Sharing Time: Tymshare, 1965

Greenberger’s article was in part an outgrowth of his organizing a conference several years earlier on computing’s future as part of celebratory events for MIT’s 100th anniversary—participants at this conference included many of the foremost figures in computing: Norbert Wiener, Vannevar Bush, Marvin Minsky, Jay Forrester, Robert Fano, Herbert Simon, Grace Hopper, Alan Perlis, Claude Shannon, John Kemeny, and Gene Amdahl.⁵ Greenberger’s views undoubtedly were also shaped by his witnessing pioneering digital computer, programming, and networking accomplishments at MIT for the past decade—Lincoln Laboratory, SAGE, Compatible Time-Sharing System (CTSS—operational in 1961, the first significant time-sharing system), and the 1963 ARPA IPTO-supported Project MAC (Multiple Access Computer/Machine-Aided Cognition—which funded development of CTSS’s ambitious follow-on Multics time-sharing system as well as artificial intelligence research).⁶

Greenberger’s “information utility,” for individual home users did not materialize during the time-sharing era (from the mid 1960s to the mid 1980s). The computer utility concept for home users arguably has been realized to a degree in the World Wide Web era with cloud services, but this has fallen far short of complete standardization and uniformity we associate with utilities. Nonetheless, on certain individual predictions, he was quite prescient. He forecast time-sharing providers would serve many businesses and industries and offer custom programming and software products (the latter a precursor to recent Software as a Service, or SaaS). With regard to suppliers of time sharing, Greenberger was correct that the time-sharing industry would be dominated by a few firms.⁷ What he did not anticipate was that this would only be true after a shakeout of roughly 100 startups, and that one of these survivors, Tymshare, Inc., would become the second largest time-sharing company in the world. The largest, General Electric (GE), sold its mainframe computer business in 1970 to Honeywell, but retained and expanded GE Information Services, or GEIS, which for two decades focused on time-sharing services.

Time sharing was a disruptive technology that sparked many debates in trade and academic publications in the second half of the 1960s regarding the relative efficiencies of batch processing and time sharing. Ultimately, underlying goals for applications and how computers were employed made the most difference as to whether users succeeded at optimizing processing resources. As was mentioned briefly in chapter 3, in the mid 1960s the computer industry consultant Herbert Grosch made an observation famously labeled as Grosch’s Law—that stated the costs of computer systems increase at a rate equivalent to the square root of their power. This statement on economies of scale—the most powerful computers being more efficient—certainly highlighted the importance of supercomputing and utilization. It also lent some weight to the time-sharing model and computer utility idea. At times, computer use was purely a question of efficiency, but there were also qualitative differences in expanding experiences as true “users.” And while batch was efficient for processing data, time sharing offered programmers greater efficiency in perfecting and debugging software. Time sharing of the 1960s facilitated a level of user interaction with machines not previously possible (as computer operators tended to run machines in batch environments)—and not equaled or exceeded before the era of the personal computer.

Time sharing was adopted by most of the largest service bureaus during the middle to late 1960s. This included IBM’s subsidiary Service Bureau Corporation, Control Data Corporation’s Data Services Division/Professional Services Division, and programming services specialist C-E-I-R (which CDC acquired in 1967). These enterprises, which appealed particularly to scientific and engineering computing users, typically would offer time sharing along with traditional batch processing, and networking, remote batch processing (batch with input-output delivery through telephone line-based networks rather than mail or courier). Over time, the time-sharing industry redefined and to a significant degree subsumed the service bureau industry. Non time-sharing service bureaus continued to exist in the 1970s and in subsequent years, but tended to be small stand-alone centers. And a meaningful portion of “time-sharing” firms’ revenue generally was derived from services that did not necessarily utilize time-sharing applications.

IBM offered limited forms of time sharing in the mid 1960s, but these efforts had minimal impact. The company only entered strongly into the field with its “CALL 360” service in 1968, which it quickly transferred to subsidiary SBC because of scrutiny by the US Department of Justice.⁸ Time sharing had both offensive and defensive strategic elements for IBM. A thriving time-sharing industry could reduce the demand for customer computer installations—some firms opting for purchasing computer time on their premises rather than installing a purchased or leased computer, or using time sharing for periods of peak demand to avoid obtaining additional computers. Restraining the development of an independent time-sharing industry was in IBM’s interest. At the same time, time sharing was an opportunity to tie firms not yet able to buy or lease IBM equipment to IBM’s family of compatible computers. IBM’s lack of foresight and planning with time sharing for its planned System/360 was a major factor why Fernando Corbató and others at MIT (who had used an IBM 7094 on CTSS) rejected IBM in favor of GE for hardware on Project MAC’s Multics. IBM’s Time-Sharing System for System/360 Model 67 was not released until October 1967, shortly before CALL 360 was introduced as a time-shared service for the IBM System/360 (models 40 and 50).⁹

As historian Martin Campbell-Kelly and economist Daniel Garcia-Swartz argue in an important article on the economic history of the time-sharing industry, this segment has been marginalized by scholars and was highly significant to the overall computer industry.¹⁰ This chapter seeks to complement their work with a business history-focused case study of Tymshare, Inc., a remarkable firm launched in 1965 very shortly after industry leader GE essentially kicked off the time-sharing industry (GEIS).¹¹ It briefly discusses the academic (and nonprofit research corporation) origins of early time sharing as well as GEIS. It then focuses attention on Tymshare before concluding with highlights on a broader industry context.

