Mary Croarken and Nathan Ensmenger, Editors
Because of space considerations in the January-March 2002 issue, we were unable to feature the following items in the Events and Sightings department.
CBI conference--"Unbundling history: The emergence of the software product"
September 2000, the Charles Babbage Institute (CBI) hosted a
conference, bringing together software company founders and historians
to discuss the formation of the software products industry and the
impact of IBM·s unbundling decision in 1969.
There is a delightful irony in the fact that software companies, which arguably comprise the most visible industry in the US, if not the world, are essentially first-generation businesses. That is, scarcely any software CEO has inherited his or her company from a parent. Of course, as noted by Computer Associates founder and CEO Charles Wang, the reason for this is that the industry is too young to have hand-me-down companies.
Wang was but one of many software pioneers assembled by the CBI at the Xerox Palo Alto Research Center in September 2000 to commemorate, recollect, and most importantly, interpret the significance of IBM's 1969 unbundling·separate pricing of software and hardware·on the emerging software products industry. His comment, made as a humorous aside, set a tone of personalizing industry events, presaging that the conference would not dwell just on numbers, statistics, or product specifics.
Repeating a well-known phrase, keynote speaker Wang said, "Victory has many fathers, but failure is an orphan." Many of those ·"fathers"· were among the 75 audience members who during the conference shared important memories, explained reasons for decisions, and described the impact of things that were both within and beyond their control. The conference was one of those rare events where the history makers and historians could directly interact, enriching the ongoing process of recording and interpreting the events of this dynamic industry.
CBI Director Arthur Norberg opened the conference by remarking that the CBI is "on the cusp of both a new decade and a new century, with new interpretations about the industry." The conference was divided into three segments: the first explored early products; the second, the scenario surrounding the deliberations about separate pricing; and finally, postunbundling growth.
This report on the CBI conference focuses on material that chronicles events of the 1960s and 1970s, which laid the foundation for products that would reach into every facet of commerce and culture. Although some of the speakers included information from the later PC and Internet eras, the principal subject was the formation of the industry, with IBM·s unbundling of hardware and software pricing as a fulcrum.
After examining the impact of IBM·s unbundling, many speakers concluded that there was no single moment of conception for the industry. Although unbundling is considered a validation of the software products industry, it is noteworthy that successful software products existed before 1969, the year when IBM signaled its change in direction.
In his keynote speech, Wang pointed out that around 50 years ago, when he became interested in the computer field, "The concept of the computer was so new that hardware companies had to provide the software in order to sell the machines," which, of course, were useless without it. "Like any new technology," he continued, "each manufacturer had its own approach to software, which at that time was truly a dark science." The result, Wang noted, was a proliferation of many different, closed, proprietary systems. Customers for a particular system had no choice but to comply, which led to account control by the manufacturers.
Wang made the case that account control created strong disincentives for hardware vendors to develop more efficient software solutions or to make their software faster or more user-friendly. In retrospect, Wang observed, inefficient software meant customers needed larger, more expensive computers. Hardware companies were able to generate even more revenue as a result, and by making system software complex and incompatible, software could remain proprietary, just like their hardware.
However, many customers had program requirements not met by the manufacturers. And, as nature abhors a vacuum, the independent software industry sprouted in the void. Hardware vendors initially failed to recognize this paradigm shift. Once they did, they moved too slowly, Wang asserted, "refusing to relinquish their proprietary model quickly enough to remain dominant."
The transition from custom programming services to software products was a long one, and some note that the pendulum is swinging back toward services today. Back in the mid-1960s, Wang recalled, he and his colleagues at the Standard Data service bureau recognized a phenomenon that may have nurtured most of the early software product ideas: Every customer for essentially the same application wanted something just a little different. Customers would promise "If you only do it this way ... we·ll buy," whatever the product was. So the developers at Standard Data (which Wang and his colleagues purchased in 1976 and which evolved into Computer Associates), as at other companies, "pulled a lot of all-nighters and wrote a lot of code."
Eventually, the Standard Data developers observed a condition that had been described approximately 60 years earlier as Pareto·s law, the 80-20 rule. As applied to computing, this law meant that 80 percent of the potential value of any business activity could be achieved from just 20 percent of the effort. The idea was to discover which 20 percent of the code fulfilled 80 percent of every customer's requirements and make that 20 percent the core of a software product (in the case of Standard Data, a generic file converter).
·"We figured if this solution helped us, it would help others, and maybe we should package it and sell it as a product," Wang explained. Other companies would later apply the same model to their own sets of priorities and develop new systems and applications products.
