first example of “lock-in.”
(36) Largely through an accident of history, the System Development Corporation (SDC) was to play a major role in supplying programming manpower.

SDC, the “University of Programmers”
(37) The Valley Committee's report recommended that the entire air-defense system be upgraded, with improved interceptor aircraft, ground-to-air missiles, anti-aircraft artillery, comprehensive radar coverage, and automatic processing of data in computer-based command-and-control centers. The last of these requirements was to result in America's largest computer and software project.

SAGE example of rhetoric of Licklider human machine symbiosis.

(37) The new system was called SAGE, for Semi-Automatic Ground Environment. The term “semi-automatic” emphasized the interaction of man and machine: the computers would perform high-speed data processing, while humans would be responsible for high-level information processing.
(38) Whereas the Cape Cod system had been programmed as a one-of-a-kind system by a select group of engineers, for SAGE it would be necessary to use inexperienced programmers to develop a generalized program that could be reconfigured and rolled out to 20 or more sites. In the absence of a private-sector contractor willing to take on the programming challenge, the RAND Corporation, a nonprofit government-owned research organization, was given the task.
(39) It was necessary to recruit raw talent suitable for training in programming. The majority of applicants were men between the ages of 22 and 29, mostly college graduates in a wide variety of disciplines. (“Music teachers were particularly good subjects.”)
(39) By 1959, there were more than 700 programmers working on SAGE, and more than 1,400 people supporting them. This was reckoned to be half of the entire programming manpower of the United States.
(39) The SAGE program, when completed, contained more than a million instructions and was by far the largest program of its time.
(40) Direction centers came on stream at the rate of approximately one every 2 months, SDC supplying a program installation team of up to 100 programmers. Once the system was up and running, about eight technical personnel were permanently assigned, along with a few instructors to train military personnel in the use of the computer.
(41) By 1963, SDC had a staff of 4,300 and offices in seven locations in addition to the headquarters in Santa Monica, and its computer center—with two SAGE Q-7s, four other hefty mainframes, and several mid-size IBM 1401s—was the largest in the world.

SABRE, “The Kid's SAGE”

Challenge of real-time teleprocessing pushed limits of early commercial technology.

(41) However, real-time projects pushed the technology to its limit. They required a computer to respond instantaneously to external inputs and to process many transactions simultaneously, a requirement for which IBM used the term “teleprocessing.” Examples included airline reservations, bank automation, and retail systems.
(42) The first and classic civilian real-time project was the IBM-American Airlines SABRE airline reservation system. Though SABRE did not become fully operational until 1964, it was the outcome of more than 10 years of planning, technical assessment, and system building.
(44) Though SABRE was called “the kid's SAGE,” it was by far the largest civilian computerization project to date, involving 200 technical personnel for 5 years.
(44-45) The operational programs for SABRE contained half a million lines of code, but as much again was written for testing modules and simulating test environments for program debugging. Real-time software was much more difficult to construct than conventional programs.
(45) Analysts estimated that SABRE earned a 25 percent return on investment; the intangible benefits in terms of customer service and brand recognition may have been even greater.

System Integrators
(45) SAGE and SABRE were the leading examples of computerized “systems integration” projects in the defense and civilian sectors, respectively. These projects, and those that followed, gave rise to the
systems integrator, a new kind of company with the ability to engineer complex systems incorporating a range of electronic, computer, and software technologies.
(47) JOVIAL went on to be used by thousands of programmers in the “closed world” of military computing well into the 1980s, although it was scarcely used in the civilian sector.

Waterfall software development technique diffused into civil computing from military applications.

(47) Other software technologies developed for military applications did, however, diffuse into civil computing. . . . [quoting TRW historian] The “waterfalltechnique, for example, provided for comprehensive reviews at key points in the development of the software as it proceeded from requirements analysis and definition through coding to test and evaluation.
(48) When one compares the $8 billion cost of SAGE with the $30 million it took to build SABRE, one gets a sense of the magnitude of the cost of establishing real-time technology and the benefit to the civilian sector. This indirect form of “technology push” by government was unique to the United States, and it helped established US dominance of the computer industry.
(48) After the airlines, the next industry to attempt real-time computing was banking, one of the most information-intensive business and a leader in information technology throughout the twentieth century.

Software Contracting Startups

The Computer Usage Company

The Computer Sciences Corporation
(52) The Computer Sciences Corporation (CSC) has been by far the most successful of the 1950s startups.
(53-54) Nonetheless, the cost overruns and delivery slippage in developing FACT were much talked about in the industry, and clients became more and more reluctant to offer open-ended time-and-materials contracts, preferring fixed price terms, ideally with penalties for late delivery.

Summary


3
Programming Services, the 1960s

Chapter 3 begins with verbose text advertisement for the Independent Software Houses.

Four sectors of computer services and software industry: programming services, processing services, facilities management, teleprocessing services.

(57) The computer services and software industry constituted four main sectors: programming services, processing services, facilities management, and teleprocessing services. By the end of the 1960s, however, the term software industry had largely taken on its present-day meaning, signifying commercial organizations engaged in the production of programming artifacts.

Computer Software and Services
(57) Most of the entrants came into programming services much as firms such as CUC and CSC had in the 1950s. They were typically founded by entrepreneurially minded individuals who had acquired technical skills with a computer manufacturer or a major user and who had a network of contacts that led to the crucial first contract.
(58) However, ADR's place in the software hall of fame is due primarily to Autoflow. Introduced in 1965, Autoflow is generally credited with being the first software product.
(58) Though Informatics was a major player in the early software contracting business, it made its mark on the history of software in 1967 by developing Mark IV, which would be the world's best-selling independent software product for 15 years.

Few early software firms left a historical trace; then there were too many to analyze individually.

(58) they [ACT, ADR, Informatics] were rather good at getting their stories in the media or because they took the trouble to preserve their history. In contrast, the great majority of software firms vanished, barely leaving a historical trace.

Role of industry analysts in organizing and creating historical sources in place of scholars.

(59) Indeed, if anything, the problem is that there are too many firms to analyze individually. . . . Hence, probably the best one can do is select a group of firms and let them stand as proxy for the rest. This was, in fact, the approach taken by contemporary analysts. Table 3.1, taken from what is perhaps the earliest surviving analysts' report on the consumer software and services industry, presents data on 17 firms. The analyst never explained how the table was derived, and some prominent players (including CUC and CAI) are missing.
(59) In the 1960s the software industry was too small an entity to warrant analysis in its own right. For this reason,
programming services was conflated with three other activities to form what was known as the Computer Software and Services Industry (CSSI). The other sectors were processing services, facilities management, and teleprocessing, the precise terminology varying over time and between analysts.
(62) By far the biggest independent player in processing services was Automatic Data Processing (ADP), established in 1949 by the entrepreneur Henry Taub as Automatic Payrolls Inc. in Clifton, New Jersey.
(62)
Facilities management was, in a sense, the opposite of processing services. A facilities management firm did not provide a computing service; it managed a data processing installation on behalf of the firm that owned it. . . . The concept was brought to the computer industry by H. Ross Perot, who established Electronic Data Services (EDS) in 1962.
(63) The [teleprocessing] sector came into its own in the mid 1960s with the advent of commercial time-sharing services. . . . However, the arrival of inexpensive minicomputers in the 1970s devastated the industry. The survivors were firms, such as Tymshare and Comshare, that refocused on network operations and on-line databases.
(63) Of the four sectors, programming services had by far the lowest barriers to entry; consequently, it attracted the most new entrants.

