We have signed a collaborative agreement with the Smithsonian Institution, National Museum of American History, which promises to enhance the computing collections of both institutions.

"This is the first such formal joint collecting agreement the Smithsonian has made with a museum of the stature of The Computer Museum," said Dr. Arthur Molella, chairman of the Department of the History of Science and Technology, of the Smithsonian's National Museum of American History and a member of The Computer Museum's Board of Directors. He further explained, "The field is so large and there is so much to do that it's necessary for us to make agreements in important collecting fields with the leading specialized museums."

The joint arrangement with the Division of Computers, Information, and Society of the National Museum of American History is broad in scope, affecting historical research, preservation and exhibitry. We will cooperate in creating a common catalog and database of our collections. This is being carried out by a group composed of David Allison and Jon Eklund from the National Museum of American History and Gwen Bell and Lynn Hall from The Computer Museum.

The common goal in our collecting agreement is to make sure that all the important artifacts are preserved. Considering that computers are now entering their fifth generation and that the classes range from supercomputers to personal computers, the amount of material worth saving is growing. Cooperative collecting is essential for preservation.

The Scientific Instrument Commission of the Union Internationale d'Histoire et de Philosophie des Sciences is also cooperating in the effort to develop a complete listing of computer artifacts held around the world. While the collection of The Computer Museum is international, many national and specialized museums preserve many significant machines from their regions. For example, The Science Museum, Kensington, has much of the known Babbage equipment and The Deutsches Museum, Munich, has a collection of the machines built by Konrad Zuse. The Computer Museum is proud to have the NEAC 2201, one of the first transistorized computers in Japan built by NEC, components of EDSAC, Maurice Wilkes' Cambridge University computer that is the first fully operational stored program machine, and other non-U. S. computers. One of the goals of The Computer Museum is to show that computer innovations are not unique to one country, to one company, or to any one institution.

Saving the history does not just mean collecting artifacts. For each artifact, a technical file is also needed. Such material includes manuals, notebooks, photographs, and other accounts of the development and use of the machine.

This report lists new acquisitions to the collections of the Museum. The listing illustrates the diversity of the collection. We have chosen to feature some ephemera on the cover, because this material is often thrown out or thought to have little value. On the contrary, ephemera are important because they can quickly evoke the spirit of an era. Don't throw out your memorabilia. Send it to us. Do it all at once, or one at a time. Several times a year, we receive a small envelope from Phil Dorn - it always has a surprise spec card from an early computer or some other piece of ephemera. When Lynn Hall, the registrar and I open it, we generally smile the rest of the day. You too can make our days happy.

Gwen Bell
Founding President

John M.M. Pinkerton

John Pinkerton was the chief architect of LEO, which stood for Lyons Electronic Office. LEO, the world's first commercial data processing computer, is a direct descendant of the architecture outlined in draft EDVAC Report via the Maurice Wilkes' EDSAC. LEO I was being used for payroll and office data processing in early 1954, prior to similar use by GE of their UNIVAC 1.

John Pinkerton tells his own story, that of a young physics graduate of Cambridge, who joined The Lyons Company and was inspired by Comptrollers John R.M. Simmons (who died in 1985) and T. Raymond Thompson (who died in 1972) to build the first data processing computer. G.B.

This is an abstract of a talk given by John Pinkerton at The Computer Museum on October 4, 1987. A paper on which this talk is based is copyrighted by John MM. Pinkerton.

John M. M. Pinkerton and Gwen Bell at the talk. Pinkerton presented the Museum with a marketing film to sell LEO II computers that was made by Lyons in 1957.

The Lyons Company

How did it happen that J. Lyons & Company Limited, a wholesale food and catering business, came to build a computer for its own use and then go into the computer business?

In the 1890s, the Salmon and Gluckstein families started a business at Cadby Hall in West London to cater for functions at the Olympian exhibition halls next door. By the beginning of World War II, Lyons had a high reputation for efficiency with the general public. They ran a variety of wholesale food businesses distributing tea and coffee, ice cream, bread and cakes, as well as other lines throughout the UK. Lyons did not believe in using wholesalers but sold and delivered directly to twenty or thirty thousand small retail shops. They also had a chain of about 150 tea shops, four or five large hotels, various restaurants, and an outdoor catering division. They employed 30,000 people. All this meant routine accounting for a vast number of small transactions with clerical efficiency since margins were small.

In the early twenties, the need for effective methods of accounting was recognised and resulted in the recruitment of a young mathematics graduate from Cambridge, John R. M. Simmons. A few years later he recruited another Cambridge mathematician, T. Raymond Thompson. By the mid-thirties they had rationalized the Lyons office practice and brought it under their management. In LEO and the Managers (1962), Simmons wrote, "...the curse of routine clerical work is that without exercising the intellect, it demands accuracy and concentration . ...(He was) looking forward to the day when machines would be invented which would be capable of doing all this work automatically."