Computer Time Sharing’s Academic Origins

The concept of time sharing emerged in the late 1950s as a new and redefined instance of certain elements that had occurred years earlier—using a computer at a distance (a lone individual using a George Stibitz early 1940s relay computer more than 100 miles away) and multiple simultaneous system users using a mainframe for a single purpose (terminals with cathode ray monitors networked to SAGE computers in the late 1950s). Time sharing differed in that it involved multiple users for different applications simultaneously sharing a computer. Conceptually, two early articulations of time sharing stand out. In 1959 National Research Development Corporation’s (Great Britain) Christopher Strachey presented a paper (at a UNESCO conference) on debugging a computer while it was used by another individual in normal application mode. Independently, earlier that year, MIT’s John McCarthy circulated an influential MIT internal memorandum with multiple debuggers simultaneously utilizing a modified IBM 709 system.¹²

Many important figures contributed to MIT’s pioneering work with time sharing. Herbert Teager and subsequently John McCarthy led MIT computer planning committees that first discussed time sharing and launched an early time-sharing project. Robert Fano, the first leader of Project MAC, was a major force with MIT’s time-sharing research and development.¹³ MIT Computation Center scientist Fernando Corbató, however, stands out in becoming the principal leader of both CTSS and Multics, the school’s influential time-sharing systems. The latter effort began as a partnership between MIT, Bell Laboratories, and the electronics conglomerate and computer manufacturer General Electric.¹⁴

CTSS had no true rival in late 1961 when it was first successfully tested (at least in the open community outside the confines of the National Security Agency and the classified world). Bolt Beranek and Newman, Inc. had a complete DEC PDP-1–based time-sharing system (that both John McCarthy and J. C. R. Licklider helped develop) in September 1962.¹⁵ And System Development Corporation (SDC) developed an important time-sharing system for AN/FSQ-32 (Q-32) computers operational in the middle of 1963, which received an enhancement contract from ARPA IPTO Director J. C. R. Licklider. Licklider and his successors made time sharing a major focus of IPTO funding in the early to mid 1960s and hoped Project MAC could influence developments on the East Coast and SDC and the Q-32 time-sharing system could do likewise for the West Coast.¹⁶ As part of this effort, Licklider also funded Project Genie at the University of California at Berkeley to develop a time-sharing system for a Scientific Data Systems (SDS) 930.

Figure 7.1 A Q-32 time-sharing installation in the Santa Monica headquarters of System Development Corporation. Courtesy of Charles Babbage Institute, University of Minnesota.

In 1964, as IPTO funded Project Genie, the National Science Foundation funded John Kemeny and Thomas Kurtz of Dartmouth College to further an effort they had conceptualized a couple years earlier: the Dartmouth Time-Sharing System (DTSS). The Dartmouth System, operable by mid 1964, initially used a donated GE 225 and GE Datanet-30 message switching machines—the latter, the corporation had developed for Chrysler and could accommodate 40 simultaneous users.¹⁷ DTSS, which by 1966 accommodated about two hundred simultaneous users, also benefited from the corresponding programming language that Kemeny and Kurtz created for the system, Beginners All-purpose Symbolic Instruction Code (BASIC). Among the computer manufacturers, GE, SDS, and DEC played early leadership roles with time-sharing hardware, GE standing out in becoming the first major time-sharing services provider.

In the second half of 1965, working closely with Kemeny, GE designed a parallel system to DTSS. GE Corporate Headquarters in New York City and selected GE computer customers were trial users late that year. General Electric’s Information Processing Center (IPC) was the services side of the GE Computer Department, and initially focused on augmenting hardware contracts—it was losing money as a batch processing services business. IPC’s new manager in 1965, Warner Sinback, however, succeeded in launching a time-sharing enterprise during his first year at the helm, and importantly, gained approval to separate it from GE’s Computer Department (significant for Simback to have free reign as a manger and also because GE several years later would sell its Computer Department to Honeywell). In January 1966, with the formal split with the Computer Department, the business unit was renamed GE Information Services (GEIS) and moved to Bethesda, Maryland. Early customers included Hughes Aircraft, Chrysler, Boeing Aircraft, Bechtel Corporation, and AT&T.¹⁸

As Warner Sinback recalled, IPC/GEIS ran an inexpensive full-page advertisement in a small-circulation accounting trade publication with the heading “Would you like access to a $1 million Computer?” According to Sinback, IBM executives were “so perturbed that they ran [an ad] in the Wall Street Journal within a month that said ‘Would you like access to a $5 million dollar computer?’”¹⁹ Just as Ross Perot’s facilities management model threatened IBM’s relationship with customers and offered more efficient usage of expensive processing time, so too did the new time-sharing business. As United Data Centers founder and future Tymshare executive Bernard Goldstein put it, time-sharing companies were “offering raw machine time for firms to avoid installing their own computers.”²⁰

In the second half of the 1960s, GEIS aggressively entered most major metropolitan areas of the US. By late 1968, GEIS had secured 40 percent of the $70 million time-sharing market. It developed a nationwide network with a few regional hubs to effectively utilize computer processing resources.²¹ Although GEIS stands out, oral histories, archival documents, and other resources provide a unique opportunity for a richer analysis of a highly innovative Silicon Valley startup—Tymshare, Inc.—that created an equally influential network (TYMNET), grew rapidly, and became the longtime and quite formidable number-two firm in an industry that evolved into an oligopoly.

