Introduction

The 48 Moore School lectures were held from July 8^th to August 31^st 1946, at the Moore School of Electrical Engineering, University of Pennsylvania, under the auspices of the Office of Naval Research, U.S. Navy, and Ordnance Department, U.S. Army. They were designed to disseminate the current knowledge and progress in electronic computation, in which the Moore School was pre-eminent.

The lecturers were drawn from all facets of computation, including the tradition of mathematical computation by hand, electro-mechanical computers (e.g. the Harvard Mark I), the ENIAC, and the design of the proposed new generation of stored-program electronic computers.

Invitations to attend were extended to interested parties in the U.S. and beyond. 28 people attended from 20 different organizations, including Maurice Wilkes from the University of Cambridge and I.J. Good on behalf of Max Newman of the University of Manchester.

The common pattern of the course was to have one (or sometimes two) of the 48 formal lectures in the morning, and to have informal seminars and discussions in the afternoon.

Background

The Moore School had become the centre for the development of electronic computation during the war. In 1942 an outpost of the U.S. Army's Ballistic Research Centre was set up there, led by Lt. Herman H. Goldstine, to make extensive use of their two Bush Differential Analyzers. Staff were drafted in to the Moore School from other institutions, and the ENIAC project was started in 1943 to build the first large scale, programmable, purely electronic computer, primarily for the calculation of ballistic tables for the Ordnance Department of the U.S. Army. The idea of ENIAC derived from ideas circulated in the Moore School by J.W. Mauchly in 1942.

The shortcomings of ENIAC as a general purpose computer were well understood before it was completed, but the design had to be frozen in order to produce a useable computer for the war effort as soon as possible.

In the end ENIAC was not fully completed till the end of 1945, but work had already started on its successor for the Ordnance Department, named EDVAC, which was to be a stored-program computer.

The "First Draft of a Report on the EDVAC" was produced in June 1945. It was written by John von Neumann, based on the thinking of the ENIAC team, which von Neumann had recently joined. A progress report was produced in September 1945, which outlined a number of possible designs. Early in 1946 it was decided that the Moore School should start work as soon as possible on a small stored-program computer, to gain experience, and that the Institute for Advanced Study, at Princeton, should carry out a study for a large-scale comprehensive computer. A contract was signed for the Moore School project in April 1946, and by October the Moore School presented three possible detailed designs, of increasing sophistication, and a decision was made to produce a machine at the lower end of the range. The EDVAC designs were based on a Mercury Acoustic Delay Line ("MADL") memory.

A preliminary report on the proposal for an IAS machine was produced in June 1946, by A.W. Burks, H.H. Goldstine and John von Neumann. This planned to use for its store the Selectron Tube being developed by RCA. This was the most suitable memory method proposed at the time, being electrostatic, with immediate access to any bit, and no periodic refreshing of the tube required.

There were other organisations interested in building their own variation of the stored-program computer, for example the Naval Ordnance Laboratory. They were currently investigating a third method of storage, which was electrostatic, but based on the Cathode Ray Tube, as used in radar and television (in contrast to the Selectron, which was more like a multi-faceted vacuum tube); this required periodic refreshing, which was inconvenient. It was the proposed storage mechanism closest to the one that Williams and Kilburn got working.

Note that by the summer of 1946 the top level of the ENIAC "team" had effectively broken up, with Eckert and Mauchly having formed their own company and von Neumann working on the design of the IAS machine, back at Princeton.

The other major piece of design for a stored-program electronic digital computer at that time had been carried out by Alan Turing at the National Physical Laboratory near London. By the end of 1945 he had produced a detailed design for the proposed ACE computer, using a Mercury Acoustic Delay Line store. By the summer of 1946 he had produced a number of enhanced designs, and the NPL, who had no electronics personnel of their own, were looking around the U.K. for personnel to help build the design. By August 1946 they had approached Freddie Williams of TRE. He had recently visited the Moore School (but did not attend the Moore School Course), and showed interest in getting involved with computers. But in the end he declined the offer, preferring to concentrate on just trying to provide an effective store for use in computers, based on the Cathode Ray Tube he was so familiar with from his wartime work in radar. This was incompatible with the NPL project, as Turing's design was closely tied to the MADL storage mechanism and to overcoming its disadvantages.

