Assuming that the decision to implement a SERC Common Base Policy is agreed, any audit must establish whether this was achieved in reasonable timescales in the period from October 1981 when the Memorandum of Understanding between ICL and SERC was signed.
The main features of the Common Base were:
The main features of the Memorandum of Understanding were:
With the exception of the upgrades to the I/O Board, the PERQ hardware in the Common Base and the MOU were honoured. In particular:
The main failures were:
Insufficient progress on the reworked I/O Board which seriously hindered the Cambridge Ring work. Three Rivers were doing the changes and ICL were very slow at passing on requirements.
The reason was the decision by ICL not to go with the local UK LAN standard but to base its policy on the emerging de facto ethernet standards.
Blame can be attached to ICL who failed to honour its commitments in this area, particularly as the DoI office pilot was dependent on the work and this was a contract between ICL and DoI.
In retrospect, it was a mistake to base the UK local area network strategy on Cambridge Ring without putting the effort into ensuring an agreed BSI standard specification that was pushed through ISO as an international standard. The situation was not helped by Logica changing the design half way through and Cambridge who had a negative interest in seeing the Ring become a product.
These activities were eventually done, but too late, by the JNT and DoI.
A major conclusion must be that the UK must not base major policy on de facto UK standards without investing the effort needed to achieve international standardisation of these UK activities.
Lack of involvement by SERC in the future development plans, as agreed in the Memorandum of Understanding. ICL treated SERC as a large customer but no more. There was never during this period an active presence by SERC in ICL's hardware developments.
Even so, the failures were minor compared with the successes and did not impede progress during the period 1981-1982. They would have been more a problem if other failures had not masked them.
The Cambridge Ring hardware developments at RAL were completed and the commercial Logica Polynet nodes were successful both in the UK market place and abroad. By September 1982, Logica Polynet Cambridge Rings were in use as far away as Australia. The UK had a world lead in ring technology which still exists today.
By the end of 1982, instead of a large virtual memory UNIX system, a basic UNIX system was available with a crude virtual memory system and a reduced set of utilities. This was a major impediment to progress during this crucial period. Its lateness of arrival meant that PERQ systems had to be shipped out with the POS operating system, causing a support problem that RAL could not cope with. Section 24.12 has given a summary of the problems and possible causes.
The major reasons for the failure to get UNIX available in reasonable timesca1es are as follows:
Lack of Manpower: in the period until March 1982, the level of manpower in the project was totally inadequate. On the SERC side, this was due to none being provided by Science, Nuclear Physics and ASR Boards. On the ICL side, attempts to get people to move to Bracknell were unsuccessful.
RAL is convinced that if the manpower available early on had been doubled, the project would have been completed by mid 1982.
Two versus One UNIX Implementations: the decision to do two UNIX implementations rather than one obviously wasted effort. A decision to mount only one UNIX would have been preferable, but which one? In hindsight, the decision should perhaps have been to go for Microcode UNIX with paging and virtual memory support. That would almost certainly have been an acceptable product but it would have taken a lot longer. RAL doubt if such a route would have produced a viable UNIX system before the autumn. If all the effort had gone into Accent UNIX, would the performance problems of program start-up have been solved? ICL certainly had the effort and expertise to solve them if the CMU design was viable. The Accent UNIX route did produce the required system in all aspects but one.
RAL do not believe that major blunders were made in this development. Both ICL and RAL staff worked extremely hard and got no rewards or recognition of this because they did not deliver the goods.
It was clear that the Accent route was attractive but had higher risks, being dependent on a product under development. Was the decision to go for a back-up as well wrong or sensible management?
Unfortunate Decision at ICL: the decision by ICL to stop work on Microcode UNIX at a time when SERC's Accent UNIX was incomplete and no benchmarks done was a mistake. It was triggered by the need for an agreed future operating system between itself and Three Rivers, and believing Three Rivers' claims of the state of Accent UNIX. ICL were also under financial pressure at the time.
To ICL's credit, the decision to mothball the product in a state where it could be resurrected in quick time was a major factor in getting started on that product again quickly.
Both FORTRAN and PASCAL were available on Accent UNIX. The move to Microcode UNIX caused the delay and was inevitable once the switch was made. By the end of 1982, neither FORTRAN or PASCAL were available. If the product had not been mothballed earlier in the year, FORTRAN would probably have been available before the end of 1982.
The lack of PASCAL highlighted again the problems due to different objectives. ICL were keen to mount their range-compatible ICL PASCAL system as this would allow existing office automation software to run which used the extended ICL facilities. SERC required a standard ISO PASCAL.
The aim was to use the York Box as a PERQ-WAN connection. A contract with York had been set up. Without UNIX little progress could have been made. When UNIX arrived, this product became available. RAL do not believe there was anything wrong with this strategy. There were delays in the York box caused by ICL's inability to supply essential documentation and system sources to York.
A project to connect a Cambridge Ring to the Wide Area Network via a gateway was eventually set up by the JNT and produced a product.
Any delays in this area were dependent on other areas and could not be avoided.
Both NAG and GKS became available on the PERQ as fast as they did on its competitors. RAL do not believe the project could be faulted in this area.