Tymshare’s Origins

Tymshare originated at the hands of Thomas O’Rourke and David Schmidt, who left General Electric to launch the startup time-sharing company in 1965. O’Rourke, the more senior of the two, had an electrical engineering degree from University of Washington and had been a district manager in Los Angeles for GE’s Computer Department. He rose to become a regional manager for the entire West Coast (working out of the Phoenix Computer Department headquarters and later Sunnyvale, California) before losing the position to the son-in-law of one of GE’s vice presidents. As a concession, O’Rourke was offered a position running GE’s military electronics group in Washington. In May 1965, having moved his family many times during his four years in the Air Force and his thirteen years with GE, and wanting to stay on the West Coast (and near Sunnyvale), O’Rourke left GE. He recruited David Schmidt, a top programmer, to join him in forming Tymshare Associates.²²

O’Rourke had been an advocate of donating computers to Dartmouth eighteen months earlier and had carefully followed the DTSS project. He believed there was a viable business opportunity in time sharing. Both O’Rourke and Schmidt withdrew their retirement holdings and put up their life savings—$15,000 and $10,000 respectively. O’Rourke, having worked on ERMA in Los Angeles in the late 1950s, contacted Bank of America’s George Quist (later a founder of influential venture capital firm Hambrecht and Quist, subsequent underwriters for Apple Computer, Netscape, and Amazon.com). Bank of America and Quist provided half a million dollars in venture capital for a 50 percent ownership stake.²³

O’Rourke and Schmidt ordered a GE 225, which was to be delivered in the fall of 1965, but the order was canceled and the down payment returned when GE informed them that their credit was insufficient. O’Rourke believed that the true reason for the cancellation was that GE had decided to enter the time-sharing business and didn’t want to sell a machine to a future competitor.²⁴ Warner Sinback of GE offered O’Rourke a senior post in helping to run GE’s time-sharing business, but O’Rourke had already launched Tymshare, partnered with Schmidt, and taken venture funding; there was no turning back.²⁵

In January 1966, with the new name Tymshare, Inc., the company was formally incorporated in California, setting up shop in dilapidated rented office space on Distel Drive in Los Altos. The lone bathroom did not have a roof.²⁶ O’Rourke was president and chairman of the board; Schmidt was executive vice president and became the general manager for technology.

O’Rourke asked the distinguished venture capitalist Arthur Rock to provide additional financing to secure a Scientific Data Systems (SDS) computer. Though Rock did not invest directly, he helped get an initially reluctant Max Pavlesky, SDS’s president, to offer a generous lease for a new model SDS 940 (then under development); it was to be delivered in the second half of 1966.²⁷ The SDS 940 was derivative of pioneering work done by Mel Pirtle, Butler Lampson, Peter Deutsches, and others at the UC Berkeley to modify an SDS 930.²⁸ As part of the agreement with SDS, Tymshare promised to help line up four other SDS 940 sales. O’Rourke delivered on that promise by arranging for the acquisition of one SDS 940 each by two other time-sharing startups (Comshare of Ann Arbor, Michigan, and Dial-Data of Newton, Massachusetts) and one each by SRI and Shell Oil.²⁹ Initially, owing to expensive long-distance telecommunications/data charges, time sharing was thought of as a local business—an office serving a metropolitan area. Thus, at the start Tymshare and Comshare were not direct competitors, and they partnered on early efforts in systems programming.

9375_003_T7.1 — Table 7.1 Established firms and other enterprises in the time-sharing industry, 1963–1966.

Before Tymshare got its SDS 940, both Tymshare and Comshare accessed the University of California’s SDS 930 between 2 and 6 a.m.—the only time the machine was available to them. Comshare’s leader, Robert Guise, sent Rick Crandall and his wife, both Comshare programmers, to team up with the Tymshare programmers Ann Hardy and Verne Van Vlear, the latter a former systems analyst from GE.³⁰ Along with O’Rourke, Schmidt, Neil Sullivan (a marketing specialist), and a secretary, this was Tymshare’s entire workforce for months. As they designed and built a system with commercial applications in mind, the small two-company software development team benefited from the time-sharing operating system designed and programmed by two University of California doctoral students, Peter Deutsches and Butler Lampson: the Project Genie Berkeley Time-Sharing System.³¹