Inside the Software Contractor
(65) The most interesting projects were the big real-time systems that were then at the leading edge of technology. . . . Supplying manpower (sometimes referred to in the industry as “body shopping”) was lucrative. Even a lowly coder could command $10 an hour, about 3 times the direct labor cost.

Marketing
(66-67) Informatics' marketing, “circumscribed by the personal circle of acquaintances of the principals,” was typical of the software contracting industry, and it was extremely cost effective. . . . Firms often touted their “star” creative employees (in this case, programmers), much as advertising agencies did. . . . Another technique for raising a firm's profile was to sponsor meetings and conferences that got the firm known by potential clients. . . . Another method was for the founder to become a prominent personality. Martin Goetz of ADR was the most quoted software pundit of the 1960s, and his name was rarely out of the trade press.

Project Management
(67) For a software contractor, the ability to complete a programming project on time and within budget was the most critical competence.
(67) Early software projects got an exaggerated reputation for cost and time overruns and poor reliability.

Familiar sequence of tasks for programming projects arose in 1956 lecture and diffused from SAGE project programmer exodus, which became the waterfall model; should be topic of critical programming studies for its reflection of social and cognitive norms, social construction, as well as likely reflexive relationship to the evolution of technological systems along with human thinking. Could call for refinement of Turkle hard mastery and bricoleur programming styles. Note paradigm shift from flowcharting to structured programming in 1970s.

(67-69) During the 1960s, managing software projects was something of a black art. He [John F. Jacobs] described these techniques in a seminal lecture at the Lincoln Labs in November 1956. . . . In the seminar, Jacobs divided the programming project into a sequence of consecutive tasks (figure 3.2). . . . This development technique set the pattern for all the programming activities of the SAGE project. The exodus of programmers from the SAGE project in the late 1950s caused this project management style to diffuse through the software industry.
(69) A common theme of these techniques was dividing the programming project into a temporal sequence of tasks, freezing specifications between them, and using a hierarchy of labor that cascaded down from systems analyst to senior programmer to programmer to coder. . . . As was noted in chapter 2, at TRW this was called the “waterfall model” because of the irreversible cascading of tasks.
(69) In the 1960s, systematic project management was a touchstone of certainty in an uncertain world, but it was a social as well as a rational construction. For example, the need for program flowcharts was an article of faith. . . . In the early 1970s, a fad for “structured programming” swept aside 20 years of flowcharting history in about 3 years.
(70) It took 10 years or longer for a programming manager to become fully seasoned in a particular application domain. . . . Rank-and-file programmers were relatively easy to find or train, but project managers had to be grown.

Economies of Scope
(70) Invariably, the niche selected by a firm was determined by the history of its founders.
(71) As firms picked up more and more contracts in the same application domain, the knowledge was effectively captured by software tools and code assets that could be endlessly redeployed for different clients.

A Little Bit of Everything”: Diversification of the Software Contractors

The International Scene
(75) In the 1960s, the European computer scene lagged the American scene by 3-5 years in number of computers, size of computers, and development of computer and software industries.
(75) By the mid 1960s, there was great political concern, particularly in Britain and France, over the “American invasion” of high-tech industries such as aerospace, nuclear power, and computers.

The Go-Go Years: From Boom to Bust
(83) The software firms hit worst were those that had diversified into teleprocessing services in 1968 and 1969, including CAL, CSC, and UCC. Each of these firms had effectively bet its survival on the emergence of the “
computer utility.” This concept was always somewhat nebulous, its exact meaning depending on who was doing the promoting. In essence, however, the idea was that eventually users would own neither computers nor software; they would run applications software on remote machines through private or public data communications networks.

Failure of computer utility as forerunner of cloud model before the infrastructure was in place overcome by arrival of minicomputers. Mentioned in summary of Chapter 5.

(83) The reason the computer utility market never materialized to the extent anticipated was the arrival of small-business systems and minicomputers in 1969 and 1970.
(85) By 1971, half of the estimated 3,000 computer software and services firms that had been in business at the height of the 1966-1969 stock market boom had gone out of business or “simply faded away.”

Summary
(87) The irrational euphoria for computer stocks and the failure of the computer utility were nearly unique events in the computer industry, and it was a generation before a comparable phenomenon occurred with the Internet boom at the end of the century.
(87) Software contracting remains the most popular way of participating in the software industry, programming services enterprises outnumbering software products firms by 2 or 3 to 1.


4
Origins of the Software Products Industry, 1965-1970

Chapter 4 begins with Autoflow advertisement as a Christmas gift for a wife.

The Rise of the Software Product

Striking claim that graphs depicting computing costs seldom based on empirical data, often plagiarized and embellished, as if no oversight in analyst community.

(91) For more than 15 years, one of the most persistent beliefs of the software community was that there would be a dramatic shift in the balance of hardware and software costs of running a computer installation. The graphs in figure 4.2 all tell much the same story: In the mid 1950s, 80 percent of the cost of running a computer had been for hardware, 20 percent for programming; at some point in the future, it would be 20 percent for hardware and 80 percent for software. Few of these graphs were based on empirical data, and they were plagiarized, with random embellishments from one author to the next, well into the 1980s.
(94) There were several software crises in the 1960s, apart from the programmer shortage: the low productivity of programmers, the poor reliability of programs, and cost overruns.
(94) Most of the software costs for large projects were not in fact for writing the code per se but for testing and debugging.
(95) Estimates of programming manpower required—and therefore project duration and cost—often proved optimistic, and often by “a factor of 2.5 to 4.” Again, the most celebrated example was IBM's OS/360. That project started in 1965 with a team of 150 programmers, quickly slipped 6 months behind schedule, so more programmers were added; this happened time and again, and eventually a thousand people were working on the project.

Recommendation by SPREAD Task Group for single series of machines with compatible software led to System/360.

(96) The [SPREAD Task Group] report recommended replacing IBM's existing computer models with a single series of machines, software compatible throughout the range from the smallest to the largest. The recommendation was adopted, and the announcement of System/360 in April 1964 was a watershed in the software industry.

