The Influence of Harwell
By Professor Ian Pyle, Founding Professor of Computer Science
Introduction
In May 1972, the University of York invited applications for a new chair of “Computational Science”. At that time I had been at the Atomic Energy Research Establishment (AERE), Harwell, for nearly fourteen years, and had come to know (for reasons explained below) that the teaching of computing in the UK (at least) was grossly inadequate. So I applied, assuming that the title of the chair was a stilted attempt to identify serious computing (exposing the limited awareness of the field by the University).
The University appointed me, and changed the name of the chair (at my suggestion) to “Computer Science.” I did not know what I was letting myself in for, and neither (I suspect) did the University! My only experience of a university had been as an undergraduate and then a research student at Cambridge, and I had enjoyed the excellent ethos and management structure at Harwell. It took me several years to establish a similar condition at York. In this note, I try to explain how it happened.
Computers at Harwell
Although the primary objective of AERE Harwell was to do the science and engineering necessary for nuclear power reactors, this entailed a great deal of computing (e.g. studying the partial differential equations of flow for neutrons in a fission reactor and of plasma in a fusion reactor). As a direct consequence, Harwell was at the forefront in the field of computing in the U.K.
Electronics Division had already designed and brought into use two new computers (the Dekatron and the CADET) before I came in 1958 to join the Computing Group in Theoretical Physics Division: not making computers, but using them. There I became involved with the detailed support for all staff at Harwell with their computing needs, initially using a Ferranti Mercury computer, which was significantly more powerful than CADET. My role was to help users in all aspects of computing, from numerical analysis, the extension of mathematical models, to the debugging and modification of programs to work on our computers.
At that time, the only computers were “main frames” which executed programs as a sequence of “jobs.” As well as using the Mercury locally, much work was sent for execution at the more powerful IBM computer at AWRE Aldermaston, and later the IBM 7090 at Risley, the headquarters of the UKAEA's Industrial Group. This was an extensive computing service, including several computers at several sites: boxes of cards had to be taken, and outputs returned, by car – or later by telephone lines (before there was any thought about data links or interacting computers).
During my time there, we were using the most powerful computers available, and contributing to the introduction of the Ferranti Atlas computer, particularly by a team (led by me) writing the FORTRAN compiler for it, so that Harwell's computing work could be easily transferred. Subsequently we chose an IBM/360 computer (for which I was involved in the negotiation) and I developed a multi-access system, HUW, so that users could contact it more easily.
From 1965, the introduction of small computers such as the PDP8 brought a significant change for Harwell scientists (as elsewhere): computers could be used to control and monitor experiments, and greatly assist data collection and retrieval. The extensive use of small computers in Nuclear Physics Division created a wealth of expertise in computing, which became an important new asset for the establishment. The Electronics Division had already developed units (CAMAC) to connect computers to experimental apparatus, but there was little software to support them.
In other words, Harwell had not only traditional “computation” - scientific (or business) calculation of numbers, but also the new “real-time systems” in which computers interacted directly with their environment, becoming important in industrial computer systems: factory automation and numerically controlled machine tools. This recognition of the potential uses of computers was the vision that Harwell gave me to bring to York.
Collaboration
Although some work at Harwell involving nuclear processes was classified under the Official Secrets Act, most (including all in electronics and computing) was open, and there was positive encouragement for collaboration with others. This applied particularly in computing, and between Harwell and similar establishments elsewhere in Europe. Thus software was freely exchanged, and common problems discussed and solved. Travel to conferences and between sites was made as easy as possible, so that I became a member of a community familiar with the scientific uses of large and small computers.
I had already become involved in the setting up of a European Workshop on Industrial Computing Systems (EWICS), and this continued after I came to York, eventually leading to the development of Ada.
Air Defence
In about 1968, at the height of the “cold war”, there was a political row about the system being developed to protect the UK from air attack, by monitoring and controlling air traffic above and around the United Kingdom. Major radar stations had been build across the country, feeding their data to a command centre near London Airport, where a computer system with a team of operators was supposed to integrate and interpret them. This was called “Linesman.”
After installing all the computers (specially designed to be resilient, capable of working continuously 24/7/365¼), it was discovered that no-one knew how to write the software to make them work! The main contractors (Plessey) had tried and failed to find a British software company willing and able to produce the software. (This was all revealed several years later in a report by the Science and Technology Select Committee, 1970-71)
Harwell had already started “diversification” - moving into new areas of research (because the reduced need for Nuclear Physics research), and the Computing Group was told to investigate this problem (which was classified under the Official Secrets Act), to identify the difficulties and to deal with them. I was involved in this from the beginning. It turned out that the system was badly specified, and that no-one knew how to program (in a sequential language) interactions between computers or between operators, or of anything significant about Software Engineering.