Ann Hardy had senior technical and managerial responsibilities virtually from the start at Tymshare. She later became the first female vice president of the company, a position she held for roughly a decade, and was one of the earliest female vice presidents of a major IT firm. Hardy was a Chicago native who had moved to New York to take education courses at Columbia University after completing her bachelor’s degree at Pomona College. On the advice of a friend, she took IBM’s Programmer Aptitude Test in 1956 and was hired as a programmer to work on the STRETCH supercomputer development project in Poughkeepsie. Lawrence Radiation Laboratory (in Livermore, California), after acquiring a STRETCH in 1961, soon hired Hardy. After seeing a Tymshare advertisement in Datamation at the start of 1966, she phoned the office of Tymshare, Inc. to inquire into the possibility of leasing a terminal for the Hardys’ house (her husband, Norman, also worked at the lab) to enable them to work remotely. Tymshare’s leaders had never thought of a home connection, and the conversation quickly took a different path.³² Ann Hardy soon visited Tymshare, and was hired in February 1966. She had the task of programming (along with Verne Van Vlear) a time-sharing operating system with a “Monitor” (an effort managed by Hardy) and an “Exec” (managed by Van Vlear).³³ This work (aided by the Crandalls from Comshare) was done first on UC Berkeley’s modified SDS 930, and later in the year on a prototype at SDS in Santa Monica (to enable time-sharing operability for the forthcoming SDS 940).³⁴

When the SDS 940 was delivered in fall of 1966, Tymshare moved to larger space, on East Meadow Drive in Palo Alto, that accommodated the mainframe. The firm’s rapidly growing staff soon included Norman Hardy (who joined in 1967) and several influential sales managers.³⁵

At the start, O’Rourke concentrated on marketing to engineering firms on the peninsula and Bay area. This limited geographical focus had the advantage of avoiding or lessening long-distance charges. Tymshare—like GE—initially provided free time on an experimental basis, but switched customers to pay service by the last months of 1966. In doing so, O’Rourke could then show his investors that checks were being sent and that time sharing, in fact, was a viable business.³⁶

GE had been aggressive in making Berkeley their second time-sharing location (after Phoenix) in order to compete directly with Tymshare. In concentrating on engineering applications, Tymshare was able to secure a small group of customers. Owing to Tymshare’s initial focus on engineering, Los Angeles aerospace companies, such as Lockheed, became a major base for the company. GE’s discounting or eliminating phone charges to Southern California firms pushed Tymshare to quickly open their second office in the Los Angeles suburb of Inglewood in early 1967.³⁷ Later that year they expanded to the East Coast with an office in Englewood Cliffs, New Jersey—initially served via phone lines, but by early 1968 possessing a SDS 940.³⁸

In 1967 Tymshare was the first vendor for a time-sharing operating system for the SDS 940 that could pass stringent “acceptance tests” of Harvard University—they did this in a matter of several days after numerous others over the past five months had failed. As Ann Hardy remembers, the SDS salesman thought he could “never make the sale” to Harvard until the Tymshare team arrived and aced the tests.³⁹ Before the end of 1967, Tymshare was also aggressively targeting business data processing applications.⁴⁰

In its early years. Tymshare benefited from having as sales managers John Jerrehian, Ray Wakeman and Ron Braniff, who each came to manage one of its first three sales regions, Northern California, Southern California, and the Northeast. O’Rourke worked closely with the sale managers.⁴¹ Braniff, like others, emphasized O’Rourke’s personable nature, straightforward style, and high expectations for his employees. Braniff described O’Rourke as tough, ethical, informal, and fun to be around. “[T]he culture of our company,” he said, “grew up around his personality.”⁴² This continued even as the company expanded, the combined staff growing to more than 100 for Tymshare’s three locations (with a total of five SDS 940s) during 1968.⁴³

While O’Rourke focused on sales, Schmidt concentrated on overseeing the technology. In 1969, Schmidt became enamored with what he believed to be a great opportunity for Tymshare: building time-sharing computers (and thus competing with SDS, GE, and DEC). In 1969, unable to get the board’s support for adding that business, he left Tymshare to launch Multi-Access Systems Corporation (Mascor) to fulfill his vision. Schmidt’s timing could not have been worse. The 1970 recession hit IT particularly hard. The $20 million Schmidt was promised was pulled away by financiers, and his vision was not realized, though he later had a successful career with several IT businesses.⁴⁴

Tymshare’s fortunes played out far differently during the recession of 1970. Tymshare reached the $10 million plateau, having grown by 60 percent since the previous year and achieving its first of many continuous years of positive net earnings.⁴⁵ With strong results in the first half of 1970 the company had a track record to complete a highly successful initial public offering of stock in September of that year that quickly sold out then rose to a premium above the $6/share issue price. Tymshare IPO raised $3.6 million, which it used to acquire additional equipment and facilities, as well as to retire debt.⁴⁶

Tymshare’s Tymnet, Acquisitions, and Applications Software

In Tymshare’s first several years, expanding to target new areas—Los Angeles, Englewood Cliffs, Dallas, or Washington (the latter two added in 1968)—involved securing offices and computing equipment in a new city. One way to speed this process was to acquire existing time-sharing companies—of which Tymshare initially looked to other enterprises that were based on the SDS 940. In 1968, Tymshare acquired Dial-Data—and, with it, offices in Newton, Massachusetts and five SDS 940 computers. This also brought new managerial talent in Dial-Data president Lou Clapp, who became a Tymshare executive vice president with Schmidt’s departure.⁴⁷

That same year, Tymshare hired LaRoy Tymes from Lawrence Radiation Laboratory. He was recruited by his friends Ann Hardy and Norman Hardy.⁴⁸ Tymes possessed not only a fitting last name for working at the company, but also considerable programming and network design acumen. The pioneering work of Tymes and Norman Hardy in developing a network, TYMNET, freed Tymshare of the yoke of largely limiting geographical markets to metropolitan areas in which Tymshare had offices and mainframes.

