by Kurt W Beyer
The Instructional Tape Preparation Table was split into lower and upper sections, the lower section holding the more commonly used routines. The Table could print the content of an instructional tape in order to facilitate debugging before uploading the tape into Mark III’s internal memory. If a change had to be made, the coding machine could overwrite previously entered instructions.84
Example of operating instructions for Mark I. Courtesy of Archives Center, National Museum of American History, Smithsonian Institution.
Mark III’s Instructional Tape Preparation Table was a mechanized version of many of the concepts discussed in Wilkes, Wheeler, and Gill’s 1951 book on subroutines. The coding machine embodied programming innovations that Hopper and Bloch had developed between 1944 and 1946 while programming Mark I and Mark II. The question arises: Did programmers in the Harvard University Computation Laboratory and the Cambridge University Mathematical Laboratory have any contact between 1944 and 1951?
SUBROUTINES: AN EXAMPLE OF SIMULTANEOUS INVENTION
When historians consider the path of a technical innovation, they come across a strange and unexpected phenomenon: Often, different inventors invent a technology simultaneously in different geographic locations. Further investigation uncovers no other connections between these disparate groups, as is the case of such well-documented examples as the telegraph, the light bulb, and the turbojet.85 Without any formal communications between the different inventors, it is tempting to write off the occurrence as chance or coincidence.
But the frequency of simultaneous invention dismisses serendipity on statistical grounds. Why simultaneous inventions occur appears to be contingent on a variety of factors, including the society’s overall scientific and technical acumen and the process of how technological problems are defined and chosen. In the case of subroutines, the problems associated with programming were similar on the two sides of the Atlantic because the hardware was similar. Early pioneers in large-scale computing machines criss-crossed the Atlantic, attended joint conferences, and shared published and unpublished documents, the most famous being John von Neumann’s “First Draft.” Because of the similar hardware solutions, comparable programming problems arose as these machines became operational.
The more difficult question that arises is why members of the Harvard crew and the Cambridge staff came to such similar solutions. Since the concept of technological determinism carries little weight with most historians, it is not wise to conclude that subroutines were the best or the most logical solution to the problem. Yet why “closure” occurs for one expression of technology but not for another is far from transparent.86 Technical “closure” has its roots in the concept of scientific certainty. During the process of scientific inquiry, consensus within the scientific community tends to form around a certain interpretation or theory. A similar process occurs when consensus emerges around a particular technology design. Communities of engineers, designers, and consumers interact until a technology design stabilizes and reaches closure. The closure process has produced rear-chain bicycles, QWERTY keyboards, and gasoline-powered automobiles. It is important to remember that these technologies and others could have stabilized with other design attributes.87
In the case of subroutines, closure may have occurred because there was much more sharing of information between Harvard and Cambridge than first meets the eye. During the summer of 1946, Wilkes accepted an invitation by Howard Aiken to visit Harvard after the completion of the Moore School lectures.88 Wilkes arrived in Boston on 11 September and remained there for a week. The young Brit recalled observing Mark I in full operation as well as having a variety of conversations with Harvard staff.89 Hopper remembered his visit years later: “We had gotten to concepts of subroutines [by 1946] and I think a little bit had influenced Wilkes, who had been there from England.”90
Another connection to Wilkes runs through British mathematician Leslie Comrie. Comrie managed the British Nautical Almanac Office from 1930 to 1936 and founded the Scientific Computing Services Ltd. in 1937. As a leading proponent of calculating technology in Britain, Comrie kept abreast of developments in America and passed them on to Maurice Wilkes. It was Comrie who provided Wilkes with a copy of John von Neumann’s report on the stored-program concept, which heavily influenced the design of the British EDSAC.91 As the historian James Cortada observes, “his help to Wilkes was typical, and that of the EDVAC report was only one in a long series of favors to Wilkes.”92
In a review of Hopper’s Manual of Operation in the 26 October 1946 issue of Nature, Comrie noted that he was extremely impressed with the “brains of the machine” (that is, the sequence-control tapes). What struck him was the fact that these tapes represented a sequence of processes that were independent of the actual figures used, thus allowing the same tape to be used multiple times. He explicitly mentioned the Computation Laboratory’s “tape library,” which significantly lessened the coding burden and made Mark I more efficient than a well-equipped manual computer.93 Though it cannot be confirmed that Comrie passed knowledge of the Harvard tape library to Wilkes, we do know that Comrie was aware of Hopper’s and Bloch’s coding innovations and that he kept Wilkes informed of developments in the United States.
Wilkes returned to the United States in the summer of 1950 in order to learn more about a new machine being constructed by the Eckert-Mauchly Computer Corporation: the UNIVAC I. By that time, Wilkes’s Cambridge colloquium on programming had been meeting for a year. In Philadelphia, Wilkes got to spend some time with Grace Hopper and her programming staff. Hopper had left Harvard in 1949 to join the EMCC as a senior mathematician. Wilkes remembered particularly stimulating conversations with Hopper and her group: “I found that they had a full appreciation of the importance of programming and of the need to develop organized and disciplined methods. I felt that I was among people who looked at things in the same way that I did myself.”94
Whether or not Hopper and Bloch influenced Wilkes’s development of subroutines and subroutine libraries for stored-program computers, Wilkes was the first to publish on the topic. Though Hopper discussed subroutines in her 1949 Description of a Relay Calculator,95 she admitted that Wilkes, Wheeler, and Gill systematically defined the programming concept and disseminated it to a wide audience.