From Software Package to Software Product
(96) One of the abiding legacies of user groups was the perception of software as a free good.
(98) As the 1960s unfolded, software packages for all the major industries were developed by IBM and other mainframe makers. These software packages, like other customer services, were “bundled” with the cost of the computer. Software development was always perceived simply as a marketing expense; there was no thought of recovering costs by selling or leasing packages.
(98) In a path-breaking tactic, the 200 series was also made software compatible with the IBM 1401, so that users could readily migrate to the more price-competitive Honeywell series, taking their existing applications (whether written by the users or developed using IBM's software packages) with them. . . . This was possible only because IBM's program packages were supplied free of charge, with no intellectual property protection.

The First Software Products
(99) The term “software product” firmly entered the computing lexicon in 1966. . . . The new firm's principal activity was to publish
ICP Quarterly, a catalog of software products in which vendors could list their software packages free of charge.
(99-100) The distinction between a software package and a software product was that the latter was a traded capital good. A software product was a discrete software artifact that required little or no customization, either by the vendor or by the buyer; it was actively marketed, it was sold or leased to a computer user, and the vendor was contractually obligated to provide training, documentation, and after-sale service. In contrast, the manufacturer's “free” software package was supplied with few contractual obligations, and no matter how well the customer might be supported in practice, there was no legal requirement that the manufacturer supply a fully operational application.

Autoflow, the first software product, was designed for software development to automatically produce flowcharts for documentation, very early automatic writing; likewise Engelbart intelligence augmentation focused on improving programmers.

(100) The most prominent if not actually the first program to meet all the criteria of a software product was Applied Data Research's Autoflow. . . . RCA wanted a software package that would enable programmers to produce flowcharts mechanically as a by-product of punching their decks of program cards, thus minimizing the documentation chore for the programmer and maximizing the chance it would get done.

Pioneering in the Software Products Industry: Informatics Mark IV
(103) From its launch in November 1967 until its peak in the early 1980s, Mark IV was the most important and best-selling product from an independent software vendor.
(103) More than any other product, Mark IV shaped the fiscal, marketing, and after-sale practices of the independent software products industry.

Product and Platform

Financing
(106) Bauer therefore encouraged Postley to seek out some customers who would be willing to fund the development by buying the product in stage payments before delivery.
(106) The market for Mark IV would be determined largely by its selling price. . . . Informatics settled on a price of $30,000. The company got just one detail wrong: It greatly underestimated the cost of maintenance and the potential revenue from upgrades. It took several years to build these factors into its price structure.

Intellectual Property Protection

Ontic status of software became an open question once IP protection was sought.

(107) The law offered several forms of intellectual property protection: patents, copyright, trade secrets, and trademarks. . . . Exactly what kind of an artifact software represented was an open question.
(107-108) Copyright law also had significant limitations for protecting software. . . . Informatics decided to rely on trade secret law for protection. It required its customers and its employees to sign non-disclosure agreements.
(108) Informatics did, however, fully utilize copyright law to protect the supporting documentation for Mark IV as conventional literary works.
(108) Informatics also made good use of trademarks.

Marketing

IBM's Unbundling Decision
(109) In 1967 the Antitrust Division of the US Department of Justice began an investigation of IBM, which was widely reported to have a 70 percent share of the domestic computer market. Since a high market share was not itself a violation of the Antitrust Act, attention focused on IBM's commercial conduct.
(109-110) Against this background, on December 6, 1968 (only days before the CDC lawsuit), IBM announced its intention to unbundle its prices and to charge separately for the five service elements then included in the overall price charged to the customer; systems engineering, education and training, field engineering, programming services, and software packages.

Unbundling stopped at the operating system, raising suspicion that its inefficiency forced users to lease more powerful computers than necessary; obvious connection to Microsoft Windows/Intel era.

(111) The decision not to charge separately for the operating systems—the most controversial of the unbundling outcomes—maintained IBM's operating-system monopoly into the late 1970s, when at last the operating systems were priced separately. Critics argued that bundling discouraged innovation in operating systems and made it difficult for manufacturers of IBM-compatible equipment to compete. There was also suspicion of a darker motive: that IBM benefited from an inefficient operating system because it forced users to lease more powerful computers than they would otherwise need.
(113) Seventeen products were initially announced, including language processors (Assembler, FORTRAN, COBOL, and PL/I), the GIS file-management system, the IMS database, and the CICS data communications program.
(113) More generally, ADR had argued that, by the practice of bundling, IBM had suppressed a huge potential software market for a decade.

After Unbundling
(114) The historian JoAnne
Yates has described unbundling as the “crucial inflection point” in the development of the software products industry.
(115) The immediate beneficiaries of unbundling were the existing computer services and software firms—not only those with products, such as ADR and Informatics, but also those with unrealized software assets.
(116) In all, Autoflow had a life of about 15 years. That turned out to be typical of the software products industry, and it was much longer than early entrants had expected.

Summary
(118) The term “product” was consciously adopted by vendors to imply a new kind of software artifact for which the vendor took contractual responsibility for performance and reliability in exchange for the license fees paid by users.


5
The Shaping of the Software Products Industry, the 1970s

Chapter 5 begins with SyncSort advertisement.

(121) The software products industry is best understood through two vectors of analysis, with the organizations that produced software on one axis and their products on the other.

The Business Environment
(122-123) A vast new market for software was created by the rise of small computers. These systems sold in the tens of thousands to medium-size firms that did not employ computer programmers and were therefore entirely reliant on packaged software.
(124) Until then [1975], the conventional wisdom had been that the life of a software product was about 5 years, and software was priced to recover development costs in that period. When the life of System/360 was extended, the life of its software packages was extended too. With regular product upgrades, a life of 10 years, 15 years, or longer turned out to be not uncommon. This was perhaps the most significant reason why the major vendors of software products were able to survive the 1970s.

Suppliers of Software Products

Computer Manufacturer and Captive Revenues

Independent Software Vendors
(126) The most interesting firms, perhaps, were the big software products vendors established in the late 1960s that grew to prominence in the 1970s—e.g., Cincom, Cullinane, Information Sciences, and Pansophic. Another group of firms had longer histories as programming services and computer services firms but had also become significant players in software products—e.g., Informatics, University Computing Corp., and MSA [Management Science America].

Turnkey Suppliers

Interesting that an unnamed example of a turnkey supplier is Toptech Systems, which supplied terminal automation solutions to the petroleum industry from the 1990s onward, from which a services company also named Triad Systems started by former Toptech employees to support the extremely complicated TMS5 software product of the same name now competes for a share of the third party support market.

(128) A turnkey vendor addressed these two problems by supplying a complete package of hardware, software, and ongoing support. A prominent late-1970s example is Triad Systems.

Software Brokers

Failure of COSMIC early free software mechanism even without original development cost due to poor products, documentation, and lack of customer support offerings.

(130) COSMIC was established in 1966—still in the era of free software—to make programs that had been developed for NASA at public expense available as a national resource for industry and education. . . . COSMIC was a spectacular failure. A contemporary report commented: “[COSMIC's] major problem—like any giveaway program—has been in giving it away.” Generally the programs were so poorly documented and productized that they could not be used without major investments in programming time.
(130) INSAC was a much more significant venture. It was formed by the British government in 1977.
(130) The reason all these attempts at software brokerage failed was that competitive pricing could not compensate for poor products and lack of customer support.

