The focus of George Dyson's well-written, fascinating but essentially misleading book,'Turing's Cathedral', is curiously not on celebrated mathematician, code-breaker and computer theorist Alan Turing but on his equally gifted and innovative contemporary John von Neumann. Von Neumann, whose extraordinarily varied scientific activities included inter alia significant contributions to game theory, thermodynamics and nuclear physics, is especially associated with the early development of the electronic digital computer (i.e. the 'EDC'), an interest apparently sparked by reading Turing's seminal 1936 paper 'On Computational Numbers' which attempted to systematize and express in mathematical terminology the principles underlying a purely mechanical process of computation. Implicit in this article, but at a very theoretical level, was a recognition of the relevance of stored program processing (whereby a machine's instructions and data reside in the same memory), a concept emanating from the work of mid-Victorian computer pioneer Charles Babbage but which demanded a much later electronic environment for effective realization.
What Mr Dyson insufficiently emphasizes is that, despite a widespread and ever-growing influence on the mathematical community, Turing's paper was largely ignored by contemporary electronic engineers and had negligible overall impact on the early development of the EDC. Additionally, he omits to adequately point out that von Neumann's foray into the new science of electronic computers involved a virtual total dependence on the prior work, input and ongoing support of his engineering colleagues. Invited in August 1944 to join the Moore School, University of Pennsylvania, team responsible for ENIAC, the world's first general purpose computer being built for the US Army, von Neumann was quickly brought up to speed courtesy of the machine's lead engineers, J. Presper Eckert and John Mauchly. As early as the fall of 1943, Eckert and Mauchly had become seriously frustrated by the severe processing limitations imposed by ENIAC's design and were giving serious consideration to implementing major modifications, in particular the adoption of Eckert's own mercury delay line technology to boost the machine's miniscule memory capacity and enable a primitive stored-program capability. These proposals were subsequently vetoed by the School's authorities on the quite understandable grounds that they would seriously delay ENIAC's delivery date; instead it was decided to simultaneously begin research on a more advanced machine (i.e. EDVAC) to incorporate the latest developments. As a new member of the group, von Neumann speedily grasped the essentials of the new science and contributed valuable theoretical feedback, but an almost total lack of hands-on electronic expertise on his part prevented any serious contribution to the nuts and bolts of the project. Relations with Eckert and Mauchly rapidly deteriorated when an elegantly written, but very high-level, document of his entitled 'First Draft of a Report on the EDVAC' was circulated among the scientific community. Not only had this document not been previewed, let alone pre-approved, by Eckert and Mauchly, but it bore no acknowledgment whatsoever of their overwhelming responsibility for much of the content. By default, and in view too of his already very considerable international reputation, the content was therefore attributed exclusively to von Neumann, an impression he made no attempt thereafter to correct, the term 'Von Neumann Architecture' being subsequently bestowed on the stored program setup described in the document.
The public distribution of von Neumann's 'Draft' denied Eckert and Mauchly the opportunity to patent their technology. Worse still, despite academic precedents to the contrary, they were refused permission by the Moore School to proceed with EDVAC's development on a commercial basis. In spite of his own links to big business (he represented IBM as a consultant), von Neumann likewise opposed their efforts to do so. All this resulted in a major rift, von Neumann thereafter being shunned by Eckert and Mauchly and forced to rely on lesser mortals to help implement various stored-program projects, notably the IAS computer at Princeton. The following year (1946) Eckert and Mauchly left the School to focus on developing machines for the business market. Before doing so, they jointly delivered a series of state of the art lectures on ENIAC and EDVAC to an invited audience at the School. Among the attendees was British electronics engineer Maurice Wilkes, a fellow academic of Turing's from Cambridge University, but with relatively little interest in the latter's ongoing activity (by this time Turing, a great visionary, had also turned his attention to designing stored-program computers). Blown away by Eckert and Mauchly's presentation, Wilkes returned to England to forge ahead with a new machine called EDSAC, which was completed in May 1949 and represented the first truly viable example of a stored program computer (an experimental prototype christened 'Baby' had already been developed at Manchester University the year before). Back in the US, Eckert and Mauchly continued their efforts, but persistent problems with funding and also Eckert's own staunch refusal to compromise on quality delayed progress, their partnership finally culminating in the development of the UNIVAC 1, the world's first overtly business-oriented computer, delivered initially to the Census Bureau in March 1951.
Mr Dyson is quite right of course (and he does this well) to trace the beginnings of the modern computer to the stored program concept, but his obsessive focus on von Neumann's role obscures the impact of Eckert and Mauchly's vastly more significant contribution to its development. The triumph of the EDC depended almost wholly on the efforts and expertise of utterly dedicated and outstanding electronics specialists like them, not on mathematicians, logicians and generalists like von Neumann or even Turing. Never one to deny credit where it was due, Wilkes (who later spearheaded advances in software, became the doyen of Britain's electronic community and ended his long and distinguished career as professor emeritus of computer science at Cambridge) unceasingly acknowledged his major debt to Eckert and Mauchly. Hopefully, Mr Dyson, a writer of considerable talent, might one day decide to tell in full their story and set the record straight.