Two Unknowns and Their Ambitious New Computing MachineDecember 7, 1999
ENIAC: The Triumphs and Tragedies of the World's First Computer Company. By Scott McCartney, Walker and Company, New York, 1999, viii + 262 pages, notes, index, $23.
Who invented the computer? It seems such a straightforward question, yet the answer is complicated---with a number of people vying for the honor. Charles Babbage conceived of an Analytical Engine in the mid-19th century that embodied most of the functional concepts of the modern computer. John Vincent Atanasoff completed in 1942 a special-purpose electronic calculator for solving systems of linear equations at Iowa State College. Harvard professor Howard Aiken conceived, and IBM built for him in 1943, a general-purpose, electromechanical, automatic calculator. J. Presper Eckert and John Mauchly built the electronic calculator ENIAC and began work on a stored-program computer, EDVAC, at the University of Pennsylvania during the later years of the Second World War. John von Neumann wrote the Draft Report on EDVAC, which set out the first design for a stored-program computer, and subsequently built a computer at the Institute for Advanced Study in Princeton. Maurice Wilkes, at Cambridge University, was the first---in 1948---to place in operation a fully functioning, full-scale, electronic, digital, stored-program computer. Other claims have been made for Konrad Zuse in Berlin, M.H.A. Newman at Manchester University, George Stibitz at Bell Telephone Laboratories, and others.
High-speed calculating machines are of course complex systems, and many different models were designed and constructed before the electronic/digital/stored-program architecture design triad became standard. The historian is not so interested as the practitioner in determining "firsts"---and in fact sees the concept as relative. Which machine obtains the honor as "the first computer" depends on the answers to several questions: Which technical features are considered most important? Was the machine merely conceived, or was it actually built? Was a prototype or a full-scale machine placed in operation? How was the machine used?
No matter which machine is believed to have been the first computer, the construction of the ENIAC is one of the signal events in computing
history. The United States designed a multitude of new guns and ammunition for the Second World War, and gunners needed tables that would tell them how to aim, based on target distance, wind speed, humidity, temperature, and other factors. Teams of human "computers" were falling steadily behind in their race to calculate the entries for these tables. The University of Pennsylvania's Moore School of Electrical Engineering held a contract with the Army's Aberdeen Proving Ground to carry out some of these calculations because it was one of the few places to have a powerful computing device---a differential analyzer copied after one designed at MIT. But even this machine was not enough to keep up with the demand for calculations.
Thus, the Army was willing to take the advice of Lieutenant Herman Goldstine, a young PhD mathematician who was serving as Aberdeen's liaison to the Moore School, and risk funding the construction of an extremely ambitious new computing machine, the ENIAC. The designers of this machine were two unknowns: J. Presper Eckert, a promising electrical engineering graduate student from a wealthy Philadelphia family, and John Mauchly, a mid-career physicist from a small college who had come to Penn first to enroll in and later to teach a wartime course in electronics.
Eckert and Mauchly were not able to complete the construction of the ENIAC until late 1945, after the war had ended. The machine's importance thus lay in its design rather than in its usage. The ENIAC was the first calculating machine to incorporate the high speed of electronics into a design with a scale large enough to solve important real-world problems. Moreover, their experience with the ENIAC led Eckert and Mauchly to work on a second machine, the EDVAC, with the capability to store and modify its own instruction---a machine that could be "programmed" in the modern sense. The EDVAC was not completed until 1952, long after Eckert and Mauchly had left the Moore School to form their own computer company. But the proof of principle of electronic computing shown by the ENIAC and von Neumann's famous technical report on the EDVAC's design inspired people around the world to begin to build computers. Eckert and Mauchly's commercial efforts were also important, spawning the UNIVAC line of computers.
The growing literature on the ENIAC and other aspects of early computer history can be divided into four categories, as shown in the box on this page. The book under review here falls into the popular-advocacy category. McCartney is a staff writer for the Wall Street Journal who has written books on organ transplants and nuclear weapons. It is his intention to show that "the distinction [of inventing the computer] rightly belongs to . . . Presper Eckert and John Mauchly" rather than to John von Neumann.
