Browse Books

Reviewer: James Cecil Hammerton

Aspray's work is an addition to those books that try to delineate for us the origins of computing. It is a scholarly work addressed specifically to the contributions made by John von Neumann. The text occupies 249 of the book's total pages, and is supported by 83 pages of notes, 19 pages of general bibliography, 11 pages listing von Neumann's writings, and a modest 9-page index. The text relates the impact of computing on John von Neumann and his impact on it. Chapter 1, “A Mathematical Research Career,” outlines his early years from 1903, the year of his birth, until he became a citizen of the US in 1937. “By the time von Neumann arrived in the United States he had already established an international reputation based on his contributions to operator theory, logic, and the mathematical foundations of quantum mechanics” (p. 13). He was also well known for his published works in the fields of mathematical economics and game theory. He continued his contributions in these fields, interrupted somewhat by World War II, and in 1944 co-authored a book with Oskar Morgenstern [1]. Von Neumann was not a one-dimensional person, however. “By the late 1930s, von Neumann was established as one of the leading international mathematicians and mathematical physicists. He had not yet had any significant contact with computing, though he was knowledgeable about numerical methods…” (p. 16). Chapter 2, “An Education in Computing,” describes von Neumann's war work and his subsequent heightened interest in high-speed computing. Through his work for the Ballistics Research Laboratory, he had contact with the design engineers of the ENIAC at the Moore School of Engineering at the University of Pennsylvania. This work led to his association with the development of ENIAC's successor, the Electronic Discrete Variable Arithmetic Computer (EDVAC). In the spring of 1945, he wrote his “First draft of a report on the EDVAC.” This report, which has been a focal point of the dispute about who invented what and when, “presented the first written description of the stored-program concept and explained how a stored-program computer processes information” (p. 39). The report was not published formally. It was distributed to the Moore School staff in late June of 1945. The author notes that “a mere 100 pages of mimeographed text gives the fundamentals of the stored-program computer” (p. 41). (Perhaps the lack of publication accounts for discrepancies between the author's quotes and a version of the report appearing on pages 355–364 of a book edited by Brian Randell [2]. This version, entitled “First draft of a report on the EDVAC,” is divided into sections on “Definitions,” “Main Subdivision of the System,” “Procedure of Discussion,” “Elements, Synchronism Neuron Analogy,” and “Principles Governing the Arithmetical Operations.” It quotes Contract No. W-670-ORD-4926 under which the work was done at the Moore School, and is dated June 30, 1945.) The report defines what has since become known as the von Neumann architecture, which has been the basis for all stored-program computer designs. The controversy revolves around the extent to which von Neumann can claim to be the inventor of the architecture and, in particular, of the idea of storing a program in a computer as well as the values of the variables on which the program works. It is difficult to recapture the context of those times when rows of mechanical desk calculators with human operators were the standard way to perform mathematical computations; when these components were superseded by punched-card calculators; when the programming was done through the medium of a plug board that had to be set up for every computation. ENIAC adopted this method of programming. Von Neumann observed that to advance beyond this primitive and time-consuming way of programming, some means of storing the program in the computer must be arranged. The difficulty, of course, was the limited size of computer memories. The computer memories of those years blossomed in a variety of forms: mechanical switches, cathode ray tubes, mercury delay lines, and eventually core memories. The concept of the stored program was hostage to the established technology. The controversy was also complicated by the dictates of wartime security. Engineers, scientists, and mathematicians in the early 1940s did not rush into print with their customary enthusiasm because their work was classified. One apparent contender for the first stored-program machine was not declassified until 1975. The COLOSSUS was an electronic computing machine used at Bletchley Park in England for cryptographic work. In his paper [3], Brian Randell wrote, “It seems fair to classify the COLOSSUS as a special-purpose program-controlled electronic digital computer.” On the other hand, he continued, “it was externally programmed and there is no question of its having been a stored-program computer.” In an attempt to resolve the controversy, the September 1990 issue of Computing Reviews carried a first—an invited paper [4] by Saul Rosen of Purdue University on “The Origins of Modern Computing.” The paper was accompanied by comments from most of the persons identified by Rosen and then a final response from Rosen. John Atanasoff, whose name is associated with the Atanasoff-Berry Computer (ABC), did not contribute a comment but his son stood in for him. Whether this attempt contributed anything to the resolution of the question of who invented the first electronic digital computer is not clear. Aspray's book does not resolve the issue, either. Perhaps it is not important that it be resolved. We are not, as far as I know, intending to build a statue to the inventor of the stored-program idea. Chapter 3, “Planning a Computer for Scientific Research,” recounts von Neumann's efforts to secure funding for a computer for scientific research from his position at the Institute for Advanced Study. Chapter 4, “Engineering a Computer,” discusses the result of his efforts—the progenitor of several copies of the machine built and used throughout the world. Chapter 5, “The Transformation of Numerical Analysis,” takes up von Neumann's interest in numerical analysis. “By the 1930s numerical analysis was in a state of decline. It was regarded as an unfashionable subject in the mathematical community and was not attracting much top mathematical talent. The introduction of the computer changed this situation” (p. 117). Chapter 6, “The Origins of Numerical Meteorology,” and chapter 7, “The Computer as a Scientific Instrument,” continue the record of von Neumann's vision of and work with high-speed computation as a practical, indispensable means for solving problems. Chapter 8, “A Theory of Information Processing,” documents his contributions to information processing, biomedical computing, the theory of automata, the practical task of building automata, and the connection between the computer and the brain, leading to his contributions to cybernetics. After the war, von Neumann was in demand by governments and industry. He consulted for, among others, IBM, the National Research Council, the Los Alamos National Laboratory, and the Oak Ridge National Laboratory. In 1957, still planning further research in “high speed computing machines and, especially, in their application to problems of geophysics and meteorology” (p. 251), von Neumann was admitted to the hospital in California and died of cancer on February 8, 1957, at age 54. This book is a valuable contribution to the origins of computing. It is not for those who prefer to skim the surface. Von Neumann by this account possessed a brilliant, disciplined intellect. Whatever his contribution to the invention of the stored-program computer, he left his mark on the many problem areas affected by high-speed stored-program computing.