Time-Sharing Services
(131) Time-sharing services were mostly used by engineers and scientists as “super calculators” for stress analysis, linear programming, matrix analysis, and other mathematical operations. . . . For most program authors, however, writing programs for time-sharing services never amounted to more than a hobby or a part-time job. It was, however, the most accessible route into the software industry before the emergence of the personal computer.

Market Structure and Taxonomy
(131) In July 1974,
ICP Quarterly listed 2,928 software products from 740 vendors. Journalists and industry analysts used a variety of classification schemes to impose some order and coherence on this fragmented market.
(133) Roughly speaking, the ratio of systems packages and application packages was 1:2. This was true whether measured by the number of packages, by the number of firms, or by the total value of the market (table 5.5), and it has remained approximately true over time.

Systems Software Packages
(134) Table 5.6 shows the revenue split in 1980 for the five divisions of the systems software package market: operating systems, database systems, teleprocessing monitors, programming aids, and utilities.

Unix the only significant cross platform operating system in period studied.

(134) With the important exception of Unix, users had the choice of exactly one operating system for any particular computer configuration and application.
(136) In 1976, Computer Associates introduced CA-SORT, a competitor to SyncSort. There was room in the market for both products, and CA-SORT turned out to be the springboard for the growth that made Computer Associates the second most significant independent software vendor (after Microsoft) of the 1990s.

Industry-Specific Applications
(136) Measured either by number of packages or by market value, industry-specific applications constituted the largest sector of the software products industry. . . . Financial services, the most information-intensive sector of the economy, had the most to gain from computerization and was one of the earliest adopters. . . . By 1980, finance, utilities, and manufacturing had emerged as the most computerized areas of the economy. Distribution and public agencies were still relatively undercomputerized, but during the 1980s they began to catch up.
(138) The primary competence needed by a new entrant was an in depth knowledge of the target industry; program-writing skills were relatively unimportant. (In the systems software sector, in-depth programming skills were critical.)

Cross-Industry Applications
(139) In 1974,
ICP Quarterly listed 777 cross-industry applications, the second largest category (table 5.8).
(139) Payroll was one of the first and easiest tasks to computerize, and hence it had a disproportionate presence in the early software taxonomy. . . . In general, the cross-industry applications of the mid 1970s were dominated by routine, batch-oriented back-office tasks that were easy to computerize.
(139) In the 1970s there were two major markets for turnkey equipment: the market for word processors and that for computer-aided design systems.

[divider symbol]
(141) In the second part of this chapter, the software “product space” is explored by means of a historical account of the evolution of the most important genres.

Operating Systems

The Captive Operating System
(143) IBM eventually unbundled its operating systems in the late 1970s. As new versions of the operating systems were introduced, IBM froze the specifications of the earlier versions (which remained forever free), designating the new versions as paid-for system products.

Unix and Open Systems
(143) Unix is the only non-proprietary operating system of major significance and longevity.
(143-144) The history of Unix is well documented. . . . Indeed, it was probably because Unix was virtually free that it proliferated in the cash-poor, manpower-rich university environment.

Unix culture produces a new generation of programmers supplanting the IBM ethos; what are the cultures of the subsequent generations seems an important consideration for critical programming studies, along with questions of how the well documented history of Unix, versus other early free software, coincide with its proliferation in university environments?

(144) By the mid 1970s, Unix was in wide use on college campuses. Thus, a generation of computer science graduates grew up in the Unix culture rather than in the IBM ethos of the previous generation.

Database Systems
(145) The central concept of the database was that there would be exactly one copy of any item of data, and that this would guarantee its integrity.
(145-146) Had history unfolded differently, the database might have been a fundamental property of the operating system; there would then have been no database industry at all.
(146) Cincom's TOTAL, IMS's chief competitor, was the first true database to be produced by an independent software vendor and enjoyed an enormous first-mover advantage.

CICS: IBM's Teleprocessing Monitor
(149) There is no better example of the invisible software infrastructure that runs the modern corporation than CICS, which for more than 30 years has been the world's best-selling computer program. . . . Whenever an ATM is accessed, a travel reservation is made, or a credit card is used in a retail purchase, the changes are that CICS is involved.

Industry-Specific Applications

Banking Software Packages
(154) By 1990 it was clear that integrated software packages for major banking operations were not viable. The world's major banks have continued to rely on custom written software for their core operations and have remained among the largest employers of programmers and analysts outside the software industry.

ASK and MRP Software
(155) ASK's turnkey system [MANMAN] was a hit, primarily because it could be located on a machine-shop floor rather than in a corporate mainframe suite. In effect it became another machine tool under the direct control of the manufacturing operation.

Cross-Industry Applications

MSA and Accounting Software

Wang Laboratories and Office Automation
(160) Wang's 30-second TV commercial during the 1978 Super Bowl was probably the first instance of a computer firm's appealing to office workers rather than to data processing managers.

Computer Vision and CAD Systems
(160) What the word processor did for the ordinary office, the computer-aided design system did for the engineering design office.
(160) CAD systems were expensive, costing perhaps 10 times as much per user as word processing systems.
(161) They remained a significant sector of the computer market despite competition from personal computers.

Summary
(162) Selling software through time-sharing services was a fascinating strategy, but before it could become a significant form of delivery it was overtaken by the collapse of the computer utility in the mid 1970s. The idea has recently resurfaced in connection with the Internet; the suppliers are called “application service providers” (ASPs). Because the concept was never fully tested in the 1970s, history has few lessons to offer to the ASP industry.


6
The Maturing of the Corporate Software Products Industry, 1980-1995

Chapter 6 begins with advertisement for Software News promoting itself as news service for the fastest-growing industry.

Can the glut of software reopen personal programming projects as a mode of comportment to the technological lifeworld?

(165) By the mid 1980s, it could be said without hyperbole that “few users write their own software anymore”--and with a reported 8,000 products from more than 3,000 vendors, there were plenty to choose from.

The Ascendancy of the US Software Products Industry

Suggest a cultural/linguistic analysis to probe why the various countries did or did not participate in the software products market, with a stereotypical nod to the disciplinarity of the German programming milieu with respect to the success of SAP R/2.

(166) Overall, the United States had two-thirds of the world's software market. But because most overseas activity was in custom programming and in computer services, the United States supplied at least 95 percent of the software products market. . . . In fact, Germany was the only country besides the United States to have a significant player in software products [Software AG and SAP].
(168) Paradoxically, the high rate of M&A activity has led to neither a much higher concentration nor to a decrease in the number of firms.
(168) Most of the readable journalistic accounts focus on individual large firms, treating them in isolation. In this chapter, I opt for a middle course, focusing on the three largest firms that came to prominence in the 1980s: Computer Associates, Oracle, and SAP (all founded in the 1970s). . . . It also happens that the three firms were the most successful exponents of the three phenomena that shaped the 1990 corporate software products industry: consolidation, the rise of the relational database, and Enterprise Resource Planning software.
(172) ERP software effectively hit the middle ground. It was a body of generalized, integrated software that could be customized for virtually any large business.