Norman Hardy, a UC-Berkeley-trained mathematician, joined Lawrence Radiation Laboratory in the mid 1950s. There he found a “beautiful combination” of mathematics, physics, and computing. He joined IBM for a time to work on the development of STRETCH, but rejoined Lawrence Radiation Laboratory once STRETCH was delivered. In the mid 1960s, he had also had an opportunity to visit MIT and interact with the Project MAC Multics team, after which he convinced Sidney Fernback, the director of Lawrence Radiation Laboratory’s Computation Department, to fund a project to design a time-sharing system for the newly acquired CDC 6600—a project Hardy then worked on with Tymes. After a brief return to IBM to work on the Advanced Computer System, Hardy joined his wife at Tymshare in 1967. What drew Norman Hardy to Tymshare was the exciting new technology of time sharing. He saw O’Rourke as a “businessman,” and had “never … had a lot of respect for businessmen,” but he believed Schmidt to be “computer savvy … with a fundamental talent for knowing what was [technologically] feasible.”⁴⁹

Hardy and Tymes conceptualized and created TYMNET. One key to making it work, as Norman Hardy reflected, was the idea of statistical multiplexing. The first time-division multiplexing was not optimized for efficiency; it did not take advantage of the fact that “most teletypes were idle most of the time.” Statistical multiplexing allowed there to be variable amounts of buffering “in the machines, and in the nodes.”⁵⁰

In 1968 Hardy and Tymes, through many conversations, convinced O’Rourke to back the networking project that became TYMNET. Tymes, who had both a bachelor’s degree and a master’s degree in mathematics from California State College at Hayward, firmly believed that the existing business model of opening new offices to target metropolitan areas was not economically viable—it was inefficient and limited the potential customer base. Ultimately this insight proved true for Tymshare, GE, and other companies that built networks, as well as for faltering competitors that didn’t build networks.⁵¹

Although Tymes had a vision for the full network, he sold the idea to O’Rourke in pieces, solving intermediate problems. He began with the terminals; the existing SDS 940 computers had relatively expensive Customer Terminal Equipment (CTE). Tymes had the idea of a minicomputer at a fraction of the cost of CTE, using one to be programmable to accommodate different terminal types and baud rates (information transfer in communication channel). Thus the hardware could be far cheaper and utilization optimized. They initially used inexpensive General Automation, Inc.’s SPC-12 minicomputers—a “12-bit, very anemic, special purpose processor,” as Tymes recalled. After quickly demonstrating the concept, they graduated to more powerful 16-bit Varian 620i minicomputers serving as both the base (to interface the network to the host, or time-shared computer) and remote (to drive the terminal). The nodes of the TYMNET, as it became fully operational in 1970, were interconnected by full duplex synchronous leased private lines at 2,400 and 4,800 bits per second.⁵²

Figure 7.2 The Cupertino time-sharing facility of Tymshare Inc., with Varian 620/i minicomputers used as “remotes” to enable TYMNET system (circa early 1970s). Courtesy of Charles Babbage Institute, University of Minnesota.

The most critical element of TYMNET was the “supervisor,” which built circuits, performed diagnostics, kept statistics, as well as other functions. Multiple supervisors were run for redundancy with one in active mode. Tymes programmed the supervisor in late 1969. Every node on TYMNET contained a “leprechaun,” which issued high-priority diagnostics and passed log-in information to the supervisor in active mode. A master user directory (MUD) was programmed by Van Vlear. This contained all passwords and access control information. It was one-way encrypted, offering significant security.⁵³

TYMNET facilitated the delivery of computer time sharing far more broadly and economically. By 1972 it connected more than 40 cities in the United States, a number that continued to grow rapidly domestically and internationally. TYMNET was developed contemporaneously to the ARPANET, but as Tymes later reflected, different in important respects. First, TYMNET was privately funded and had to be “profitable right from the very beginning.” Second, the ARPANET “originally did not have terminal interfaces, it was entirely computer to computer.” Third, ARPANET could not have a centrally controlled point like TYMNET, it was distributed. It needed to be distributed because, as Tymes articulated, a central control point for ARPANET was “politically not viable.” Fourth, Tymes emphasized, the ARPANET, like today’s Internet, is vulnerable to “logjams, thrashings.” “Nobody,” he claimed, “ever successfully hacked into TYMNET.” Finally, Tymes added, TYMNET had no use for TCP, it had “flow control, node to node, on every section of every circuit, so it was not possible to flood a particular node with excess data. … There is no such mechanism in the internet.” While TYMNET never getting hacked is probably an exaggeration, it may have been true through Tymes’ involvement with it from the late 1960s to 1981. Regardless, TYMNET’s creators, Tymes, Norman Hardy, and others, considered security concerns far more seriously than protocol groups for the more interoperability-focused Internet.⁵⁴

