Grace Hopper and the Invention of the Information Age
Page 23
The surging demand for UNIVACs in 1953 and 1954 added to Remington Rand’s growing customer support problems. Many talented members of Grace Hopper’s staff were assigned to clients or were hired away by clients. Hopper asked John Mauchly to inform her superiors at Remington Rand of her growing attrition problem. Mauchly put it this way:
Some of the members of Dr. Grace Hopper’s staff have already left for positions with users of IBM equipment, and those of her staff who still remain are now expecting attractive offers from outside sources. . . . The Eckert-Mauchly Division has not, however, been able to make offers sufficiently soon enough, or good enough, to prevent the depletion of her staff , because there is no budget allowance in the Eckert-Mauchly Division for such personnel.11
Pressure from Mauchly, Hopper, and others prompted Remington Rand to organize more structured UNIVAC customer support. During the summer of 1953, a UNIVAC National Sales Headquarters was established in New York, with a training office, a customer service bureau, and a maintenance office. Paul Chinitz left Hopper’s programming group to head up the initiative. The training office offered an intensive two-week course on computers that included an overview of UNIVAC I’s capabilities and seminars on programming and systems analysis. According to Chinitz, the fee-based course attracted mostly middle and upper-middle managers who were familiar with the data-processing requirements of their companies.12
The customer service bureau, under the direction of Arthur Katz, helped potential clients determine their computing needs. The service bureau conducted surveys, prepared proposals, and organized customer-specific demonstrations. A demonstration usually involved a critical aspect of the client’s proposed need. Once defined, the problem was sent to Hopper’s team in Philadelphia, and two or three programmers worked on it for a month or two. The finished program was then run for the potential customer, either at the New York service bureau, at the Philadelphia factory and laboratory, or at the Census Bureau in Washington. If the demonstration was a success, the service bureau followed up to sign a contract.13
Eventually the maintenance department came under the direction of engineers from the newly acquired Engineering Research Associates (ERA). Remington Rand purchased the Minnesota-based company in early 1953 for a variety of reasons. First, ERA had 400 experienced technicians and engineers who could be used to augment Remington Rand’s customer service staff. Second, ERA’s president, John Parker, would be appointed head of the National Sales Office, so Presper Eckert could remain in the laboratory. Third, ERA’s magnetic drum memory technology was the most advanced in the industry. Finally, ERA’s computers were scientific in their orientation, thus augmenting Rand’s product offering and complementing UNIVAC I’s more commercial slant.14
Though Remington Rand’s efforts to reorient itself were commendable, the 1950 decision to cut production capability came back to haunt the firm. The UNIVAC engineer Lou Wilson later acknowledged the almost impossible situation in which he and his fellow engineers had found themselves during 1953 and 1954: “In spite of having the production capability of only six machines a year, within 2 years they sold 42 machines. The result was they couldn’t deliver the damn things.”15 And if Remington Rand did deliver on time, what showed up was not always a fully functioning machine. Gene Delves recalled, for example, that a UNIVAC was to be delivered to a General Electric facility in Louisville in December 1953. With the delivery date quickly approaching, the machine was still in production. “So December 15 they just crated it all up and brought it to Louisville and delivered it along with all the engineers who were still building it.”16
Despite manufacturing challenges, in the winter of 1953–54 Grace Hopper believed that the dearth of trained programmers was the chief limit on the growth of the computer market. As has happened often in the history of technology, labor scarcity became a central external driving force behind technical innovation, prompting Hopper to ask Remington Rand management to allow herself and her team to dedicate themselves full-time to finding a way to automate the programming process.17 In light of Remington Rand management’s unwillingness to make the necessary budget allowances to maintain an adequate programming workforce, Hopper’s 31 December 1953 proposal was approved and the Computation Analysis Laboratory was renamed the Automatic Programming Department. The new department shed its more mundane task of UNIVAC customer support and focused entirely on automatic programming research and development related to automatic programming. Hopper was named the department’s director.18
IBM BENEFITS FROM THE COLD WAR
The historical evidence shows that during the first half of the 1950s Remington Rand competed successfully with IBM. Not only did UNIVAC lead in sales; even such traditional IBM customers as General Electric and Pacific Mutual picked the UNIVAC over the IBM 701/702 on technical grounds. Furthermore, all competitors in the computer market were faced with the same shortage of talented programmers. Unwillingness to pay a premium to keep programming talent may have cost Remington Rand a chance to build a commanding lead within the industry, but the firm’s efforts to build a customer training and support system were commendable, and senior management’s willingness to support Hopper’s automatic programming research was farsighted.