IBM and Other Computer Manufacturers as Suppliers of Software Products
(173-174) The biggest seller of corporate software products was not an independent software vendor; it was IBM. . . . Not until 1998, when Microsoft overtook it, did IBM cease to be the world's largest software supplier.
(176-177) One of the paradoxes of the software industry was how IBM came to be so dominant despite the mediocrity of its software. . . . Leasing, rather than one-off sales, gave IBM a secure, constant revenue stream and a way to raise its prices from time to time.
(177) IBM's main advantage over its competitors, however, was that of what now was called “bundling.” In its new sense, this word now meant selling a bundle of software packages that were known to work harmoniously together.
(177) IBM's biggest software initiative of the 1980s was is Systems Application Architecture (SAA), announced in March 1987. . . . Indeed the SAA announcement was too arcane for most computer professionals, never mind the public at large. The firm that had charmed the world with the Charlie Chaplin advertisements for its PC had reverted to type when it came to corporate software.

IBM SAA story waiting to be told well.

(178) The solution was SAA, a set of rules and interfaces that enabled a program written for one platform to be easily “ported” to another. This sounds easy, but in effect it took 3 years of negotiation between IBM's globally distributed software and hardware development centers. The SAA story--”IBM's Moment of Truth”--is, in its way, as compelling as the development of System/360 or that of the PC, but it has never been told well.

Computer Associates, the Ultimate Consolidator
(178) In February 2000, Computer Associates acquired Sterling Software for $4 billion. The two largest consolidators in the software industry were now combined into an organization that enshrined a veritable history of the software products industry.
(180) As befits its anonymity to the general public, the literature on Computer Associates is sparse, and most of what has appeared focuses on the management style and idiosyncrasies of its founder, Charles Wang.
(184) In effect, Computer Associates had become a one-stop software shop for customers who could now buy their hardware from a commodity-type market.

Oracle and Maturing of the Database Market
(185) Though comparisons are often drawn between Microsoft and Oracle, they are very different firms with very different business models. . . . It is the difference between consumer goods and capital goods.
(186) While the phenomenon of “regional advantage” is well known in the hardware industries, it is much less common in software. Northern California's advantage in regard to relational technology was due largely to the fact that the technology was little published and non-obvious (certainly relative to programming languages and word processors, which had spread like wildfire), and so the ideas were slow to diffuse.

Relational database a disruptive technology, enough so that a code snippet is given.

(186) The invention of the relational database was a classical example of what some Harvard economists have called a “disruptive technology.”
(187) In brief, the underlying mathematic basis for the relational database was developed by Edgar F. Codd, an IBM research scientist, in 1970. The particular advantage the relational database offered was its ability to handle complex queries.

SAP and ERP Software
(191) A common characteristic of ERP software was extreme complexity. Few corporate information systems departments were capable of installing ERP software without the assistance of external consultants.
(193) Having a real time program was a significant advantage for SAP, whose US rival's software had generally evolved in earlier, non-real-time environments.
(193) The development of R/2 owned a great deal to the German programming milieu, which has always been more formal and more disciplined than its American counterpart.

SAP spread via Trojan horse effect via German subsidiaries of multinationals, entering US after IPO.

(194) The Trojan Horse effect of R/2's having been installed in the offices of German subsidiaries of multinational firms was probably a more effective sales device. . . . Thus, the SAP package was completely generalized, and was tailored to an individual country and firm by setting thousands of parameters and switches. . . . Its 1988 IPO in Germany provided the capital for entry into the United States, the world's biggest software market.
(195) Whereas R/2 had been a mainframe product, R/3 had been designed for client-server computing. It was released in Germnay in 1992 and in the United States the following year. The R/3 product was distributed on two CD-ROM discs, which somehow concealed that the fact that the software was stupendously complex and had reportedly cost $920 million to develop.
(197) In many ways, SAP is a more potent monopolistic threat to the United States than Microsoft.

SAP R/3 critical to Western industrial economies, and irreplaceable in short term; too big to fail?

(197) If overnight R/3 were to cease to exist (say, if its licenses were made intolerably expensive), the industrial economy of the Western world would come to a halt, and it would take years for substitutes to close the breach in the networked economy. Were Microsoft's products to vaporize overnight, it would take only days or weeks to find substitutes, and the economic disruption would be modest.

Summary
(198) In software, one can discern strong parallels with other historical episodes of system building, from railways in the nineteenth century to airlines in the twentieth.
(198) That the database industry exists at all is something of a historical accident. . . . Fortunately, software is infinitely malleable, and incumbent firms were able to retain most of their customer bases. There are not many other fields where this would have been possible—it was like a manufacturer of piston engines suddenly having to switch to turbines.
(199) The reason European firms were able to make this [ERP] market their own was related to the cycle of infrastructure investment. . . . At this time, European organizations were undercomputerized relative to American ones, so ERP software from indigenous firms was able to fill the software vacuum when computerization took off in the 1980s. In the United States, widespread computerization had taken place 5 to 10 years earlier than in Europe, and organizations were locked into earlier, discrete software packages that had been painfully integrated by in-house staffs—an infrastructure that is usually graced with the name “
legacy software.” . . . It may be that the United States, having been first to create a software infrastructure, will be the first to face the problem of renewing such an infrastructure.


7
Early Development of the Personal Computer Software Industry, 1975-1983

Chapter 7 begins with Visicalc advertisement, emblematic of personal computer software industry.

Computer industry did not perceive microprocessor-based stand-alone devices as threat because they were developed in the electronics industry.

(201) The microprocessor contained all the essential parts of a processor on a single chip, and, when combined with memory chips and peripheral equipment, was capable of performing as a stand-alone computer. However, because the microprocessor had been developed in the electronics industry, the computer industry did not perceive it as a threat to its products.

The Origins of the Personal Computer Software Industry

Bricoleur, dilettante origins of microcomputer software development practices; examine emergence of Altair 8800 and folk history of PC.

(202) Although some professional software development practices diffused into microprocessor programming, much of the software technology was cobbled together or re-invented, an amateurish legacy that the personal computer software industry took several years to shake off.
(202) The first microprocessor-based computer (or certainly the first influential one) was the Altair 8800, manufactured by Micro Instrumentation Telemetry Systems (MITS). This machine was sold in kit form for assembly by computer hobbyists, and its appearance one the cover of
Popular Electronics in January 1975 is perhaps the best-known event in the folk history of the personal computer.
(202) The transforming event for the personal computer was the launch of the Apple II in April 1977.
(203) The launch of these genuine consumer products created a significant consumer market for personal computer software. Software packages were typically priced between $30 and $100.