Martin Goetz opened the early-products segment of the conference, recollecting the story of Autoflow and other product offerings of Applied Data Research, the company that he and six other individuals founded in 1959. Like other software companies, ADR began its early years as a custom programming shop with fixed-price contracts. "We originally had no intention of marketing program products," he said, "as software was either free or exchanged." The company's 1965 initial public offering, which raised the healthy sum of $5 million, provided the money to develop the market for ADR's first actual software product, an automatic flowcharting system called Autoflow. Among its many distinctions, Autoflow would eventually become the second software product to be patented, after Goetz's "Sorting System," also from ADR. (See Goetz's memoir about his experiences in the 1950s and 1960s in the January-March 2002 issue.)
Next at the podium, at least figuratively (as he was unable to travel to the conference), was John Postley, progenitor of the Mark IV product, certainly one of the first general-use programs of any kind. By telephone, he related his recollections of Informatics and the evolution of Mark IV. Postley's narration at the conference was part of his memoir published in IEEE Annals of the History of Computing in 1998 ("Mark IV: Evolution of the Software Product, a Memoir," vol. 20, no. 1, pp. 43-50).
Postley related that, in 1960, he and Bob Hayes formed a company called Advanced Information Systems (AIS), whose first product was a tape retrieval system they built for Douglas Aircraft. This first work by AIS on a file management system was designed to retrieve data from files that were stored in several nonstandard formats. Funding came in increments of $3,000 to $5,000, and there was no formal contract. His presentation (and the 1998 Annals article) traced the Mark product family evolution from the custom project for Douglas through the company·s acquisition by Howard Hughes. Ultimately, that company (Hughes Dynamics) decided it did not want to be in the software business, and it sold AIS to Walter Bauer's new company, Informatics.
In concluding the early-products session, Duane Whitlow related how his work on a data management system yielded one of the first·and most successful·competitors to IBM software. In fact, the generalized sorting program sold by Whitlow Computer Systems was so successful against IBM's "free" software that the company's name was changed to that of the product, Syncsort. It is still installed in thousands of companies today.
Whitlow said he had perceived an "essential flaw" in the IBM software that was already used on many systems:
IBM appeared to have taken a tape sort and modified it so that the data that was previously stored on tapes was stored on disk. I thought that if a sort could produce better times, it would take hold in the market.
Technically, Whitlow explained, a major execution delay in a sort results from waiting for the next set of records to be brought into memory or waiting for a write to complete, or both. He continued,
One improvement would be to provide buffers to anticipate records for input as well as a spare for output. But because the buffers used, and thus reduced, memory available for the other functions needed, they also reduced performance.
capability Whitlow sought resulted in synchronizing those activities of
reading and writing to disk, such that when a block of record was
written, memory space allowed a block to be read (thus the name
Syncsort). This approach also dramatically reduced the disk space
required and the number of disk accesses.
"Discovering this was an incredible experience. We were both frightened and exhilarated," he recalled:
I knew it was patentable because Marty Goetz had started the ball rolling in that direction. We had found one instruction in the System/360 computer that made a dramatic improvement in performance. Although there were significant other aspects of the sort that are included in the patent, I knew this aspect was a breakthrough and anchored the patentable methodology.
individual speakers, the first session culminated in a panel
discussion, during which Luanne Johnson, director of the Software
History Center, observed that bundling was really about account control
more than about direct business. Customers were "very reluctant" to make
their account managers angry by going outside IBM, she observed,
adding, "Syncsort helped break down that barrier." Even after
unbundling, she asserted, IBM offered software by lease-only and cheaper
than that from independent software vendors. Customers believed that
IBM could produce anything it wanted to, a fact that discouraged them
from buying from would-be competitors. So the issue was much more
complicated than just "free" software.
Informatics founder Walter Bauer, also on hand for the panel, reflected that Mark IV received the first "Million Dollar Award" from International Computer Programs, in 1971. In fact, ICP would also award the first $10 million and first $100 million citations to Mark IV.
Software unbundling: The fulcrum
Many observers, participants, and historians consider unbundling to be a pivotal point in establishing the software products industry. Having heard from competitors of IBM's free·though ultimately paid-for·software products, in the second conference session attendees heard from someone who was at IBM at the time, and who, in fact, participated in many of the company's software unbundling decisions.
Watts S. Humphrey, who now heads the Software Engineering Institute at Carnegie Mellon University, described his personal perspectives on what IBM did, and why, from the vantage point of a manager and executive in the IBM systems development organization. (Humphrey's article on this topic appears in the January·March 2002 issue.)