Ten years before her bold proclamation that she had written the first subroutine in 1944, Hopper candidly admitted in an interview that innovation at Harvard did not follow a smooth, logical path, and that at times Hopper and her Harvard colleagues were not fully conscious of the significance of their creations: “Those were essentially relative subroutines, though we did not know it [in 1945]; we didn’t realize it. In fact, I wasn’t aware of it fully until I got down to UNIVAC, when I had input-output under my control, and could store those on tape and call them back into a program without having to copy them.”96 Her confession does not take away from the accomplishments of the Harvard crew, but it reminds us that sometimes inventors do not understand the significance of their creations until well after the fact.
5 THE BEGINNING OF A COMPUTING COMMUNITY
During World War II, the crew of the Harvard Computation Laboratory constituted a significant portion of the emerging computer community. The members of the crew were bound together by shared ideas and knowledge, but being a member of the crew went beyond the sharing of technical information. The group cultivated its own style of being and sense of purpose. Howard Aiken, Grace Hopper, and the external influence of the Navy were all instrumental in developing this disposition, demeanor, and outlook. Ultimately the Harvard Computation Laboratory became more than just a place to house a large calculating machine. It was a community of practitioners capable of generating, capturing, and disseminating highly esoteric technical knowledge.1
WARTIME ISOLATION
During the war, the Harvard Computation Laboratory functioned in near isolation. One obvious reason was the fact that there were only a few other programs. In England, Alan Turing’s work at Bletchley Park
on the code-breaking machine Colossus was shrouded in secrecy. In the United States, Eckert and Mauchly’s ENIAC project at the University of Pennsylvania was in its infancy at the time of Mark I’s dedication. “To a considerable extent,” Robert Campbell recalled, “the Harvard project was sort of self-contained. It pretty much went its own way, and there was not an awful lot of interchange.”2
Hopper recalled hearing about a computing program at MIT during the war, but admitted that none of the Harvard crew ever interacted with their MIT counterparts. It seems strange that two programs so close geographically to one another would not share information, but a confidential memorandum to Harvard president James Bryant Conant from MIT president Karl Taylor Compton (dated 7 February 1940) sheds light on the matter. Compton wanted to make sure that Conant was mindful of two facts. First, for several years there had been a undisclosed program at MIT that had developed numerical computing devices. Second, the MIT program and its members were interested in obtaining patents for their work. As a result, according to Compton, “there could be no cooperation between Dr. Aiken and the MIT group as concerns the methods involved.”3 Compton goes on to question the wisdom of developing a computing program at Harvard: “The lack of a practical engineering viewpoint permeating the institution” looked like a significant handicap to the MIT president, both in the development and utilization of such computing machines. Dependence on IBM was believed by Compton to be another notable shortcoming, for “the utility of such a machine, born fully developed, would be impaired by the fact that the necessary personnel would not have passed through the evolutionary stages with the machine.” In contrast, the growth of the computing center at MIT had been an evolutionary process. It had started with Vannevar Bush’s analog differential analyzer. As of February 1940, the center had a variety of calculating machines and a seasoned staff, all housed in a 5,600-square-foot space. The tone of the memorandum forcefully suggested that the development of computing devices should be left to MIT, though that wasn’t stated explicitly.4
Compton’s implied advice went unheeded, largely because Howard Aiken persevered and because IBM and the Navy backed him. On 7 August 1944, Aiken celebrated his triumph with a formal ceremony marking the transfer of the Automatic Sequence Controlled Calculator (Mark I) from IBM to Harvard University and the founding of the Harvard Computation Laboratory. Aiken captured the event in pictures , and the Computation Laboratory worked closely with the Harvard News Bureau to prepare a news release.
The release was picked up by Boston’s newspapers (the Herald, the Globe, the Daily Record, and the Post), and also by the New York Times, the Christian Science Monitor, and the New York Herald Tribune. Headlines such as “Navy Man Inventor of World’s Greatest Calculator” and “Aiken’s Calculator” infuriated IBM’s president Thomas Watson and began a well-documented feud between Aiken and IBM that would last through Watson’s lifetime.5 Despite Watson’s accusations and objections, Aiken’s crew rallied around their leader. When questioned about the feud in the years to follow, Hopper maintained her staunch support of Aiken, insisting that the overall concept of the machine, as well as the design of the sequence mechanism and interpolators were his, not IBM’s. “IBM never gave the credit they should have to Aiken,” she stated.6
Aside from the tensions between Aiken and IBM, the formal dedication on 7 August 1944 marked the start of a new era in mechanical information processing. Harvard faculty members and administrators and high-ranking Navy officers attended the event. Most notable among the latter were Rear Admiral Edward L. Cochrane (chief of the Bureau of Ships), Rear Admiral A. H. Van Keuran (director of the Naval Research Laboratory), and Rear Admiral Julius Furer of the Office of Naval Research. Those three naval organizations would play particularly important roles in the future development of the computing industry.
