The effect of the Flowers Report on the Laboratory over the next five years will be twofold:
The first point should have a serious effect on our future. In general, the policy of the Laboratory so far has been based, quite rightly, on the assumption that our machine, being much more powerful than the University machines, provides a service for Universities to run programs too big for their own machine and to ease general overloading problems. This has meant that a large number of compilers have had to be available on the machine and that these compilers have been flexible to allow for the difference in dialects in the major languages. Hartran, Algol and EMA all exhibit this aim of providing maximum compatibility with other dialects.
When the Universities start feeling the effects of the Flowers Report this will not be the case. The updated KDF9s should provide facilities, as far as Algol, Fortran and EMA is concerned, of the same order as the ones obtained on Atlas at present. The same should apply to the large 1900s.
I feel we are moving into an era where the Laboratory will have periods (ie just after a new machine has arrived) when there will be a definite increase increase in speed to be had by coming here but that in general this will not be the case. Therefore I think it is more reasonable that if we are supplying a service to the Universities that it should be more on the line of quality rather than quantity. At present a large number of the users come to Atlas because of facilities available rather than machine speed and I think that this is the direction we should go. Work on these lines could be in the following areas:
This can be divided into three subsections:
There are two or three obvious candidates here although I am sure a lot of others are possible. I am thinking of hardware which is too expensive for local universities but which should be available in a Regional Centre.
The laboratory already has a fairly large library of Algol and Fortran routines which could be extended. The new regional centres will require such a library and if a joint structure for such a library could be set up it would be very advantageous. The present library suffers being rather poorly vetted especially in the case of Algol and a large amount of work could be put into to producing a decent library of well-tested and documented routines. This should be machine-based possibly with the aid of the SC4020 and consoles.
It is essential that somewhere on the British Isles a good Library of Computing Periodicals and Books exists. It is also essential that coupled to this is a good method of Information Retrieval. The Laboratory would be a good place for such a library as we have the basis of one already and we have the computer facilities for doing the Information Retrieval. The main drawback at present is the lack of staff to run such a library and the present incompetent method of obtaining books and periodicals. The 'assistance' of the Rutherford and Harwell Libraries would have to be removed and all ordering, cataloguing etc done ourselves. This will require staff and room to accommodate such a library. In addition a proper Xerox instead of the one we have at the moment would have to be hired for reproduction from both books and periodicals.
The second main effect of the Flowers Report will be more machine time available for the Laboratory's own use. The service projects outlined above should take up a certain amount of this time but, even so, there should be sufficient time available for research projects to be initiated in the Laboratory independent of our service work. the difference between service work and research work is rather academic as service work in a large number of cases may be treated as research. For example, I have already mentioned under the heading of service such items as:
Additional topics will depend to a large extent on the Fellows we have and the programming staff available. If any additional Fellows are made, it would be desirable that they should be associated with Computer Theory rather than branches of Mathematics or Physics which incidentally use computers. Partly, I feel, because we have sufficient non-computer Fellows and also for fertilising the programming group.
The amount of work suggested above could easily take up the complete time of 20-30 programmers. These need not all be of the SO grade and above and it is possibly an advantage to have some lower grade staff in the programming group. A fair proportion of the work being done by the programming group is very menial in nature and ought not to be done by programmers of SO grade and above. The present set up is rather a waste of highly-paid manpower. In this connection I feel that the support activities in the Operations Group should be expanded to two or three people to provide a buffer between the users and programming group. Too much of the programming group's time is wasted by elementary questions from users. The answers are usually available in the literature which has not been read. It is the job of the Programming Group to document their work. It is not their job to teach people to read. The probem is one of educating the user and, to a large extent, I feel it is the job of the Operations Group.
The only research project that I would like to see started in the Laboratory (apart from ones already described) is some work on Automatic Learning or Game Playing. Little work has been done in this field over the last few years and I am sure worthwhile results could be achieved given sufficient machine time and people interested in the subject. There is a fair amount of interest in the subject in the Laboratory and, with the aid of a reputable Fellow (not a crackpot), I am sure the effort would be worthwhile. The field id fairly wide open and we have a powerful machine available with possible some surplus machine time.