Effect on subjectivity of fiddling with spreadsheets; see discussion of killer app hypothesis on 212.

(203) Perhaps the second most celebrated event in the folk history of the personal computer is the fall 1979 release of VisiCalc, a software package that transformed the perception of the personal computer as a business machine. Up to that time few people had even heard of a “spreadsheet,” but within a few years the term had entered the lexicon of business and fiddling with spreadsheets had transformed the working lives of middle managers in the Western world.

First Movers in Systems Software: Microsoft and Digital Research
(203-204) Before there could be a software industry for the microcomputer, however, there had to exist the foundation of an operating system and programming languages. The market for these systems programs was shared by Microsoft and Digital Research.

Programming Languages
(204-205) According to Gates and his many biographers, in his sophomore year at Harvard he saw the January 1975 issue of (sic)
Practical Electronics, which had the Altair 880 kit on the cover, and saw in a flash the opportunity to become the leading vendor of programming languages for microcomputers.
(205) Payment by royalties was effectively a new business model for the software industry.

Operating Systems
(206) The first vendor of microcomputer operating systems was Digital Research, founded in 1976 by Gark Kildall (1942-1994). . . . He called the system CP/M (Control Program for Microcomputers).
(206) The cost of developing an operating system, perhaps $50,000 or $100,000, was a major barrier to entry to the manufacture of personal computers. But CP/M (essentially the Unix of the microcomputer world) could be adapted at minimal cost to run on any 8008-based machine, and this resulted in dozens of new entrants into the burgeoning personal computer market in 1977 and 1978. Digital Research grew rapidly with the market, creating new versions of CP/M to run on different microprocessors. By enabling any CP/M-compliant application to run on any computer using the CP/M operating system, Kildall liberated the market for personal computer software (much more so than Microsoft's programming languages).
(206) The turning point in the relative fortunes of the two companies came with the award to Microsoft of a contract to develop the operating system for IBM's new personal computer, launched in August 1981.
(207) How Digital Research came to pass up the opportunity to create the IBM operating system has become one of the most poignant episodes in the folk history of the personal computer.
(207) To speed the development, they obtained the rights to a locally developed operating system for a reported $100,000. They quickly re-engineered it into what became know as MS-DOS (MicroSoft-Disk Operating System). The operating system was supplied to IBM on a royalty basis. It was estimated that Microsoft received $10 for each copy sold.
(207) Microsoft was already becoming the most opportunistic and diversified of the microcomputer software firms, its many products including add-on hardware cards and computer games in addition to programming languages.

The Making and Selling of Microcomputer Software
(209) By 1981, retailing software was a $150 million business, and software-only stores had begun to flourish.
(209) The distributor had a somewhat unglamorous role, shipping boxes of shrink-wrapped software from the publishers to the retailers.
(209-210) However, the crucial link in the software supply chain was the software publisher. The role of the publisher has been understated in the history of the industry, most of the glamor going to the “creative” software developers.
(210) In many respects the relationship between a software publisher and the developer was like that between a book publisher and an author.

Unacknowledged role of software distributor, with focus usually on developer as creative force, ignoring importance of marketing efforts; nice link for humanities comparing to publisher.

(210) Perhaps the most interesting revelation was that marketing consumed 35 percent of costs.

VisiCalc and the “Killer App” Hypothesis
(212) VisiCalc is often described as the “killer app” that unleashed the corporate personal computer revolution.

Killer app hypothesis mixes social constructionism and critique into technology studies.

(212) The “killer app” hypothesis argues that a novel application, by enabling an activity that was previously impossible or too expensive, causes a new technology to become widely adopted. . . . Outside the realm of computers, The Jazz Singer has been credited with establishing sound cinema in the late 1920s.
(214) By late 1983, an astounding 700,000 copies of VisiCalc had been sold. Almost entirely on the basis of income from VisiCalc, Software Arts and Personal Software had become significant firms.

Productivity Applications: Spreadsheets, Word Processors, and Databases
(215) Productivity applications, constituting more than 70 percent of the business package market, drove the personal computer software industry.

Spreadsheets

Word Processing Software
(216-217) No market better illustrates the paradigm shift that the personal computer represented than that for word processing software. The bestselling word processing program, WordStar, was priced at $495. IBM's comparable mainframe package, ATMS-III, leased for between $330 and $616
a month.

Databases

Small-Business Software

Cross-Industry Applications

Industry-Specific Applications
(223) Although industry-specific applications were marketed as “packages,” they were much more like corporate software products that required some customization and after-sale support.
(224) In the boulders-pebbles-sand model of the software industry, the microcomputer-industry-specific application vendors were some of the finer grains of sand.

Consumer Software
(224) One of the reasons for the slow development of the home computer market had been the dearth of software titles.

Entertainment Software

Home Management Software
(226) By about 1983, word processing had emerged as the most important use of a home computer.

Educational Software

Summary


8
Not Only Microsoft: The Maturing of the Personal Computer Software Industry, 1983-1995

Chapter 8 begins with 1-2-3 spreadsheet advertisement for maturing personal computer software industry.

Microsoft and the Software Industry
(232) IBM, at its peak, in the 1960s, had a three-fourths share of the worldwide computer industry—hardware, software, and services. Microsoft has never even had a 10 percent share of the software market.

The IBM-Compatible PC Standard
(237) It should be noted that the success of System/360 was largely independent of its original technical merits. It was the mere fact that it was a standard platform that made it desirable.
(238-239) A fabrication plant cost about $1 billion before the first microchip rolled off the production line, and this limited competition to a few major players, such as Advanced Micro Devices, Inc. (AMD) and Cyrix. In contrast, Microsoft had numerous competitors because the capital requirements for producing an operating system were relatively modest.
(240-241) There was considerable market uncertainty in the industry, and within Microsoft, as to whether MS-DOS would forever be a product in its own right or would eventually be subsumed into a Unix-style operating system.

Strategic Understanding in the Personal Computer Software Industry: The Case of Autodesk

Difficulty of getting access to corporate archives; Autodesk an exception.

(243) But until Microsoft opens its archives to independent scholars, we have only some tantalizing hints of the company's strategic thinking and the extent to which Gates was responsible for it.
(243) In the meantime, another firm, Autodesk has made its early strategic thinking publicly available in a book titled
The Autodesk File.
(245) More than 600 educational institutions taught engineering drawing using subsidized AutoCAD software, thus locking in the rising generation of engineering graduates.

Paradigm Shift: The Graphic User Interface
(247) In 1982 and 1983, a consensus emerged that the best way to achieve multitasking was by means of a windowing system.

Conflation of multitasking with user friendliness.

(247) As a result of the rise of the graphical user interface, multitasking became conflated with the secondary issue of user friendliness.
(250) Only Microsoft and IBM had the resources to persist with a graphical user interface,which would have to wait for the next generation of Intel microprocessors.
(251) Windows 3.0 received unequivocal market acceptance, and the paradigm shift had finally occurred.