TYMNET opened up entirely new possibilities to serve customers effectively in new geographies as well as to build new applications and provide networking infrastructure for companies. In the early 1970s the National Institutes of Health approached Tymshare about providing network access to a large database they had created on poison antidotes. This turned into a major application, operable in 1972.⁵⁵ Later in the decade, they sold underlying software and services to assist TRW with establishing a large internal computer network—TRWNET. In 1976, joint-ownership with Taylorix-Tymshare in Frankfurt and Marubeni in Tokyo, furthered TYMNET’s geographic reach to include 90 percent of Western Europe and much of Japan.⁵⁶

As TYMNET grew rapidly in the first half of the 1970s, the Federal Communications Commission suspected that Tymshare was evolving to become a “communications company.”⁵⁷ This led Tymshare to proactively spin off TYMNET as Tymnet, Inc. a wholly owned subsidiary in 1976. That year Tymnet, Inc. applied to become a “common carrier,” which the FCC granted. With this status, Tymshare’s TYMNET had no restriction on being a “message communication” service, which was previously limited under its “shared use” designation.⁵⁸ This was the overwhelming upside, the limitation (which proved not to be onerous) was, as a common carrier, Tymnet, Inc. was subject to regulation on rates.⁵⁹ Throughout the 1970s, TYMNET, along with software applications and various acquisitions, propelled the firm’s rapid growth.

With the development of TYMNET under way in the late 1960s, Tymshare secured a location on Budd Road in Cupertino, California that soon became the computer center to house the SDS 940 computers. The company rapidly expanded at this location and the emergent Cupertino campus became Tymshare’s corporate headquarters at the start of the 1970s.

Table 7.2 Tymshare’s revenue and net income (in thousands).

	Revenue	Net income
1972	$27,175	$1,617
1973	$39,513	$2,973
1974	$52,682	$3,638
1975	$64,412	$5,064
1976	$81,837	$6,713
1977	$101,174	$8,008
1978	$149,559	$10,594
1979	$193,042	$14,644
1980	$235,854	$18,743
1981	$289,687	$15,670
1982	$297,025	$8,809

Source: Tymshare, Inc. annual reports, 1976–1982 (University Libraries, University of Minnesota).

In its first years, Tymshare was engaged in software development applications internally, which it complemented with software packages and programming capabilities acquired through acquisitions. Initially (and internally) it developed engineering application packages. By the end of 1968 it had three time-shared enabled applications packages completed: COGO, CAPT, and Conversational ECAP. COGO was a coordinate geometry system to solve civil engineering problems such as highway design, right-of-way survey, and bridge geometry. CAPT was a conversational language that enabled programming and operating numerical control in manufacturing. Conversational ECAP was for electronic circuit analysis. Like business data processing applications later, these early systems helped differentiate Tymshare’s time-sharing services to customers.⁶⁰

Tymshare used acquisitions to expand rapidly during the 1970s and the early 1980s. By 1968, more than forty companies had entered the time-sharing industry.⁶¹ A few quickly faltered and disbanded, but most new entrants survived until the nearly year-long recession of 1970, which led to an initial shakeout in the industry. Much of 1969 saw unparalleled exuberance in software, computing, and time-sharing companies and their stocks. There were more new time-sharing companies formed that year (26) than any other in the industry’s history. The number of new firms dropped to 16 in 1970, 5 in 1973, and only several others were launched between 1974 and 1978.⁶² Tymshare was well established by 1970 and faired far better than most time-sharing companies. In fact, it expanded through joint-ownership agreements into both Canada and France that year—taking minority ownership in Tymshare, Canada and CEGOS-Informatique and Credit Lyonnaise.⁶³

Tymshare’s acquisitions accelerated rapidly throughout the remainder of the decade. Some acquisitions were to reach new geographies (even with TYMNET, geographical coverage was important to both costs and serving customers). Of equal and sometimes greater importance, acquisitions allowed the firm to gain software packages targeted at specific industries, as well as to secure talented applications programmers. Some acquisitions achieved both geographic expansion and software/industry capability goals—especially the 1974 acquisition of United Data Centers. It brought fifteen new offices, primarily in the eastern United States and one in Montreal, as well as important software applications.⁶⁴ Like most Tymshare acquisitions, it was an all-stock deal. It represented its largest acquisition up to that point, with a value of approximately $6.1 million.⁶⁵

United Data Centers’ founder and leader, Bernard Goldstein, who had previously worked at Control Data, launched UDC to become a chain of batch-focused data centers. He had grown it through acquisitions with an eye toward business applications—including UDC’s acquisition of Dynafacts (in Wichita, Kansas), which had an automated tax return software package. Goldstein later recalled that Tymshare wanted UDC for its business software applications, its capabilities to offer “solutions for the marketplaces.”⁶⁶ After the acquisition of UDC, Goldstein became Tymshare’s mergers and acquisitions specialist for several years.