Von Neumann's name has been associated widely with the invention of the computer---the most common computer design, for example, is called "the von Neumann architecture." During the war, von Neumann, who already had a well-established international reputation as a mathematician, was much in demand. He led an itinerant life, traveling by train between Los Alamos, Aberdeen, and several other locations to consult on mathematical aspects of ballistics, shock waves, and weapon design. Los Alamos needed help with the numerical modeling of a triggering device for the atomic bomb, and von Neumann made several inquiries about the availability of high-speed calculating machinery for this purpose. Nobody thought to mention the ENIAC project to him, probably because Eckert and Mauchly did not have established reputations and the design seemed hopelessly ambitious. However, when Goldstine ran across von Neumann in 1944 on a train platform in Aberdeen, introduced himself, and told von Neumann about the ENIAC project, von Neumann took an immediate interest. Goldstine quickly arranged von Neumann's appointment as a consultant at the Moore School, adding yet one more job to his list of wartime consulting assignments.
By the time von Neumann joined the project, the ENIAC design was set and construction was well under way. Eckert, Mauchly, and others had already been meeting occasionally for more than half a year to discuss the design of the successor machine, the EDVAC. Von Neumann joined in on these discussions when he was available, every month or two. During an extended stay at Los Alamos, he wrote the Draft Report on EDVAC, which Goldstine distributed widely. Much to Eckert and Mauchly's annoyance, von Neumann's name was the only one to appear on the document. The report gave a lucid description of the functional design of the stored-program computer, using the abstract language and concepts of neural nets that had recently been invented by Warren McCulloch and Walter Pitts to emphasize the "logical design," rather than discussing specific engineering implementations. Once the war ended, von Neumann returned to the Institute for Advanced Study, where he ran his own computer project to test the value of the computer in scientific research.
Historical literature is often a product of its times. To succeed, a history must address the concerns and reflect the attitudes of its period. Thus, historical figures, from Isaac Newton to Richard Nixon, are interpreted and reinterpreted by successive generations of scholars and writers. In at least two ways, McCartney's book is a product of its times. The growth of the personal computer and the Internet has brought home to the general public the importance of the computer to modern life. This, in turn, has led to a steady diet of books about the origins of the computer written at a level accessible to the home user of computers.
McCartney's particular slant may also have its origins in current attitudes. In the 1960s and 1970s, the mathematical aspects of computer science were ascendant; for example, there was a strong correlation between the rankings of computer science departments in the U.S. and the quality of their theory faculty. In the 1980s and 1990s, there has been a shift in academia toward the engineering aspects of computer science. Herman Goldstine's The Computer From Pascal to von Neumann---published in 1972, the heart of the mathematical era---argues for the prominence of von Neumann's contribution to the stored-program theory over the engineering contributions of Eckert and Mauchly. These early developments have been reinterpreted over the last fifteen years, giving more emphasis to the machine builders and less to the theorizers. McCartney is not a member of the computer community and does not seem aware of this historiographic shift, but as a journalist he has picked up on the prevailing attitude in the practitioner community.
The story of the invention of the computer is not so much a story about machines as it is a story about humans---about genius, human frailties, ambitions, politics, and institutions. The ability to write about this human dimension is a strength that a good journalist or professional writer can bring to the literature on technological creation. McCartney recognizes the importance of the human dimension, and the best parts of his book are those covering the family backgrounds, ambitions, and work styles of Eckert and Mauchly. The book is also worthy of attention because it is the first to benefit from access to the personal papers of Eckert and Mauchly.
