Once the euphoria of finally getting an agreement had died down, it was necessary to start discussions with the technical arm of ICL rather than the marketing one concerning future directions. ICL also had to set up a project team and decide how the operation would be handled internally. With uncertainty as to whether a joint agreement would be reached with ICL right up until the end of August, almost no prior planning was possible.
In the period between September 1981 and the end of the year, both organisations attempted to get people into place and a plan of action.
The project had been carried out so far with almost no manpower being used. The amount of effort used in the SERC financial years 1979/80 and 1980/81 had been about 0.5 man years each year. In the first year, this was almost completely myself and Rob Witty with some involvement of David Duce. In the second year, some systems effort had gone into producing the demonstrations for ICL management and some user support effort in terms of public relations and demonstrations.
More effort had been used in 1981/2. In particular, Tony Williams had been involved in the technical assessment and had made visits to Three Rivers, Carnegie Mellon University and Apollo during the year. Even so, the effort prior to September 1981 had been small.
A major problem was that no manpower had been allocated in 1981/2 and the only manpower available in 1982/3 was 5 MY plus whatever would be given by projects on site. The Estimates bid for resources in 1982/3 were now so far underway that this was the only sensible course open. However, at the presentations to Boards and Council, this had been clearly stated and there was every hope that manpower would be made available.
On 4 August 1981, when the project looked as though it would be starting, Bob Hopgood wrote to Dr Manning indicating that there was a need to put between 10 and 15 MY of software development effort into the project in the early years and this needed to be found from within the Laboratory. The 5 MY to be provided by Council was to manage the project and would be used up in hardware acquisition, user support etc.
The matter had been raised at the RAL Division Heads Committee and all the Division Heads were very lukewarm. The two Science Board projects on site refused to provide any effort and after long discussions with Nuclear Physics, they also indicated that they were not prepared to put effort into the project. The ASR Board had made it clear very early on that they had no effort to spare apart from any involvement that could come via STARLINK. At that time, STARLINK was run from the Computing Division and ASR did help the mounting of GKS later on.
Bob Hopgood made it clear that if effort was not going to be made available, RAL should stop the project then and not make a mess of it. With the manning levels provided by Council and no effort coming from three of the Boards, the only way the project could succeed was to borrow effort from other Computing Division projects. Bob Hopgood urged that large projects on site should provide manpower both in terms of funding and people. Staff should be seconded to the project while staying on the payroll of their existing projects.
The Laboratory was not prepared to back the project and no additional funding was made available in 1981/2 or 1982/3. Consequently, the project was forced to exist on the 5 MY to be provided in 1982/3 together with whatever could be found from other Engineering Board projects prepared to contribute effort.
In terms of people, Keith Fermor with an electrical engineering background was released to work on local area network developments, Chris Webb (an applications programmer) was released by Laser Division to work on the GKS implementation, and Jim Loveluck, a theoretical physicist with little computing background, was made available. Although having a definite learning phase to go through, Jim Loveluck and Keith Fermor both made a good contribution to the DCS local area network activities which indirectly helped the Common Base Project.
By February 1982, eight people had been diverted from existing work and became the basis for the work done on the PERQ. To a large extent, my involvement in the project steadily decreased. The Division was attempting to replace its mainframe systems. This, together with other activities associated with publicising the PERQ, meant that the main management of the group rested on Rob Witty who, at that time, was Coordinator of the Distributed Computing Systems programme, and was attempting to get the new Roberts Software Technology initiative off the ground despite having no staff in post. Consequently, senior management effort was very small.
In the summer of 1981, Bob Hopgood made a number of attempts to get a manager (Principal Scientific Officer) from elsewhere in the Laboratory to run the project without success. He had made a strong bid for Ken Robinson in August of 1981. It was not until July 1982 that Ken was finally released. As a result, the project in its first six months was very much run by part-time senior management.
The structure of the people directly involved in the PERQ Common Base activities was as follows:
The personalities of the people involved:
Len had been the main systems programmer on the PRIME systems and had been at RAL for many years. He had no knowledge of UNIX but was noted for being quick to learn new systems.
The PRIME systems were in a stable state at that time and it was felt that he could be released without too much effect on the ICF.
Colin continued to be funded by DCS. Colin had started his career as a school teacher before moving to the University of Glasgow as a demonstrator. He had acquired a good knowledge of UNIX in that post and had been employed at RAL to provide UNIX support for the DCS programme. All the DCS university grant holders had standardised on UNIX as their operating system base where possible. The DCS programme had a small PDP11/34 at RAL running UNIX and later a PDP1l/70 was added. These were used to provide a development vehicle for local area network activities associated with the programme (mainly Cambridge Ring) and installed new versions of UNIX etc.