Also in the UK, by August 1946, Newman at Manchester had obtained a grant to build a computer, and Wilkes at Cambridge had decided to build one, based on the EDVAC principles and MADL storage. But neither had produced any significant design work.

The Published Record

Plans for the Moore School Course were only made within two months of the course starting. Wire recordings were made of each lecture, and lecturers were asked to produce a manuscript of the lecture, using the wire recordings if they wanted. These would be turned into a published version of the lectures. Some lecturers in fact rewrote the lecture for the manuscript version. Some took a long time to produce them. Some lecturers never did.

Volumes I and II of the lectures were published by the Moore School in late 1947 (up to lecture 21). Volumes III and IV of the lectures were published on June 30^th 1948.

Pressure of work on the EDVAC meant it was not possible for the Moore School staff to provide summaries of the lectures (using the wire recordings) for all the missing manuscripts. The worst sufferer was Vol. IV, where 6 out of the 15 lectures have no record, and one has a two page summary by a member of the Moore School staff from the wire recording, plus a set of diagrams. The remaining records vary in length from about 4,500 words to 7,500.

A book "The Moore School Lectures" was published by the MIT Press in 1985, edited by Martin Campbell-Kelly and Michael Williams. It contains the printed versions of the lectures, and has a brief introduction to each lecture by the editors. There is also other general background and commentary on other aspects of the course, in particular drawing material from the detailed notes made by one of the attendees, Frank M. Verzuh. The book also includes a description of a number of lectures given between lectures 41 and 42 by the ENIAC staff, at the request of the attendees.

The Volume IV Digest

The lectures in Volume IV are much more concerned with higher-level considerations and possible machine architectures, rather than individual facets of the subject. They are arguably the most interesting from a modern viewpoint. So I have written a 12,000 word digest of those lectures for which there is a record. The digest emphasises a programmer's viewpoint, rather than an Electrical Engineer's or a Mathematician's -- reflecting my own interest and competence!

The style of the digest varies between (1) a straightforward summary of the material in a lecture, following its structure, (2) commentary on the material, and (3) a complete rewrite of material in an attempt to give a clear and concise version of its content. The particular style should be obvious from the context, but in particular italics have been used to distinguish commentary from (1) and (3) within a lecture. Some of the comments made derive from the 1985 reprint. Italics are also used, e.g. as presumably and I think, to indicate places where assumptions have been made about facts that are not in (or not clear in) the original text.

Note that I have not always kept strictly to 1946 terminology, most obviously using "bit" to mean a binary digit. Note also that simple sentences and paragraphs that are obvious to us now may have taken paragraphs or pages in the original lecture! Note also that throughout the lectures "EDVAC" is frequently used to refer to any proposed new-generation stored-program machine, not just to a machine close to the EDVAC eventually built (which used a Mercury Acoustic Delay Line store).

The Volume IV Lecture List

Acknowledgements

As luck would have it I found a copy of (just) Vol. IV at the The (UK) National Archive for the History of Computing (having reached there via our Departmental Library). The NAHC is in the next building to me! My thanks especially to Jon Agar of the NAHC for providing me with the extended access needed to produce the digest of Vol. IV. A further special thankyou is due to Michael Williams, co-editor of the 1985 reprint, who has since lent me a copy of the book. This should (already has!) enabled me to correct and improve the digest at leisure, and possibly extend it.

My thanks are also due to the Moore School itself, for Paul Shaffer for showing me the ENIAC Museum there, and Professor Mitch Marcus for providing helpful information and advice.