Unfortunately, the book has a number of limitations. There seems to be an overreliance on oral histories, which are uncritically accepted, and a lack of attention to traditional archival sources. There is also a problem with Mc-Cartney's understanding of basic concepts. For example, he confuses the meaning of the terms "real-time computing" and "analog computing." He refers to the work of Herman Hollerith, the inventor of the punched-card tabulating system and the founder of a company that was the major predecessor of IBM, as real-time computing. But Hollerith's punched-card systems are the embodiment of batch processing, which is the antithesis of real-time computing. In a batch-processing system, the work is stored until there is enough to make maximum use of the machine; no effort is made to do the processing when the input is received. McCartney calls Howard Aiken's computers at Harvard "analog" machines---presumably part of his effort to differentiate them from the ENIAC. However, the Harvard Mark I was digital, just like the ENIAC; it differed from the ENIAC in being electromechanical rather than electronic.
In preparation for writing this book, McCartney appears to have given inadequate attention to the growing literature on other early computers. As a result, he has some trouble situating the ENIAC and EDVAC in their proper historical place, in characterizing what is special about the machines built by Eckert and Mauchly. For example, McCartney argues that Penn was working in the digital (read "modern") tradition, whereas the major American academic computing powerhouse, MIT, was working in the analog (read "outmoded") tradition. There is no question that MIT faculty members Vannevar Bush and Samuel Caldwell had been the principal originators of analog computing in the 1920s and 1930s. Yet McCartney's assessment misses the fact that, soon after the end of the Second World War, the MIT administration made the decision to return a Rockefeller Foundation grant that was to have paid for the recommencement of its analog computing program. MIT went on instead to back the digital work then starting up in the Servomechanisms Laboratory and in the Research Laboratory for Electronics.
McCartney voices a commonly heard view: If only Penn had been smarter and kept Eckert and Mauchly at the university, the university would have been the leading computing center in the U.S. But even if Eckert and Mauchly had remained at Penn, it is not clear that Penn would have been the dominant force in computing. One must remember that MIT had an exceptional staff and knowledge of electronics as a result of its wartime Radiation Laboratory; it had unparalleled ties to federal scientific leadership and industry, and it had a long and strong computing tradition that was deeply embedded in the mission of the university.
Disputes over intellectual property rights drive much of McCartney's narrative. The Army had given the University of Pennsylvania the right to file patents arising from the ENIAC project, so long as the government held a nonexclusive, no-cost right to the patents. Eckert and Mauchly arranged with the university in 1943 that the patents could be held by the individual members of the Moore School team who were responsible for the inventions. Eckert and Mauchly filed the only patent application, however---for the overall computing system (and only in 1947).
By 1946, when it came time to negotiate for the rights to the EDVAC patent, the university saw that the world had changed and that it must harden its patent policy. Universities had new opportunities to become strong through research sponsored by industry and the federal government, and Penn believed that it had to take steps to protect these opportunities. McCartney argues that this change in policy was not only unfair to Eckert and Mauchly, but was also exceptional among American universities. However, the new director of research at the Moore School, Irven Travis, had been a contracting officer for the Navy during the war. He modeled the Moore School's new patent policy after the practices he was familiar with from the Applied Physics Laboratory at Johns Hopkins and the Department of Industrial Cooperation at MIT. The new policy established by Travis required that faculty members sign over all patent rights to the university.
Rather than sign over their EDVAC-related patents, Eckert and Mauchly resigned from the university and started up a commercial venture to build computers based on their Moore School experience. Goldstine and von Neumann were not sympathetic with Eckert and Mauchly's hopes to trade on EDVAC patents, and argued instead that the technology should be placed in the public domain. In a 1946 meeting, Army officials determined that von Neumann's Draft Report on EDVAC constituted a publication, and since it had been circulated publicly for more than a year, all patent claims were rendered invalid. This ruling further strained the relationship of von Neumann and Goldstine with Eckert and Mauchly.