The second speaker in this session provided a more formal economic analysis of unbundling's impact. Stephen Usselman, a professor at the Georgia Institute of Technology, suggested that it is not clear whether unbundling itself would have occurred without the threat, and indeed the reality, of an antitrust suit. He believes that market forces would have resulted in unbundling anyway, but it remains an open question as to when this would have happened on a large scale without the threat of the Department of Justice suit.
Usselman's presentation made the following observations:
We can say that antitrust may well have hastened the emergence of the software industry during the late 60s and early 70s, if perhaps a bit inadvertently. The action by the US government certainly did nothing to impede the forces propelling the emergence of a separate and vibrant services sector, and [the] Justice [Department] may well have hastened the process by causing IBM to separate its services offerings.
Usselman said that separation, however, was balanced to some degree by increased marketing by IBM of standard programs.
Usselman also said that the packaged program sector lagged behind services and the projections of observers such as Walter Bauer, who had anticipated that independent software providers would soon overtake IBM in this rapidly expanding field during the late 1970s. That development would eventually come to pass but only after the PC's emergence during the early 1980s.
Did IBM's decisions about the use of unbundled software for its PCs, made during the final stages of the long antitrust suit against it, reflect a concern deep within IBM about tying up the industry with closed, proprietary systems? Usselman asked. "After years of antitrust surveillance, had IBM adjusted its thinking and strategies to the point that it instinctively unbundled? Perhaps, but more likely, the firm simply failed to anticipate the future." As one executive close to the situation recalled, he continued, "the thinking of top management was that the PC was a small market, and one cannot make big mistakes in small markets."
In the postsession discussion, Burton Grad, who was one of the three people working on software in the 1969 unbundling task force, pointed out that IBM's primary emphasis was on systems programs, not applications. Even so, "during and after unbundling, IBM continued to believe that its fortunes would come from hardware," he commented. Indeed, he said, "Some people opposed any application announcement that did not support the requirement for new hardware" and there was no support for old versions of CICS, for instance. "IBM did not make money on applications, but it certainly did on systems-type priced programs" (for example, CICS and VM/CMS).
Whereas many speakers, including Usselman, think of the economic impact of unbundling as applying primarily to an emerging industry of independent companies, there was significant impact on IBM itself. The 3 percent decrease in hardware revenues came right out of IBM's revenue line, Grad said, perhaps meaning a 10 to 15 percent decrease in actual profit.
Postunbundling: The emergence of the software product
Although the software industry had a solid foundation that was actually laid in the 1950s, it was the 1960s, with IBM's unbundling (and, perhaps more importantly, its earlier introduction of the System/360 family of computers with two primary operating systems) that set the stage for growth. About five years after the pivotal 1969 unbundling event, Larry Welke's International Computer Programs identified 1,229 industry-specific applications products. And, although one might define applications as the reason one would purchase a computer, it should also be noted that the size of the systems software business has typically been equal to or larger than that of applications.
Also in 1974, Peter Cunningham founded INPUT (http://www.input.com), a respected market research and analysis firm concentrating on the software and services industry. Cunningham proposed that true "unbundling" would have meant that a computer vendor (IBM) establish a separate software product business organization, with its own marketing, sales, and discrete pricing models.
One other significant issue then was the very availability of software. Cunningham said, "it was often simply not there, even though there was a willing market." That the market need outstripped the industry's ability to provide solutions was not for lack of trying: R&D spending was about 20 to 25 percent of total costs in software vendor companies, twice the expenditures for R&D by systems companies, he said. "Reinvesting in products and services is really what enabled the software industry to thrive.
For more than another decade, until PCs took hold, virtually all the leading software companies were the computer manufacturers, as Table 1 shows.
|Table 1. Largest US vendors of software products in 1979. (Copyright INPUT 1980)|
|Rank||Name||Software product revenue ($ millions)|
|Control Data Corporation (CDC)|| |
|Digital Equipment Corp.|| |
|National Cash Register (NCR)|| |
|Management Science America (now part of GEAC)|| |
|Informatics (now part of Computer Associates)|| |
In 1980, INPUT forecast that by 1985, 75 percent of large-company users would use PCs, a bold prediction for its time but, as it turned out, a conservative one. The PC represented the beginning of a demand for white-collar productivity tools. The 1981 debut of the IBM Personal Computer fundamentally changed IT, essentially beginning a new era.
The next speaker in this session was Martin Campbell-Kelly, reader (professor, in US parlance) of computer science at the University of Warwick, England, presenting an economic analysis of the software products industry. He commented,
It may be true that software is the most visible industry in the world today, and it is tempting to conclude that software is unlike any other. Given that the Internet, and before that the PC, and before that the mainframe, each in its era changed the way business was conducted, it is tempting to conclude that there may be nothing to which one can compare software.
contrary, he asserted. Although software may look like an industry
different from any other, "it is important, and indeed possible, to
apply standard business rules and models to the software industry."