By the thirties, Simmons and Thompson were not only studying the accounting and calculating machines on the market but using some in intensively unorthodox ways. For instance, they were pioneers in the use of the Kodak Recordak camera for processing bakery orders. Punched card systems were not generally favored. Lyops believed that the rationalisation of clerical procedures that was needed to transfer work to cards would not prove to save costs.

When I joined Lyons in 1949, Simmons was Chief Comptroller and Thompson was Chief Assistant Comptroller. They were in charge of some 2,000 clerical staff who worked in large open plan rooms with 200 clerks carrying out routine payroll calculations, order processing or invoice passing. I found the atmosphere in the clerical department was one of high seriousness of purpose and dedicated loyalty on the part of the staff. Systematic grading of clerical jobs had been pioneered in Lyons' offices and rewards and promotion were strictly related to merit, which was regularly reviewed.

Simmons was totally dedicated to management and especially to the collection and application of management information using the computer. This came out clearly in his second book: The Management of Change (1970).

Thompson joined Cadby Hall in 1928, after working in a Liverpool department store. He had the quickest intelligence of anyone I ever met. Much of the inspiration for the hardware development as well as the software of the LEO project came from him. He could quickly visualize ways to organize anything from a complex clerical task to a language compiler down to the finest detail. He maintained the enthusiasm and set the intellectual and management tone of the Leo project from its beginning in 1947 to the merger with English Electric in 1963.

The Birth of LEO

In 1947, Lyons sent Thompson and Oliver Standingford to the USA to investigate the "giant brains" that were then being reported in the British press. An introduction was obtained to meet Professor Goldstine at Princeton who was associated with the ENIAC and specifying the EDVAC. I heard Thompson's account of this meeting several times. It apparently lasted an hour or so. In the first half hour, Goldstine explained to Thompson the principles of the stored program computer. In the second half hour, Thompson explained to Goldstine just how the computer could be employed on routine clerical work, such as payroll and invoicing.

Ironically in the USA, they learned about Maurice Wilkes' EDSAC project underway at Cambridge University. They lost no time in going to Cambridge on returning to the UK. After seeing Wilkes' project, Thompson told me he was impressed by the squareness of the pulses he saw on an oscilloscope in the lab (even though he had no knowledge of electronics). Thompson and Standingford formed a favorable impression of Wilkes' work reported for the Lyons Board. Only a fragment of this has been preserved: "We believe that they have been able to get a glimpse of a development which will, in a few years time, have a profound effect on the way in which clerical work (at least) is performed. Here, for the first time there is a possibility of a machine which will be able to cope, at almost incredible speed, with any variation of clerical procedure, provided the conditions which govern the variations can be pre-determined. What effect such a machine could have on the semi- repetitive work of the office needs only the slightest effort of imagination. The possible saving from such a machine should be at least £50,000 a year. The capital cost would be of the order of £100,000.

"We feel therefore that the Company might well wish to take a lead in the development of the machine and, indeed, unless organizations such as ours, namely the potential users, are prepared to do so, the time at which they become commercially available may be unnecessarily postponed for many years."

In November 1947, the Board agreed to contribute £2,500 to the cost of the EDSAC and to lend Ernest Lenaerts to Wilkes for six months, which turned out to be nearer a year. Lenaerts, who had worked for Lyons for several years, was employed in electronic engineering during his war service. In return, Wilkes agreed to give Lyons whatever details of the EDSAC design they might need to build a machine for their own use. Lenaerts reported to Thompson on progress.

The EDSAC being built at Cambridge University Mathematical Laboratory. Maurice Wilkes is kneeling in the center behind the mercury delay lines.

During 1948 Lyons tried, and failed, to find a contractor to build a machine like the EDSAC. Later that year they decided in principle that when Wilkes' machine had been shown to work, they would build a version of it themselves. They therefore advertised for someone to take charge of the engineering and I applied for the job.

When the anonymous advertisement appeared I suspected it was from Lyons. I had returned to Cambridge after the war to work with Mr J. A. Ratckffe at the Cavendish Laboratory only 100 yards away from the Maths Lab where Wilkes worked. I wrote a thesis on ultrasonic absorption in liquids using a pulse method which, as it happened, was an ideal preparation for work on computers using delay lines for storage. I first met Wilkes as an undergraduate through the University Wireless Society. In the summer of 1948, Wilkes told me about Lyons' interest in building a computer.

In December 1948, I went for an all day interview at Lyons which was extremely impressive. Not only was I given an excellent lunch, but Mr G. W. Booth, the venerable but alert Director of the clerical department who was over 80 years old, carne out to interview me. He asked me if I thought I could make this machine work. I optimistically said I could, but added that as it needed several thousand valves, it would be difficult to make reliable, which naturally turned out to be correct. On December 18th, I got married and in mid January, 1949, I started to work for Lyons.