If Remington Rand held its own during the first half of the 1950s in an area of relatively open competition in the nascent industry, what explains IBM’s sudden ascendancy? Paradoxically, government funding of private research and development became a pivotal factor in IBM’s rise to dominance, for it subsidized IBM’s computer investment and transferred protected knowledge from the Massachusetts Institute of Technology to IBM’s Poughkeepsie Laboratory. That funding came courtesy of the SAGE (Semi-Automatic Ground Environment) early warning defense system.19
SAGE was a consequence of the growing tensions between the Soviet Union and the United States. In 1949 the Soviet Union exploded its first atomic bomb. The communist nation’s newfound nuclear capability, combined with its long-range bomber technology, threatened the North American continent, and the Truman administration called upon America’s scientists and engineers to devise a counter to the perceived threat. The associate director of MIT’s Servomechanisms Laboratory, Jay Forrester, proposed a network of radar installations across Northern Canada and Alaska to track enemy bombers and provide real-time intercept information to friendly fighter aircraft. The Forrester proposal mapped out a 15-year research and development effort with a price tag exceeding that of the Manhattan Project.20
The only way to deal with a mass attack of bombers was to fully automate the tracking and intercept system—something that could be accomplished, Forrester believed, with high-speed digital computers. During World War II, Forrester had worked on a flight simulator for the Navy. After the war, that project retained funding under a new name: Project Whirlwind. With the continued backing of Mina Rees at the Office of Naval Research, Whirlwind became the most elaborate, expensive general-purpose computer project of its day. Forrester now saw a purpose for the theoretical machine. The Whirlwind computer—a real-time information-processing system capable of automating command and control—would serve as the brain of the SAGE radar network.21
Late in 1952, Forrester received authorization to pick a subcontractor for an unprecedented 50 Whirlwind computers. Forrester and his team spent the early part of 1953 interviewing engineers, programmers, and management from three prospective subcontractors: Remington Rand, the Raytheon Manufacturing Company, and the International Business Machines Corporation. Thomas Watson Jr. believed that the SAGE contract was the answer to the UNIVAC challenge. “I thought it was absolutely essential to IBM’s future that we win it,” Watson recalled. “The company that built those computers was going to be way ahead of the game, because it would learn the secrets of mass production.”22
Unknown to Watson Jr. at the time, efficient computer mass production was not the only secret to which the contract winner would be privy. Between 1948 and 1951, at the cost of more than $1 million, Forrester and his t
eam had invented high capacity, real-time, random-access memory. Not satisfied with mercury delay lines or electrostatic storage, Forrester’s team experimented with magnetic memory in 1949. Starting from the knowledge that certain materials could be magnetized and demagnetized by pulses of electricity, the MIT team constructed a matrix of small doughnut-shaped magnetic ferrite cores that performed much like binary switches.23 Not only could these switches hold more information than other memory technologies; the information could be accessed far more rapidly in a non-serial fashion.24
IBM executives were well aware that Remington Rand was the top candidate for the SAGE contract. Not only had Remington Rand’s machine proven itself in the marketplace; Leslie Groves, the much-respected former director of the Manhattan Project, was the head of advanced research at Remington Rand. “I tried not to worry about Groves or the other competitors,” Thomas Watson Jr. remembered. “I took Forrester to see our plants and introduce him to our most gifted people. He was under extreme pressure to get the system into production as soon as possible, and I think what impressed him was the fact that we were already building computers in a factory.”25