That said, Campbell-Kelly observed that one of the problems in studying the software products business is the dispersion of companies in the industry. Unlike cars or computers, there have been thousands of software companies, both private and publicly owned, and many of them have almost vanished from history. Fortunately, there are a handful of corporate histories from companies like Informatics.
Obtaining accurate financial statistics from companies that are no longer around·or even from some that are·is a daunting task. Campbell-Kelly said that companies should follow Informatics' example and produce corporate histories, and he encouraged them to post financial data on their Web sites to facilitate research in the field. Software companies should emulate other industries and make their financial data more accessible for economic analysis, he said.
Previewing his forthcoming book The History of the Software Industry, Campbell-Kelly stressed that there is so much software embedded in all kinds of products, from printers to storage devices, that getting a firm estimate of the industry size is virtually impossible, particularly without cooperation from the companies themselves.
As an overall technology, the software industry likely began with custom programming, which started around 1955, he related; within a human generation, there became thousands of shrink-wrapped products for the PC. The next phase will be the Internet era and paradigm, he suggested.
Complicating the research effort, he continued, various professions from engineering to consulting are software-intensive, yet companies do not put the salary costs under the heading of software:
You can't get a measure of the cost of the software effort. In construction, there are tangible things to measure (lumber, for example), but this is not the case with software. Even at HP, 70 percent of the development effort in printers is said to be for software.
exact numbers are elusive or nebulous, by 1988, nearly 20 years after
unbundling and 10-plus years into the PC era, 15 percent of
data-processing budgets were spent on software products, Campbell-Kelly
Because software has permeated every facet of business to the point that it has become impossible to estimate software's economic impact, the conference's concentration on software products as a discrete industry made perfect sense. With the acceptance of packaging, pricing, and production processes that mirror other "hard" products (not to ignore patenting), the conference helped document the reality of software, described by Peter Cunningham as the "lifeblood of business," and software products as the mainstay.
Would the industry be such a major force without IBM's unbundling? No one could say for sure, but unbundling certainly created visibility for software products, building on the foundation of the pioneers and helping to launch the most visible and fast-changing industry of the 20th (and likely the 21st) century.
As noted at the outset, perhaps no other industry is young enough to convene a conference where so many of its history makers could be present. One can be forgiven for recalling Walter Cronkite's closing words to his popular 1960s television show, "You Are There," which gave viewers vicarious representations of historical events: "It was a day like all days, filled with those events that alter and illuminate our times. And, you were there." Only in this case, the history makers literally were.
Russian Pioneer Day
8-11 October 2001, the Siberian Branch of the Russian Academy of
Sciences held an International Conference devoted to the 90th birthday
of Alexei Andreevich Lyapunov (1911-1973). The conference, held in
Novosibirsk, brought together around 200 of Lyapunov·s colleagues and
followers who presented and discussed more than 150 papers related to
various cybernetic disciplines originated by Lyapunov.
Designated Pioneer Day, 8 October was devoted entirely to the memory of Lyapunov, who was born 8 October 1911 in Moscow. During the last 11 years of his life--one of his most productive periods--Lyapunov lived and worked in the Siberian Scientific Center of the USSR (Union of Soviet Socialist Republics) Academy of Sciences, outside the city of Novosibirsk. Like Norbert Wiener, Lyapunov had a thorough understanding of disparate areas of knowledge. Biologists, geophysicists, and philosophers all considered Lyapunov one of their own. His erudition and encyclopedic knowledge of science, combined with his integrative, global approach to natural sciences, provided the background for his ideas about cybernetics.
At the beginning of the Pioneer Day session, Gyozo Kovacs made a brief introduction, describing the international computer community's tradition of celebrating Computer Pioneers and previous Pioneer Days. Lyapunov's daughter, Natalya Alekseevna Lyapunova, then reminisced about her father. She showed the audience Lyapunov's Computer Pioneer Award medal--with which the IEEE Computer Society had honored Lyapunov in 1996 for his founding of Soviet cybernetics and programming.
Several other memorial papers were presented during Pioneer Day. These included "Aleksey Andreevich Lyapunov in the History of Russian Science," by R. Podlovchenko; "A.A. Lyapunov and Mathematical Biology," by A. Fedotov; and "The Physiological Aspects of Lyapunov's Work in Cybernetics," by V. Fedorov.