Microsoft and Productivity Applications
(251) Microsoft's strategy has been to use the revenues from its successful systems products to develop applications software, without regard to short-term profitability. Invariably, these packages have succeeded only after a third or subsequent product launch.

Lotus 1-2-3 and Excel
(253) Accepting the impossibility of competing with Lotus 1-2-3 head on, Microsoft decided to develop a GUI-based spreadsheet for the Macintosh.
(253-254) However, to most observers, including Lotus, OS/2 seemed the more likely successor to MS-DOS. . . . Excel filled the spreadsheet vacuum created by Windows.

WordPerfect and Microsoft Word
(254) The WordPerfect Corporation was one of the few personal computer software publishers to have originated in the world or corporate computing.
(256) Again, as with Excel, a version for the Macintosh enabled Microsoft to perfect its graphical user interface and its laser printer technology out of sight of the mainstream IBM PC market.

Aston-Tate
(257) It should be noted that the demise of Ashton-Tate cannot be traced to Microsoft. Even best-selling products, such as dBase IV, VisiCalc, and WordStart, could fall from grace because of obsolescence or unreliability.

Product Integration and Office Suites
(259) Microsoft has never explicitly described the strategy behind Office, but analysts have speculated that the package enabled the company to push loss-making applications, such as PowerPoint, onto people's desktops and to gain market share from its one-product competitors while incurring relatively little revenue loss.

Competing with Microsoft
(260) First, some firms have published products complementary to Microsoft's best-selling packages.

Suggestion of religious rather than rational preference for programming languages links to social construction of technological systems.

(260) A second way of coexisting with Microsoft has been to occupy niche markets where Microsoft has no presence. A classic example of this was Borland, which introduced Turbo Pascal in 1982. It happens that developers prefer particular programming languages for reasons that are more religious than rational. Pascal was an elegant programming language, favored by academics and by idiosyncratic firms like Apple Computer, whereas Microsoft and most software developers preferred the more prosaic BASIC or C programming language.
(262) A third way of competing with Microsoft has been to develop packages that run on a range of operating systems and platforms, typically the big three: IBM-compatible PCs running MS-DOS or Windows, the Apple Macintosh, and Unix.
(263) The fourth way for firms to maintain market share relative to Microsoft has been grown through acquisition.

Summary
(264) Aside from his [Gates] considerable though unexceptional technical capabilities, his success has been attributed to an ability to understand and exploit the economics of increasing returns.
(265) In each case, an intuitive understanding of increasing-returns economics enabled a firm to dominate its market. Though this hypothesis remains to be proved, it is provable that a majority of the successful firms had a similar vision.
(265) However, Microsoft had more luck than just picking the right theory. The rise of the Windows operating system led to Microsoft's ultimate domination of the most lucrative sector of the personal computer software industry: office productivity applications.
(266) Microsoft had failed, as a mature company, to moderate the external signs of the very behavior that led to its success. It may be that a social historian of the future, rather than an economic historian, will explain the context of the case against Microsoft.


9
Home and Recreational Software

Chapter 9 begins with Sonic the Hedgehog advertisement, which is mostly image, very few words compared to all previous ads.

Videogames and Computer Games: Early Development and Classification
(269-272) It is not an overstatement to say that Pong, produced by Atari, was the springboard for today's vast computer entertainment industry.
(272) By the early 1980s, rare was the American home with a teenage son that did not own a videogame console. Home computers were first sold in 1977, and most of them were used to play games.

Efforts of private citizens and societies have preserved early videogames far better than early corporate software artifacts.

(272) Unlike the early corporate software artifacts discussed in this book, most of which have evaporated into the mists of time, early videogames have been well preserved through the curatorial efforts of private citizens and preservation societies.

Arcade Games

Nod to Turkle for noticing new aggressive form of play inspired by videogames; can statistics of pinball machine production refute claim that videogames largely displaced them starting in 1974?

(273) By 1974 there were 100,000 Pong-type machines around the world. They largely displaced pinball machines, diverting the flow of coins from an old technology to newer one without much increasing the overall take.
(273-274) The devotees of Space Invaders were 90 percent male and 80 percent teenage, and the new aggressive form of play attracted the attention of Sherry
Turkle and other social theorists.

Videogame Consoles

Connect this brief history of videogame consoles to Montfort and Bogost.

(276) Videogame software cartridges, largely ignored by software industry analysts, became a huge business. It was estimated that five games were sold each year for every one of the millions of videogame consoles in existence.

Games for Home Computers

Bedroom coder phenomenon fueled by popular magazines; odd not to mention any Apple computer magazines.

(276-277) The most successful home computers sold in the millions, spawning huge user communities, mostly young male hobbyists. There was a vibrant newsstand literature, with titles such as Power/Play (for Commodore users), 99 Home Computer Magazine (for users of the Texas Instruments 99/4), Antic (for Atari 800 users), and Hot CoCo (for users of the TRS 80 Color Computer). These magazines typically sold 50,000-150,000 copies a month.
(277) No additional software development system was needed, there were no proprietary trade secrets to unlock, and programs could be duplicated on the computer itself, with no need for access to a third-party manufacturing plant. The lack of significant barriers to entry led to the phenomenon of the “
bedroom coder.”
(277) Parlaying early success into a long-term business required intuition, entrepreneurial talent, and access to venture capital. Those who succeeded were in no sense “accidental millionaires.” Of these success stories, that of Broderbund is perhaps the best documented.

The Crash of 1983
(280) The declining quality of games caused the teenage denizens of amusement arcades to look elsewhere for entertainment, and the arcades' proprietors had to find new customers.

Videogames Redux, 1983-1995
(280) It was almost universally agreed that the principal cause of the videogame crash was the flood of low-quality games that saturated the market in late 1982.

Managing Risk in the Videogame Software Industry
(281) Even a hit game usually had a market life of less than a year, whereas a word processor or a spreadsheet program had an indefinitely long life and provided a constant source of upgrade fees. Hence, the appropriate business model for a producer of videogames was closer to that of the recorded-music industry than to that of the personal computer software industry. . . . Relative to personal computer software, there were few lock-in effects to prevent a good game from becoming successful.
(281) The point of genre publishing was that certain categories of game appealed to large groups of users and therefore had less need to overcome market inertia.
(283) As with movies and recorded music, one way of increasing the likelihood of a game's becoming a hit was to improve the “creative inputs,” particularly by employing “star” programmers.
(283) Besides programmers, the creative team included video layout artists, sound and music directors, and script editors. The new creative process was much more capital intensive than the old. A side benefit of the increased capital intensity was to erect barriers to entry that reduced the number of players in the industry and marked the end of the bedroom coder.
(283) Within genre publishing, another way of reducing risk was by means of licensing—the videogame equivalent of Hollywood's “bankable star” system. Sports and movie tie-ins were the most common form of licensing.
(283) Licensing of arcade games was also successful.
(284) Electronic Arts and Activision, the two biggest players, have acquired very few firms in 20 years. In contrast, the leading firms in corporate and personal computer software have typically made a dozen or more acquisitions.