Though IBM’s internal computer technology left something to be desired, Forrester was impressed by the Poughkeepsie lab’s integration of research and production. Forrester also took note of IBM’s “degree of purposefulness, integration, and esprit de corps.”26 Remington Rand’s superior technology and talented engineering and programming staff were overshadowed by concerns about the company’s production capabilities. Forrester concluded that IBM was better equipped to manufacture, in a limited period of time, 50 of the most complex machines ever produced. In April 1953 he offered the contract to Thomas Watson Sr.27
The contract to build 50 Whirlwind computers was a technological watershed for IBM, and Thomas Watson Jr. was the first to admit that “SAGE saved the 702 design team.”28 IBM immediately assigned more than 300 engineers and scientists to Poughkeepsie in order to design and manufacture what would be called the AN/FSQ-7. During the first year alone, MIT transferred more than 1,000 confidential technical documents to Poughkeepsie, while IBM personnel spent upwards of 950 person-days at MIT’s Lincoln Laboratory learning about Whirlwind.29
Not only did IBM engineers take away knowledge about random-access magnetic core memory; they also learned how the Whirlwind team had pushed the technological envelope in a number of other areas. Forrester’s staff had figured out a variety of ways to lower the frequency of vacuum-tube failure, thus increasing system reliability. Cathode-ray-tube displays were ingeniously employed to display processed information, index registers made programming easier, and real-time information from radar sensors could be processed without the need for a slow input medium such as punch cards.
Not surprisingly, Whirlwind technology quickly found its way into the next generation of IBM’s commercial computers.30 The IBM 704 was announced on 7 May 1954. Its redesigned circuitry was twice as fast as its 701 predecessor and far more reliable. Although it was shipped with two electrostatic and two magnetic drums for storage, the 704 was designed so that core memory could be installed at a later date. Then, on 1 October 1954, the IBM 705 was announced. The 705 was manufactured with fully integrated random-access magnetic core storage and a buffered magnetic tape input/output (I/O) system that could read and write simultaneously.31
To appease disgruntled customers, IBM filled standing orders for 701s and 702s with the superior 704s and 705s. The market quickly embraced the new IBM machines, and there were far more new orders than had been anticipated. “In a little over a year we started delivering those redesigned computers,” wrote Thomas Watson Jr. “They made the UNIVAC obsolete and we soon left Remington Rand in the dust.”32 By the start of 1957, IBM had 87 computers in operation and 190 on order. Remington Rand, in contrast, supported 41 operational UNIVACs and had back orders for 40 more.
In only 3 years IBM had met the Remington Rand challenge, mainly because of the benefits accrued from winning the SAGE contract. Not only did SAGE inject Whirlwind technology deep into IBM’s DNA; the magnitude of the contract permitted IBM to hire more than 8,000 more engineers, programmers, and administrators. An astonishing 80 percent of the company’s revenues from stored-program computers were generated by the SAGE contract, which amounted to more than $500 million by the completion of the project.33 In view of the Justice Department’s antitrust case against IBM during the 1970s, it is paradoxical that a large government contract appears to have been the basis for IBM’s ability to dominate the computer market by 1960.
Besides providing IBM with a market advantage, the SAGE contract transmuted computer programming from an esoteric craft to a well-funded profession. The massive programming task required hundreds of thousands of lines of code, written by more than 800 programmers. Since resources were not an issue, IBM subcontracted much of the programming work to the RAND Corporation, a non-profit, military-sponsored research center. RAND hired and trained so many programmers that the SAGE development division soon outnumbered the rest of the RAND Corporation, prompting RAND to spin off the division as the System Development Corporation.34
The 28 March 1955 cover of Time.