In 1976, Tymshare reorganized into three marketing subdivisions: Information Services, Industry Services, and Marketing Services. This formalized strategic planning and organization in targeting particular industries for customers. O’Rourke emphasized in the 1976 annual report that this would be achieved both through internal resources and through acquisitions to gain certain industry specific capabilities.⁶⁷ Three targeted industries for Tymshare acquisitions (by O’Rourke and Goldstein) in the mid to late 1970s were tax preparation, travel, and medical data processing. This strategy was implemented through acquisitions of Unitax (1976), Autotax (1978), Western Twenty-Nine (1976), Medical Data Systems (1976), and Medical Information, Inc. (1978).⁶⁸

In the early 1980s, bank card processing was targeted by Tymshare through acquisitions of Bancard of Rhode Island and TeleCheck (both in 1980). Unlike Dial-Data, new acquisitions also furthered the expansion of Tymshare to a wider set of hardware systems including models from minicomputer specialist Digital Equipment Corporation and IBM. Tymshare’s 1975 acquisition of Alan-Babcock Computing provided IBM mainframes. Tymshare’s organizational customers also developed applications software that ran on the network. Some businesses made deals with Tymshare to serve as a selling agent for their applications. Norman Hardy emphasized the flexibility of Tymshare in its early years and O’Rourke’s willingness to consider all types of arrangements that had revenue and profit potential.⁶⁹

Tymshare’s varied acquisitions brought new employees from different corporate cultures, which sometimes proved challenging. This also was true with its international joint ownership arrangements. With the latter, in the mid 1970s, Tymes and Norman Hardy were meeting with CEGOS, Tymshare’s jointly owned company in France. According to Tymes’ recollection of the corporate planning meeting, “The French people were aghast at our proposal that the names of all the French customers had to be in the MUD that was to be stored in the United States … . And Norman Hardy said, ‘why not?’ A supervisor is a supervisor; its geographic location really isn’t material. If it’s controlling the net it does have to have all the information of all the users to function.” Reflecting on the situation, Tymes said “[T]o us it was a pure technology problem, but to the French it was a matter of pride.”⁷⁰

Acquisitions and subsequent integration could be even more difficult. Many of Tymshare’s 1,500 employees by the end of 1976 had joined the company after their firm or organization was acquired by Tymshare. Lynn Sanden, who joined as a secretary in 1969, worked her way up to the position of human resources manager. She recalled the difficulties after Tymshare, in 1977, acquired a center of about 20 research scientists at SRI, Douglas Engelbart’s famed oN-Line System Center (NLS), which he founded (as the Augmentation Research Center) in the mid 1960s. For Tymshare executives, the SRI-NLS acquisition was completed to advance efforts in office automation and to develop software for the office of the future.⁷¹ Though Engelbart adjusted well to the new environment and became Tymshare’s chief scientist, many of the former SRI-NLS computer scientists did not want to be acquired, and that led to “hostile” attitudes.⁷² A number of the former SRI-NLS computer scientists resigned and joined Xerox PARC or other businesses or organizations.

In other cases disagreements materialized with regard to responsibilities, job titles, compensation, or deployment of technology. In acquiring Western Twenty-Nine, the agreement was for the newly organized unit, and its head Pat Brent (former leader of Western Twenty-Nine), to receive a portion of compensation in profit sharing. However, Brent’s unit was administratively set up in a way that it was exceedingly difficult to show a profit (internal or transfer pricing on what the unit was charged for TYMNET was high).⁷³ And with the 1978 acquisition of Medical Information, Inc., to bring time sharing and applications software to hospitals and nursing stations, what Tymshare could actually offer was never carefully considered by O’Rourke. Ann Hardy had written a memo opposing the acquisition on this basis, but Tymshare nevertheless went through with the acquisition.⁷⁴ It was a bust, and Tymshare completely divested of medical information processing several years later in 1982.

Table 7.3 Tymshare’s acquisitions and mergers, 1968–1982.

	Acquired or merged division or company
1968	Dial-Data
1970	CEGOS-Informatique and Credit Lyonnaise, France (partial ownership)
	Tymshare, Canada (partial ownership)
1971	Computer Systems Division, Graphic Control Corporation
1972	Computer Complex TS Ops
1973	Valley Computer
1974	United Data Centers
	Tronics, Inc. (merger)
1975	Alan Babcock, Inc.
	Quelex Data Systems
	Leasco Response, Inc.
1976	Unitax, Inc.
	Simplified Data Processing (merger)
	Taylorix-Tymshare, Germany (partial ownership)
	Marubeni Corporation, Japan (partial ownership)
1977	Western Twenty Nine, Inc.
1978	Auto Tax
	Medical Information, Inc.
1979	Valdata Division, TRW
1980	Medistat
	Bancard of Rhode Island
	TeleCheck
1981	Electronic Travel Services, Inc.
	Payment and Transfer Services, Inc.
	Microband Corporation of America
1982	FTC Communications

Sources: Corporate History Collection, Computer History Museum; http://corphist.computerhistory.org/corphist/view.php?s=themes&id=5 (accessed July 24, 2015).