Mark I listed as one of the ten most important scientific advancements of 1944 in Popular Science. Courtesy of Archives Center, National Museum of American History, Smithsonian Institution.
JOHN VON NEUMANN AND MARK I
Also present at the dedication of Mark I was the mathematician John von Neumann. Von Neumann played a pivotal role in the early history of computing, for he authored one of the most influential papers in the field, “First Draft of a Report on the EDVAC.”7 The “First Draft,” dated 30 June 1945, outlined the architecture of a stored-program, general-purpose machine (the basic architectural framework on which present-day computers are based). Von Neumann’s paper quickly became the working blueprint for the next generation of computers, including EDSAC and ACE in England and UNIVAC and the IAS computer in the United States.
The “First Draft” was also the center of a considerable controversy involving J. Presper Eckert Jr., John Mauchly, and John von Neumann. Herman Goldstine, the US Army’s liaison officer to the ENIAC project in 1944–45, holds von Neumann most responsible for the ideas presented in the document: “This report represents a masterful analysis and synthesis by [von Neumann] of all the thinking that had gone into the EDVAC from the fall of 1944 through the spring of 1945. Not everything in there is his, but the crucial parts are.”8 Through the years Eckert and Mauchly disagreed adamantly about Goldstine’s interpretation. The inventors of the ENIAC claimed that they had developed many of the concepts found in the “First Draft” during the spring of 1944, months before they even met von Neumann. Despite their claims, it was not until 3 months after the release of the “First Draft” that Eckert and Mauchly recorded their ideas (in a paper titled “Progress Report on EDVAC”).9
If Goldstine’s interpretation is correct and von Neumann was responsible for the architecture concepts in the “First Draft,” the question remains: what was the source of von Neumann’s inspiration? Goldstine identified the source as the famous mathematician’s “genius.” Further historical inspection, however, shows that the time frame in which von Neumann developed the ideas for the famous paper coincides with his four-month stay at the Harvard Computation Laboratory. Though it is too strong to assert that the work at the Harvard Computation Laboratory served as his inspiration, there is no denying that von Neumann served as a bridge between the Mark I and ENIAC computer projects during the fall of 1944, or that his vision of a general-purpose stored-program computer grew out of his conversations and observations in Cambridge and Philadelphia.
John von Neumann. Courtesy of Los Alamos National Laboratory Archives.
Identified during his youth as a gifted mathematician, John von Neumann received his doctorate in mathematics from the University of Budapest at the age of 22. After a three-year stint as a Privatdozent in mathematics at the University of Berlin, von Neumann became a visiting professor at Princeton University. As a result of Adolf Hitler’s ascendance to power in 1933, von Neumann decided to remain in the United States and accepted a position as a professor of mathematics at Princeton’s Institute for Advanced Study. While at the IAS, von Neumann became increasingly interested in applied mathematics, and particularly in problems concerning supersonic and turbulent fluid flows. Mathematical representations of dynamic fluids required knowledge of non-linear partial differential equations, and by the commencement of World War II von Neumann was one of the leading experts in that field. Since shock and detonation waves fell under his expertise, von Neumann was vigorously sought after by the Army’s Ballistic Research Laboratory and the Navy’s Bureau of Ordnance.
His growing reputation gave von Neumann unparalleled access to classified projects, including the top-secret Manhattan Project. For the Manhattan Project’s scientists and engineers working in secrecy in the New Mexico desert, von Neumann’s knowledge of fluid dynamics was highly valued, and he joined the group as a consultant in late 1943. The Los Alamos team was considering various ways to transform a non-critical mass of fissionable material, such as uranium or plutonium, into a critical mass as rapidly as possible. Once critical mass was achieved, the scientists theorized, an atomic chain reaction should release tremendous amounts of energy. One theory (pro
posed by Seth Neddermeyer) was to forcefully implode a sub-critical sphere by means of conventional explosives in order to compress it into a critical state. Von Neumann was called in to help develop a mathematical model for Neddermeyer’s implosion process. The resulting partial differential equations were extremely difficult to solve for the Los Alamos human computers, with their desk calculators. In March 1944, in an attempt to speed up the process, von Neumann asked his friend Emory L. Chaffee about the possibility of using the new Harvard machine. Chaffee was the head of Harvard’s physics laboratory and Howard Aiken’s former graduate advisor. With time of the essence, von Neumann replied: “The possibility of making these calculations on your new device was an exceedingly tempting one.”10
Aiken agreed to allocate machine time to von Neumann’s problem, and von Neumann arrived at Harvard for the dedication ceremony on 7 August 1944. It was, Aiken recalled, von Neumann’s first exposure to large-scale calculating machines. “He wanted to see the machine and I told him how it worked and what it did.”11 Impressed with the machine’s potential, von Neumann decided to work with the Harvard crew and attempt to solve the implosion problem on the experimental machine.