Pioneer Day concluded with a recently restored documentary film about Lyapunov--The Way to the Stars, which was originally produced in 1966 by French television. Conferees watched as Lyapunov, who had an excellent grasp of French, showed French guests around the Siberian Scientific Center at Akademgorodok.
A new book about Lyapunov1 was published just before the conference, and each Pioneer Day participant received a copy as a gift. The book contains about 60 articles, 200 letters, and numerous photos from Lyapunov·s archives. Most of these materials are published for the first time.
During the rest of the conference, numerous scientific papers were discussed at five panels corresponding to the main areas of Lyapunov's interests. These areas were mathematical cybernetics, programming, mathematical modeling in biology, cybernetics in physiology, and prospective problems of education.
Throughout the conference, a large exhibition was open in the Academy of Sciences' Scientists' House in Novosibirsk. On display were Lyapunov's own works, literature about Lyapunov, and rare artifacts from the Lyapunov memorial museum.
References and notes
Early computer mouse encounters
17 October 2001, Xerox PARC in Palo Alto, California, hosted the "Early
Computer Mouse Encounters" seminar, sponsored by the Computer History
Museum and the Swiss Science and Technology Office in San Francisco. The
seminar brought together six early developers of computer mice. In the
panel format, each speaker spoke for approximately 10 minutes and
answered questions afterward. Daniel Borel, a Logitech cofounder, noted
the tight coupling of both research and ideas on mice (and human-machine
interfaces generally) between Stanford University, the ETH (Swiss
Federal Institute of Technology) in Zurich, Xerox PARC, and SRI
International. Doug Engelbart spoke next on his lifelong work in
boosting the collective ability of groups to solve problems. In a manner
reminiscent of Jay Forrester's perfecting of workable core memory,
Engelbart's mouse was almost an afterthought, or certainly an ancillary
invention to this larger goal. In particular, Engelbart's lifework on
"improving the improvement infrastructure" of organizations--which he
calls bootstrapping--has been the true focus of his research, although
he is known (somewhat unfortunately, to his way of thinking) as the
"father of the mouse."
Bill English, one of Engelbart's colleagues at SRI, described the desire for a good pointing device to select text (in contrast to graphical elements) as the impetus behind the mouse. English convincingly demonstrated how the search for a useful input device was approached scientifically, measuring dwell and access times for typical users and input devices, such as the trackball, the Graphicon, foot controls, and even a knee mouse. The mouse, essentially as we know it today, won handily in measures of bit rate. English then described various mice variants--mice using wheels, balls, and the somewhat successful optical mouse, the latter requiring a special mouse pad on which a Cartesian coordinate system had been inscribed. Jean-Daniel Nicoud, a professor at ETH Zurich, then discussed his interesting attempts at creating mice, including a 1973 visit to Engelbart in Palo Alto. He also showed a variety of ingenious miniaturized devices he had constructed over the years.
Stuart Card, another Engelbart colleague at SRI, spoke on the specific scientific testing behind the mouse. Particularly interesting was his observation that when using the "Engelbartian interface" (chordset, keyboard, and mouse) users were so adept at inputting and manipulating information that they had to be videotaped and watched again in slow motion. Another interesting cultural observation was the US Social Security Administration's rejecting of the mouse for its offices because of concerns clients might hang themselves with its cord! The one simplifying principle behind all the mouse testing was that input devices had to employ the "high-bandwidth" muscles of humans for effectiveness. This measure made a head mouse, for example, untenable as people ended up with almost crippling neck strain injuries after only a week of use.
Niklaus Wirth spoke about his year-long visit to Stanford and how the Xerox PARC Alto computer "was a revelation" for him in its ability to change dramatically how people undertook their daily work. Not wishing to return to the batch processing of his native institution in Switzerland, he decided to build a "Swiss Alto," the Lilith.
Finally, Daniel Borel concluded with some thoughts on the corporate success of the mouse under the firm he cofounded, Logitech. The company's first product was the mouse designed by Jean-Daniel Nicoud and Rene Sommer.
In the question period, Wirth noted that "huge manuals are a sign of huge inadequacy" in terms of usability. Engelbart noted that we "are still in the horse and buggy stage" of user interfaces.
For more information, please visit http://www.computerhistory.org/events/index.php?id=1090020693 To see the original lecture announcement, including speaker bios, visit http://www.computerhistory.org/events/lectures/mouse_10172001/.
Contributors to this online issue of Events and Sightings are Edward Bride, Yakov Fet, Computing Center of the Siberian Div. of RAS; Gyozo Kovacs, John von Neumann Computer Society; and Dag Spicer, Computer History Museum.