A team of about 20 people from Harwell, led by me, (mostly Nuclear Physicists) were seconded to work with Plessey to write the software. The installed hardware was mostly large computers, but used in ways more like the small computers in Nuclear Physics at Harwell: integrating data from many and various kinds of input/output.
Making the software for this Radar Data Processing System (RDPS) involved starting from scratch: I (with two others) wrote a detailed manual explaining all the software engineering processes involved, and introduced a management structure suitable for the size of the task. I specified the structure and the top-level design of the software; Harwell and Plessey staff developed all the rest of the software, and in 1972 handed over a working system to the RAF for serious use.
This experience revealed the inadequacy of the programming skills and training that then current practice had produced. I resolved to improve this situation – I knew that it would be necessary to teach about computers and software engineering for the next generation, to avoid a perennial “software crisis.”
Flowers report
The 1966 Flowers Report established the ground rules for computers for research in universities. (No mention of teaching at this stage!) The government accepted the substantial points of the report, and set up the Computer Board to oversee the installation of computers in all the British universities, including York, which as a small university was provided with an Elliott 4030 computer.
In the present context, it is relevant to note that Brian Flowers, chairman of the Joint Working Group that produced the report, had been the Head of Theoretical Physics at Harwell from 1952 to 1958 (just before I arrived), and later Rector of Imperial College, London. The influence of Harwell and the UK Atomic Energy Authority is evident in the membership and conclusions of the group.
As well as recommending particular computers for the universities, the report specifically mentions computer software: fast Fortran and Algol compilers capable of producing good object code in relocatable binary. This follows the Harwell approach, used for FORTRAN on the Ferranti Atlas computer.
Strategy for CS at York
Approaching the responsibilities of the “Department of Computation” (renamed soon as “Computer Science”), I identified the three distinct areas of Research, Service and Teaching.
From my experience at Harwell, I had no problems with Research and Service, but Teaching was the important new area, as I had discovered in Linesman. In my view, the subject needed three strands:
- Theory, particularly mathematical logic and formal algebras;
- Hardware, particularly digital electronics and storage media;
- Software, not just programming of all kinds, but due consideration of resource constraints, scale and complexity: real Software Engineering.
Having met the existing members of the department, I saw a significant gap in our expertise: commercial data processing, and its associated concern for databases. Consequently, one of my first priorities was to recruit someone who could fill this gap. (Fortunately, I already knew a suitable person!)
Growing from the subsidiary subject status took several years, and the syllabus was fleshed out as explained by Bill Freeman.
The Computing Service needed expanding to meet the growing need of the university, as had been anticipated by the Flowers Report and the subsequent Computer Board. I was able to work with the Computer Service Advisory Committee, to produce a statement of our requirements that formed a good basis for negotiation with computer manufacturers, leading to our eventual acquisition of a DECsystem 10 computer. As at Harwell, co-operation between computing services was common, and York joined with Leeds, Bradford, Hull and Sheffield to co-ordinate our approaches to the Computer Board
Research took a back seat initially, as the other areas needed urgent attention. I had already started researching the problems of computer – computer interaction while at Harwell (generalising a Linesman problem), and had identified Real-Time Systems as a key area to work in. Broadly, the problem was how to deal with input-output, not just with human users. This involved both data representation and timing.
I was strongly aware that research was needed in relation to both Teaching and Service: the communication between computers for service provision was as important as the interaction between programmer and compiler for students learning about software. (Subsequently I lost this battle, as the attitudes in the two parts of the department diverged.)
We eventually started with Modula as a way of handling input-output, then moved on to Ada. It is pleasing to see that Real Time Systems still form a significant element of the Department's research activity.
Conclusion
Starting the new Computer Science department was not easy: the university's intention had been a continuation of the original “Computation” department, and my appointment caused a significant change of direction . The university was heavily arts-based, with science departments that in 1973 were rather old-fashioned, and no mention of engineering. The financial squeeze during the late 1970s did not help. However, the department did establish itself in the university and nationally, becoming the seventh in the UK by the time I left at the end of 1985. It is rewarding to see the department now flourishing and respected within the university.