Construction of LEO I

Lyons decided to wait to start building their copy of EDSAC after it was demonstrated to work by computing a table of primes. In the meantime, I went to a Lyons training course given to their office supervisors and spent several weeks in Cambridge absorbing the design of EDSAC and its logic and circuit techniques. Ernest Lenaerts and I set up a small workshop to build a delay line and circulate pulses in it. We also tried some ideas for transmitting pulses over long leads because we thought LEO would be physically bigger than EDSAC.

In February 1949, I started discussions on how to deal with the input and output problems revealed by the payroll program. It was recognized that input data would typically fall into one of three categories: (1) current data reflecting events since the last run of the job, (2) data brought forward from that run and (3) semi-permanent data not requiring to be repunched for each run. Similarly results would fall into at least two categories: (1) results to be printed and acted on (e.g., wages to be paid) and (2) results to be carried forward.

The EDSAC being built at Cambridge University Mathematical Laboratory. Maurice Wilkes is kneeling in the center behind the mercury delay lines.

We decided that LEO needed multiple channels for both input and output, to be fitted with buffers that could be read in a single operation and were large enough to hold all the data items of a given kind, e.g., one person on the payroll. We also decided that LEO needed means for converting and reconverting data and results automatically in each channel, rather than using subroutines within the machine. The first of these decisions turned out to be excellent but the second was bad in execution and probably in principle. Once EDSAC began to work, events moved fast at Cadby Hall. A large room in the office block was allocated, staff were transferred from other departments or hired from outside, and Lyons excellent drawing office got to work drawing up racks and chassis to carry the circuits. A contractor was appointed to build the units, and a revised design of the EDSAC batteries of ultrasonic delay lines was drawn up making full allowance for engineering tolerances.

The overall orgmization of LEO I as it was built.

Putting LEO to work From the start Lyons believed that their own staff should create programs for LEO which would be suitable for the work of their offices. In 1950, David Caminer who had been in charge of Lyons Systems Research Office was appointed to take charge of all LEO programming. Payroll was to be the first main routine task. Since a breakdown in the middle of a two hour job could be serious, the concept of putting out restart totals at the end of each department in the payroll was established early on.

Handmarked documents at J. Lyons & Company Ltd. to be automatically read and transferred to punched paper tape by LEO in the early sixties.

Punched card machines were used for all channels, except those carrying current input data, for which we chose punched tape. Binary, not decimal numbers, were punched into the cards as a compact method of carrying forward results from one run of a job to form the data for the next.

We estimated that doing the conversions by subroutines would take up 90% of the time of LEO I. But the conversion and reconversion problems were solved when Lenaerts recognised that if the binary values of 10, 100, 1,000 and so on were stored in a matrix of the new germanium diodes, then the control circuits for multiplication and division taken over from EDSAC could be adapted to control the two conversion operations. They took no more than 10% of LEO's time, an acceptable overhead. About 1,000 new germanium diodes were used, accomplishing a task that would have been impractical with hot cathode diodes used elsewhere in LEO.

In 1953, when the machine was ready for use, the entire clerical staff of Lyons, numbering more than 2,000, was brought to Cadby Hall in batches of 30 to see a demonstration of LEO. Within six months, LEO produced part of the payroll for the Ford Motor Company at Dagenham. Early in 1954, LEO started doing the Cadby Hall payroll and other jobs followed rapidly, including bakery sales and tea shop orders.

LEO Lives On ...

Lyons saw LEO I as a considerable success and it remained in service until 1954. They invested in a design of an improved model for small scale production and sale, LEO Mark II, which ran about three times faster than LEO I. Eleven LEO II's were built - all of them slightly different.

Before the end of the fifties, it became obvious that a parallel transistorized machine was feasible and we embarked on LEO III. This was a 40-bit parallel machine of advanced architecture incorporating (as we later found out) many ideas also used in the IBM 360 series. Besides the multiple, buffered I/O channels provided by LEO I and II, it had multi-radix, as well as floating point arithmetic, extremely effective checking of data recorded on tape, direct input and output to and from main store (DMA) and 4 protection tag bits to each word in store allowing multitasking with up to 15 jobs. It was, I believe, also the first machine using microprogramming to go into production anywhere in the world.

About 150 LEO III's were built and sold. However, the capital demands of a growing business persuaded Lyons in 1963 to merge the computer department with English Electric. Later they sold their half share of the joint company to English Electric. While no LEO III's remain in use, a few System 4 machines from English Electric and 2900 series models from ICL have micro coded implementations of the LEO III instruction set, thus allowing one valued customer to continue to use an interlocked suite of programmes originally written for LEO III in the midsixties.