The Rise of the Japanese Videogame Console
(285) Because the [Nintendo] Famicom and the arcade machine used similar hardware, the player experience was essentially the same on either system (an unprecedented achievement), and many arcade gamers bought a Famicon simply to be able to play Super Mario Bros. at home.
(285) It is fair to say the Nintendo was solely responsible for the renaissance of the worldwide videogame industry. However, this had less to do with Mario Bros. as a “killer app” than with Nintendo's control of the software market for the NES. . . . This was achieved by incorporating inside each game cartridge a chip that acted like a key, unlocking the NES's circuits. Only Nintendo-manufactured cartridges had the chip.
(285) Many third-party developers were encouraged to produce games for the NES, but Nintendo was the sole publisher of games for its console.
(286) In any case, Nintendo's time in the sun was coming to an end with the arrival of 16-bit home consoles and Sega's secret weapon: Sonic the Hedgehog.
(287) More than any previous console, the Sony PlayStation was strategically positioned in the market.
(287-288) Sony provided unprecedented support in the form of development systems and software libraries, thus allowing developers to focus more on the game experience and less on the writing of code. . . . The use of CDs gave Sony a major advantage over its competitors. They were cheaper to manufacture than cartridges, and their short lead times permitted almost instant reproduction in response to market demand.
(288) The Sony PlayStation has dominated the videogame market since the mid 1990s.

Texts and technology link concerning videogames as cultural phenomenon.

(288) Today videogaming is an exploding cultural phenomenon with particular appeal to people under 30. Some authors believe that its longterm significance will one day rival that of movies or recorded music.

CD-ROM Encyclopedias: Content vs. Code
(289) Beneath the somewhat theoretical debate over what constitutes a software good lies an important phenomenon: the convergence of content and program code, of which the CD-ROM encyclopedia is one of the most visible examples.
(291)
Encarta is a remarkable example of Microsoft's ability to dominate a market with a thoroughly lackluster product.
(293) The availability of low-cost CD-ROM encyclopedias had a devastating effect on the market for traditional hard-bound encyclopedias.

Personal Finance Software: Quicken vs. Money
(294) Indeed, by the early 1990s Quicken had emerged as the best-selling consumer software product of all time, far exceeding the sales of any other software package, computer game, or multimedia product.
(295) Intuit was probably the first software company to focus on ensuring that a product would be suitable for immediate use, without recourse to training, manuals or telephone help lines.

Summary

Cultural diffusion of programming compared to diffusion of writing as publishing industry 150 years ago.

(301) Programming—broadly interpreted to include development of video games, creation of templates for spreadsheets, visual programming, and database design—is becoming a widely diffused skill. Indeed, it may be as widely diffused today as writing was 150 years ago. One does not speak of a “writing industry”; one speaks of a multi-sector publishing industry whose only point in common is putting ink on paper. In 20 years, we will likely think of the software industry in the same way.


10
Reflections on the Success of the US Software Industry

Chapter 10 begins with image of 60,000 punched cards for SAGE master program to reflect on US software industry success.

Castells connection on difficulty of reproducibility of US software success in other cultures.

(303) In the last 20 or 30 years, in Europe and many other parts of the world, planners have attempted to reproduce the success of Silicon Valley in their local economies.
(303) The cultural factors are usually the hardest to reproduce.

Early Start and Market Size
(304) Commercially produced software always has developed in response to the evolving technological capabilities of computers and their diffusion. . . . Because these technological developments always happened first in the United States, the American software industry has always been ahead.
(304) In short, the American computer for which the FACT compiler was being developed was so far ahead of the British machine that it opened up a new vista of programming possibilities.
(305) Thus, no matter how one interprets the numbers, the market for software products in the United States was at least 10 times as large as that in any European country. . . . In the early 1970s, when Europe had finally caught up with the United States in the diffusion of mainframe computers, major vendors such as Informatics, ADR, Pansophic, and MSA had mature products at reasonable prices against which national suppliers could not compete.
(305) The pattern persisted into the personal computer era.
(306) While would-be producers of shrink-wrapped software outside the United States were waiting for their domestic markets to mature, an entire business and publishing culture sprang up around American productivity programs.

Information Asymmetry and Clustering Effects
(306) The clustering of firms was a major factor in the success of Silicon Valley. . . . It is now accepted that a cluster of firms is much more dynamic and responsive than a giant corporation. . . . Clusters ultimately outperformed the “command economy” of vertically integrated corporations.

Network effects on knowledge and business interactions; firm clusters and social networks result in information assymmetry and advantage Silicon Valley and other centers (Castells).

(306-307) A complex network of social relationships, consolidated in professional meetings, bars, and health clubs, enables individuals working in different firms to know in what direction the market is moving and to develop the right product at the right time. This “information asymmetry” is what makes it so difficult for firms outside the network to compete. . . . The firm cluster allows technical specialists to move from one firm to another with better prospects almost overnight without breaking up their social networks or uprooting their families.
(307) Perhaps the only well-studied example of clustering in the software industry is that of the relationship of Northern California, home to Oracle, Sybase, and Informix. . . . A second example concerns the creation of complementary products in the PC software industry.

Government Support for Research and Development

Corporate software research and development less disciplined than in other industries.

(308) The importance of R&D in the success of the software industry is little understood. Software companies typically spend 10-15 percent of the revenues on R&D, well above the percentage in most other industries. . . . In the pharmaceutical industry, much of the R&D takes place in university and industrial research laboratories, and field trials are directed by PhD-qualified scientists. In the software industry, most R&D is done by youthful programmers, usually not trained past the bachelor's degree level, who crank out code in an intuitive but effective fashion—R&D with a small r and a large D.
(308-309) By far the most important publicly funded software R&D program was the US government's sponsorship of the Systems Development Corporation in connection with the SAGE defense system.
(309) While sales of personal computers were increasing exponentially, and while Bill Gates was making his second billion, the ACARD [UK's Advisory Council on Applied Research and Development] inquiry ignored the personal computer. And one can almost sense the disdain with which the committee would have greeted any suggestion that it pay attention to videogames.
(310) Within 5 years, nine of the ten leading British software companies examined by the ACARD committee had been acquired by overseas interests or gone bankrupt. Nothing in the report, even had it been implemented, would have had the slightest bearing on the outcome.

Summary

Manpower training as universal programming instruction in public schools may represent lost generations.

(311) Among the things a government can do to foster a software industry, manpower training may be the most beneficial and may create the least market distortion.



Campbell-Kelly, Martin. From Airline Reservations to Sonic the Hedgehog: A History of the Software Industry. Cambridge, Mass: MIT Press, 2003. Print.