The sharp increase in the number of programmers sparked by the SAGE contract did not directly affect Grace Hopper and her work, but it did indirectly make it more difficult to sell the idea of automatic programming. Some of the most outspoken opponents of Hopper’s automatic programming vision through the 1950s were programmers. By automating the programming process, Hopper was threatening to make programmers redundant at a time when their numbers and importance within the computing community were growing exponentially.
10 THE DEVELOPMENT OF PROBLEM-ORIENTED LANGUAGES
On 13 May 1954, a little-known moment in computer history significantly affected programming development at both Remington Rand and IBM. Charlie Adams, a long-time member of MIT’s Project Whirlwind, presented a paper at an Office of Naval Research (ONR) symposium on automatic programming. Grace Hopper had organized the meeting in conjunction with the Navy in order to assess the state of the programming field. Adams described computer developments at MIT, including the work of J. Halcombe Laning Jr. and Neal Zierler. Laning and Zierler had applied emerging compiler techniques and had created what they called an algebraic compiler. The compiler accepted standard mathematical symbols and converted them into machine language. Thus, with a limited amount of training, a first-time user could simply write an equation and its parameters, and the computer would generate the answer.1
For the most part, Laning and Zierler’s efforts were ignored. To many the algebraic compiler sounded nothing short of chimerical. IBM programmer John Backus captured this sentiment in a paper he presented at the same conference, “IBM 701 Speedcoding and Other Automatic Programming Systems.” Backus stated that most programmers had at one time or another thought about how nice it would be to write X + Y instead of a more complicated code, but “no doubt if he were too insistent next week about this sort of thing he would be subject to psychiatric observation.”2 As of 1954, programmers in general were skeptical of most forms of automatic programming, let alone ones that claimed to be capable of the direct translation of mathematical equations. “To them,” Backus later recalled, “it was obviously a foolish and arrogant dream to imagine that any mechanical process could possibly perform the mysterious feats of invention required to write an efficient code.”3 But in fact there were dreamers in the audience that day, including the newly appointed director of automatic programming at Remington Rand, Dr. Grace Hopper. Her own A-2 compiler, though a milestone in the field, was difficult for even skilled programmers to use.
Laning’s and Zierler’s work demonstrated the possibility that computers could be operated by laymen. Two years later, while serving as the keynote speaker for the second ONR-sponsored Symposium on Advanced Programming Methods, Hopper stated that the MIT compiler represented the most comprehensive language ever developed, since it not only assisted the origin
al programming of the problem but even partially automated the debugging process.4 MIT’s achievements in the field of automatic programming effectively changed the trajectory of Hopper’s work in automatic programming between 1954 and 1956, leading to her most significant contributions to date.
THE AUTOMATIC PROGRAMMING DEPARTMENT AND MATH-MATIC
By 1954, “computer programmer” was a well-defined job description. Most programmers were mathematically trained and knowledgeable about computer hardware. The programming craft involved coaxing accurate output out of a temperamental machine handicapped by input/output and memory limitations. The complexities surrounding the “black art” of programming translated into escalating computer costs that both users and manufacturers of computer equipment hoped to lower with the implementation of automatic programming. This mounting economic quandary eventually moved Remington Rand management to back Hopper’s research efforts by forming the Automatic Programming Department. The new department afforded Hopper the chance to implement her own unique approach to management. Combining a sense of purpose and direction from the Harvard Computation Laboratory with the flexibility and freedom of the Eckert-Mauchly Computer Corporation, Hopper instituted a decentralized management style that unleashed a certain level of self-direction, encouraged playfulness in the work environment, and provided her subordinates with an overarching vision of the computing future. Time and again people who worked with Grace Hopper during this period spoke about how they loved what they were doing. “It was,” Mildred Koss recalled, “really a place where you felt a lot of loyalty to the people who were there, wanted to make it work, wanted to be a part of that exciting environment.”5 Grace Hopper had somehow captured the unity and sense of urgency that existed during the war and transferred it to Philadelphia in the 1950s.