The advent of personal computers in the mid 1970s and the acceleration of the personal computer industry in the second half of the 1970s did not go unnoticed by Tymshare’s leaders. Tymshare’s subsidiary Tymnet was not only providing the network infrastructure for Tymshare, but selling networking service to customers that were not purchasing Tymshare’s time sharing. In the early 1980s, O’Rourke estimated, half the business was coming from the TYMNET communication services-only side, and of the two that was the growth business. O’Rourke had pivoted strategically to move more of the business “to communications,” to the Tymnet subsidiary. O’Rourke later commented that outsiders missed what Tymshare was doing in the early 1980s: “They felt we were riding a dead horse with time sharing, we were frantically trying to change horses but not doing it abruptly because these guys [the time-sharing business] were cash cows … .”⁷⁵

Figure 7.3 An operator at a Tymshare 430 monitor station, circa 1981. Courtesy of Charles Babbage Institute, University of Minnesota.

Tymshare had created a valuable network in TYMNET, but Tymshare had declining net income in both 1981 and 1982. In 1983, when TYMNET linked 400 cities in 42 countries, Tymshare became an acquisition target of McDonnell Douglas’ Automation (i.e., information technology) division (commonly known as “McAuto”). McDonnell Douglas’ initial offer in 1983— $378 million—was withdrawn late in the year amid third-quarter reports showing Tymshare had lost $139,000 in the first nine months of the year. Early in 1984, McDonnell Douglas came in with an offer roughly 20 percent lower—$307.5 million—and it was accepted by Tymshare’s board.⁷⁶ McDonnell Douglas broke the non-network portion of Tymshare into many divisions, some of which were sold; it kept TYMNET for five years before selling it to British Telecom in 1989.⁷⁷

The Evolution and Fate of the Broader Time-Sharing Industry

GE Information Services (GEIS) took the early lead in the industry and never looked back. Like Tymshare, it developed a major computer network in the late 1960s.⁷⁸ GEIS’s Network I had remote hubs in Los Angeles, Kansas City, Atlanta, Teaneck, Schenectady, and Washington that linked its suburban Cleveland (Brook Park) central computer complex to buffer units (multiplexers) in more than forty major metropolitan areas of the US by late 1969. Similar to Tymshare, GEIS developed specialized software applications targeting particular industries—in part through acquisitions in the banking and energy data processing fields.⁷⁹ In the mid 1980s it became a major supplier of Electronic Data Interchange (EDI), offering business-to-business networked communication systems that linked inter-firm supply chains IT systems in the motor vehicle, aerospace, and other industries. By 2000, the former GEIS, having been renamed Global Exchange Services (GXS) a few years earlier, had revenue of about $640 million and was focused on supply management EDI operations for about 100,000 business units of roughly 30,000 multinational firms. In 2002, GXS sold a 90 percent stake to the venture firm Francisco Partners for $800 million.⁸⁰

IBM had officially entered time sharing in mid 1960s with QUIKTRAN and DATATEXT, but those were limited specialized offerings for “interpretive program execution” and “text editing” (respectively) for certain pre-IBM System/360 machines, including the IBM 1460.⁸¹ IBM participated in the field in a substantial way in 1968 with “CALL 360” services that in a matter of months served 34 US cities to sell time on IBM System/360 Model 40 and Model 50 mainframes located in Los Angeles, San Francisco, Chicago, Cleveland, Philadelphia, and New York. Virtually from the start of CALL 360, this effort caught the attention of the US Department of Justice, which, under the 1956 consent decree, had prohibited IBM from service bureau operations in forcing it to spin off its Service Bureau Division as Service Bureau Corporation (SBC), a wholly owned subsidiary that had a separate workforce and got no breaks on purchases of IBM computer systems.⁸² Early speculation in the computer trade press in 1966 about whether time sharing was another form of leasing (allowed with restrictions under the 1956 consent decree) or a new form of service bureau (not allowed) appeared to tilt toward the latter (at least in the US Department of Justice’s interpretation) by late 1968, since many service bureaus were moving to add time sharing.⁸³

While many historians have emphasized software products with IBM’s unbundling, Department of Justice officials were also heavily focused on IBM’s impact on the fast-growing computer services industry and its relationship to the hardware business, including time sharing (potential anti-competitive advantages either in hardware or services by the computer leader). In early December 1968 an IBM spokesperson stated, that for several months the corporation had been “reexamining its methods of doing business in the US to determine what support services should be separately offered and priced.”⁸⁴

Figure 7.4 Lead operator Bob Warno (left) and systems programmer Carlos Navarro (right) in the Chicago office of Comshare, Inc., 1979. Courtesy of Charles Babbage Institute, University of Minnesota.

More than a month earlier, on October 23, 1968, IBM’s leaders, in fact, had already made a major decision on services by shifting “commercial time-sharing” services away from IBM’s Data Processing Division to its subsidiary SBC; by then SBC was a slower-growing, low-margin operation.⁸⁵ In 1972, SBC had profits of only $1.5 million. (SBC’s margins in those years tended be about 2 percent.) In 1973, at the time of the transfer of SBC to Control Data Corporation as part of the settlement of its lawsuit against IBM, two thirds of SBC’s business was still from batch processing.⁸⁶ Thus, in 1973, SBC probably had less revenue and smaller profits from time sharing than Tymshare, and far less than GEIS. Adding SBC to CDC’s time-sharing business probably propelled it beyond Tymshare in revenue to second place in the industry, far distant from GEIS (though Tymshare probably regained the number-two position in the industry later in the decade as CDC’s time-sharing services business grew more slowly).

7 Sharing Time: Tymshare, 1965–1984