Since the last report on this subject progress meetings have continued to be held regularly at the West Gorton factory of I.C.T. (meetings on 21st April, 23rd. June, 9th September, 16th October).
On 9th September the position was that the handover appeared likely to be delayed by one month. I.C.T. still expected the hardware to be moved from West Gorton to Aldermaston at the end of October, in accordance with that clause of the contract. I.C.T. had previously hoped to do this at the beginning of October, thereby gaining an extra month for the commissioning process at Aldermaston.
Of the many contributory causes to the hardware delay, three appear to be worth specific mention. First, the disc file had been delivered from Data Products (U.S.A.) two months late, arriving at the end of August. On the other hand, it worked satisfactorily within a week of being received. As a result, it is almost certain that the disc hardware will be included as an integral part of the system when the handover does eventually occur. It is quite possible, however, that the software enabling us to use the disc will not be available until up to a month later than the handover date.
Secondly, the core storage had been proving troublesome. It appears that the printed circuitry associated. with the core stores (provided by Plessey) makes extensive use of a particular type of resistor which is not effectively insulated by its external coat of paint. The body of the resistor is very close to a conducting element, so that handling of the cards sometimes causes contact, with break-down of insulation. (This fault has also been experienced recently on the N.I.R.N.S. Atlas 1.) I.C.T. admit this mistake to be particularly deplorable because these resistors were previously known by I.C.T. and Plessey to suffer from this defect. To replace all the resistors would apparently involve severe penalties in time and money. I.C.T. plan to try and patch up this defect by spraying the printed circuits with varnish to provide an extra insulating layer. If this device is successful the system as a whole will not be delayed thereby, as the four banks of core storage can be dealt with one at a time, without putting the system out of action.
Thirdly, the Slave Store is not yet working. The design of this part of the machine demands unusually close tolerances on the tunnel diodes involved - closer than is generally achieved in production. Accordingly, stringent selection has to occur before a set of components can be assembled, which are good enough to make the Slave Store work. Since the Slave Store is not a necessary part of the Atlas 2 system, and in any case is likely to produce only a ten per cent increase in speed on average for our type of work, we may have to make the decision whether to accept Atlas 2 initially without the Slave Store, or alternatively to refuse to accept the Atlas at all without the Slave Store. The Slave Store is contractually a part of the Atlas 2 system, so the A.E.A. would be legally justified in taking the latter course. As a delay in delivery of Atlas 2 would cost the A.E.A. even more (in extra Stretch rent or loss of computing power) than it would cost I.C.T. (in penalties), however, it would be in the interests at the A.E.A., even more than in the interests of I.C.T., to choose the former of the two possible courses of action.
Because of the delay in availability at West Gorton of the Aldermaston Atlas 2, it has been impossible for I.C.T.'s team to keep up to schedule with the testing of their software. Accordingly, the software for which I.C.T. is responsible, including the Supervisor, will probably be one month late. Fortunately, this is the same period as the expected lateness of the hardware. It has become clear, at Atlas Owners Policy Committee meetings and elsewhere, that I.C.T. now recognise that they considerably underestimated the magnitude of the task of getting the Atlas 1 Supervisor working. As a result they have been considerably more guarded in their commitments regarding the Atlas 2 Supervisor, and, at the same time they have assembled a more powerful team to build. the Atlas 2 Supervisor. Accordingly, disappointment in the performance of the Atlas 1 Supervisor need not be reflected in predictions of the performance of the Atlas 2 Supervisor.
The S3 compiler, for which A.W.R.E. is responsible and which is essential for our use of Atlas 2, is expected to be ready by the time Atlas 2 is handed over, in spite of the fact that one of the I.C.T. men working on this task under A.W.R.E. supervision unexpectedly resigned to take up a position abroad.
A.E.A. programmers have already been informed of most of the foreseeable differences between Stretch S2 and Atlas S3, and this information should be brought to a state of virtual completion early in October.
Through the welcome co-operation of the N.I.R.N.S. Atlas Laboratory, test compilations of certain A.W.R.E. programmes will be carried out on the N.I.R.N's. Atlas 1 during the month preceding the handover of the Aldermaston Atlas 2.
The Atlas 2 system, comprising all the hardware and software contracted for delivery by December 31st, 1964 (with the possible exception of the Slave Store), is likely to be handed over at Aldermaston on January 31st, 1965. Also handed over at the same time is likely to be the disc hardware, but probably not the disc software.
Acceptance tests of the system without the disc will start immediately the system is handed over. Acceptance tests of the system with the disc will start immediately the necessary disc software becomes available.