This was not the last of the patent story. Eckert and Mauchly eventually sold their company to Remington Rand---and with it the rights to the ENIAC patent, which was in essence a patent for the general-purpose computer. Remington Rand merged with Sperry Gyroscope, and the new company, Sperry Rand, decided to use the ENIAC patent, which was finally issued only in 1964, to extract royalties from the major computer manufacturers. Honeywell balked at paying royalties, and the two companies sued one another in 1967. Five years and 32,000 exhibits later, the ENIAC patent was ruled invalid, in part because the computer had been used in December 1945 for a Los Alamos simulation, which was regarded as a public use of the computer that preceded the (1947) filing of the patent claim by more than a year. Another, more controversial reason for the ruling was that Mauchly had visited John Atanasoff in Iowa in 1945 and learned about the ideas for the Atanasoff-Berry computer, which was counted as a prior invention.
Readers interested in an objective account of these intellectual property issues are directed to Nancy Stern's From ENIAC to UNIVAC. Another account, from the perspective of von Neumann, can be found in this reviewer's book, John von Neumann and the Origins of Modern Computing (MIT Press, 1990).
McCartney dwells on Mauchly's fascination with the possibility of calculating the weather as driving his original interest in computing. As a physics professor at Ursinus College, Mauchly had his students using a second-hand desk calculator to look for correlations between rainfall and sunspot cycles. He built an analog device to do some weather calculations, and in 1939 he began to explore ways of building an electronic calculator to do these calculations. This was in fact what led to his collaboration with Eckert.
McCartney might profitably have contrasted Mauchly's efforts in weather prediction with the pioneering work of Jule Charney and von Neumann at the Institute for Advanced Study in the 1940s and 1950s. They chose weather prediction as their critical test of the computer as a scientific instrument because weather forecasting is one of those nonlinear hydrodynamical problems that is not open to analysis without a powerful calculating machine. They isolated the most critical meteorological features, developed numerical methods to handle the partial differential equations, and built the machine that proved the feasibility of numerical weather prediction. Their methods were soon adopted by the military weather services and the U.S. Weather Bureau, as well as by others in many countries. By contrast, Mauchly never really made any progress on the weather problem. Von Neumann was aware, but dismissive, of Mauchly's work in that area.
Eckert's place in history is clear. He made an extraordinary contribution by showing how reliability could be designed into a large-scale computing device comprising many unreliable vacuum tubes. Von Neumann recognized Eckert's engineering prowess and offered him a job as the chief engineer on the computer he was building in Princeton---an offer he eventually rescinded when Eckert would not set aside his commercial interests in order to build a computer for science. While von Neumann made job offers to several other members of the ENIAC team, including Goldstine, he never made an offer to Mauchly. Indeed, it is hard from McCartney's account to understand what Mauchly's contributions were. It is clear that Mauchly was an effective sounding board for Eckert and that he made some contributions to the systems design of the computers built at Penn and at his company. But exactly what these contributions are is not clear. This is particularly troubling since McCartney is trying to argue the case for a special place in history for Eckert and Mauchly.
The final part of McCartney's book tells the story of Eckert and Mauchly's commercial ventures. Despite its technical strength and employees who worked extraordinarily hard---at one time (McCartney claims) the official work week was 84 hours!---their company lasted only about four years before being bought out by Remington Rand. McCartney adopts a line of thought that historians of computing hear often: If only Eckert and Mauchly could have held out a little longer, and if only their external funding angel from American Totalizator Company had not died in an airplane crash, they could have succeeded and become the dominant company in the computer industry. Unfortunately, the business history of computing speaks strongly against this romantic view.
Virtually every technical start-up to manufacture computers in the 1950s and 1960s went bankrupt or was acquired by a business machines manufacturer that needed the technical expertise to make business computers. To succeed in the computer business, an entrant needed extensive capitalization, good management, an established customer base, and protection (such as a market niche) from the biggest competitors. Eckert and Mauchly were undercapitalized, mediocre managers (e.g., signing fixed-cost instead of development or time-and-material contracts), without a customer base, and unfocused in their business plan (selling a computer to anybody for any purpose). They had a chance to succeed as part of Remington Rand, but the company faded badly in competition with IBM---but that is another story.
William Aspray is executive director of the Computing Research Association.