Colin Prosser was the one UNIX guru owned by RAL.
A High Energy Physicist by training with an interest in interactive computing. He had worked in the group at Oxford before leaving to work at Burroughs. He had joined RAL to work on ICF being mainly involved in the SIGTIP (Tools for Interactive Programming) activities.
Tony Williams provided the main compiler expertise on the project through his work at Burroughs although both Len and Colin had expertise in this area also.
A mathematician by training who had jut completed an excellent MSc with Prof Hoare's group at Oxford on the formal aspects of computing. She had been recruited in June 1981 to work on the Software Technology project where she had very relevant experience. During the latter half of 1981 she had also taken responsibility for the day to day system support of the PDP11/70.
As no effort was forthcoming from elsewhere and Liz was known to have a good systems background, the Division had little choice but to divert her to the PERQ project. This meant that the research in the Software Engineering area was put back by well over a year.
Previously a deputy shift leader on the mainframes. He was recruited to handle all the hardware aspects of buying, installing and maintaining PERQs. He also installed new versions of systems software and provide trouble shooting back-up where possible.
Provided the main technical secretarial support to the project and was involved in the PERQ-specific Cambridge Ring activities.
Had joined the Laboratory in September 1981 after a Computer Science degree and this was her first project.
A computer scientist from Brunel who had worked in the Division for two of her three sandwich periods. She did not actually join until February 1982 but was very quickly able to get up to speed due to her previous involvement with the Division.
In addition to these people, Chris Webb worked on the GKS implementation which was a separate activity. Ian Benest, who had been involved in the SIGTIP activities and was currently involved with MMI coordination, continued to provide a low level of assistance in benchmarking the PERQ together with the Apollo. The Common Base Project had made it clear early on that it saw the need for more than one system in the Common Base. Consequently, the Apollo system that had been purchased was updated as new releases were produced and continued to be used in assessing the suitability of the system for the Common Base particularly in the ICF CAD area.
During the whole of the period of the Common Base Project, the Computing Division was under complement. In April 1982, the Division's complement was 183 with an in-post figure of 163. Consequently it was very difficult to run any project in the Division at full strength.
The Common Base project was one of the few projects that year which received all the manpower that was allocated to it. This caused resentment in the Division from other projects but was necessary to ensure any kind of achievement in this difficult period.
Due to low salaries and the improving external market, it was very difficult to recruit and retain staff. The Division was also expanding with new projects which exacerbated the problem. By April 1983, the Division's complement was 190 with an in-post figure of 171.
During the whole of the main development phase of the Common Base Project the Division was about 20 under complement. Many senior staff had left and we were only allowed to recruit at the lower levels. Consequently, the ratio of management and technically skilled personnel compared with lower grade staff was the lowest in the Laboratory.
Even though we continued to advertise extensively, we were unable to increase our level of UNIX expertise from outside until July 1982 when one other person was recruited.
To indicate the high level of turnover, six of the ten people mainly involved with the project no longer work for RAL.
ICL started to move staff on to the project in October 1981. The funds available in the first year were £580K and a manpower of 30. The project was to have software developments done at Bracknell.
Chris French was the Director in charge of the Project and the Project Officer for most of the early period was Roger Ashbrooke. ICL appeared to be no different from SRC in getting staff to work on a new project. Their existing employers were reluctant to let people go and general internal vacancy notices did not get many people willing to transfer to Bracknell.
By January 1982, ICL had managed to get 14 people on the project with 6 more in the pipeline. The major shortfall was in software development and support where there were still 7 vacancies. Even so, ICL had many more people involved in the project than SRC.
By March 1982, the number of people working on the project had risen to 19. Consequently, the number of people involved in software development both at RAL and ICL was running at about 50% of the required levels throughout the crucial period. Promises by other projects in ICL to provide the project with staff on loan were never honoured.
The project to mount UNIX as the operating system on the PERQ was slow to get off the ground due to difficulty in getting staff to work on the project both at RAL and ICL. Time lost here could not be recovered. The aim in SERC was to go for UNIX as the initial operating system on the PERQ rather than make POS available as an interim system. Consequently, rapid build-up and keeping to timescales was important. On ICL's side, the major emphasis was selling hardware early on. While agreeing with the philosophy of moving to UNIX, ICL had to sell systems and, if that could be done early on using the POS operating system, that is what they would do.
