The Annual Report covers the year ending September 1985. As the Division was formed in July 1984, this first report covers the period from the formation of the Division until September 1985.
The aim is to produce a Report primarily for internal use indicating what has been achieved in the year and by whom. It is hoped that the timing of this Report is such that the RAL Annual Report entry can be constructed from the Divisional Report.
The Computing Division was split into two separate Divisions in July 1984 for a number of reasons. The Computing Division had been growing in size and had a complement of over 200. The decision by the Computer Review Working Party to return control and funding of the Interactive Computing Facility and Single User System programme back to the Engineering Board meant that a Division of viable size could be formed whose interests were in the information systems area in the widest sense and which was funded completely by the Engineering Board. It was similar in role and purpose therefore to the Science Board funded divisions such as Laser and Neutron.
At its formation, the transfer of funding back to the Engineering Board was still in progress but by April 1985, Informatics Division was almost completely funded by the Engineering Board from the following projects:
As the MMI work had only just started, the initial Group Structure was:
Neither of the last two Groups had a group leader and it was necessary to get approval for posts to be established before recruitment could take place. In the meantime, Bill Sharpe acted as head of the IKBS Group, and the Distributed Computing Group ran as three separate sections:
Ken Robinson was responsible for developing the activities in the MMI area.
Major management changes during the period were the resignation of Bill Sharpe, and the move of Cliff Pavelin to Informatics to take charge of the IKBS/MMI area.
By July 1985, the structure was:
The format of this Report is based on the structure in position at the end of the year.
The role of the Software Engineering Group is as follows:
All but the Research activity is funded directly from the SE Director's budget, with 75% coming from DTI and 25% from SERC. The DTI funding is because these activities are services on behalf of both the academic and industrial communities.
The Research work is either funded by Alvey as part of the SE Programme or by SERC's Computing Science Sub Committee (CSSC) which is responsible for long-term research. The Alvey funded research is collaborative with industry, while the CSSC research need not be.
Prior to the Alvey Programme commencing, SERC had started a Specially Promoted Programme (SPP) of research in SE (the SERC's Software Technology Initiative). Many grants had been awarded and an infrastructure started. R W Witty was the coordinator of this SPP. At the start of the Alvey Programme, these grants were included in the SE portfolio providing continuity and involvement of the relevant SE researchers in the Alvey programme.
The SE programme currently consists of 47 new academic awards of which 11 are fully collaborative with industry. There are 15 new awards to be announced shortly. Of the 40 projects included initially, 23 are still running.
Support for the Alvey Programme consists of interaction with academic groups to assist in the preparation of grant applications and coordination of the SERC/Alvey Directorate interface.
Research supported by Alvey differs from that usually supported by SERC in that the Alvey Programme is a directed programme. This involves much closer involvement in both the preparation of the grant applications and especially during the execution stage of the research. For this purpose all projects are monitored, both technically and financially. This involves preparation of a database on all projects, collation of reports, generation of summaries and spend profiles. Regular reports are required from Monitoring Officers for collaborative projects and from industrial Uncles for academic only projects. Any exceptions raised by this monitoring process are actioned quickly.
Support is provided also for the Alvey SE Office in Millbank by providing assistance in meetings, workshops, panels and presentations. Much effort is expended in maintaining the files and reporting procedures for the programme. To assist in maintaining continuity of research, a mailshot service is operated. An increasingly important aspect of the overall support as the programme develops is the analysis of the budget and preparation of data for forward planning. For this purpose a special procedure has been developed with the objective of minimising difficulties associated with over commitments of the budget.
R W Witty, F M Russell and D C Findley, supported by Lilian Valentine and Janice Gore, undertake the above work.
ACARD (the Advisory Council for Applied Research and Development) is a high level body which reports to the Cabinet on long term issues. ACARD is currently studying the state of the UK software industry, its future prospects and the role of software engineering in all aspects of UK industrial activity (ie Manufacturing and Services).
The study is being conducted by a working group chaired by John Coplin, the Design Director of Rolls Royce. R W Witty, through his Alvey role, is a member of this working group.
The study has been conducted on a part time basis for much of 1985 and a report is likely to go to the Cabinet in 4Q86.
The period has been one of transition as the section completes previous involvements arising from DCS, STI and the Common Base and begins to build up its new research links. Day-to-day activities have concerned the completion of Cambridge Ring developments, conversion of ML/LCF to the new Standard ML, and the import and evaluation of theorem-proving tools both for our own use and on behalf of the Alvey SE infrastructure.
Theorem provers are software tools to aid the process of proving theorems. In this context we are primarily concerned with theorems about programs, for example that a particular program possesses a particular property.
Three current systems, ML/LCF, IOTA, and Boyer-Moore, have been acquired and are being studied to varying degrees. A fourth system (Gypsy) that SEG would like to know more about is not available owing to a US embargo!
The Cambridge version of ML/LCF was mounted on RLVC and then ported to the Atlas 10 under UTS (A D B Cox and A J Kinroy before he left). The UTS version has demonstrated the benefit of mainframe power for large proofs and is already being used so by external researchers at Cambridge. Proofs that previously would have been run overnight on a VAX are now accomplished in terms of minutes on UTS.
IOTA is a modular programming system, built in Japan, which includes a significant verification capability and has many features of a prototype second-generation Alvey IPSE. Experience with the approach will be of particular value for the IPSE 2.5 project, a large industrial-academic collaborative Alvey SE project.
Investigations have continued into how best to make IOTA available in the UK. IOTA is implemented in an ancient dialect of Lisp which turns out to be a forerunner of PSL (Portable Standard Lisp). PSL runs on a variety of machines (DEC20, VAX, Apollo) and has now been acquired for VAX Unix. An assessment is being made of the work required to port IOTA to run on top of PSL. A few problems have inevitably been encountered but are gradually being resolved.
The latest version of the Boyer-Moore theorem prover has been obtained from the University of Texas. We have yet to work with this system ourselves but are acting as a UK distribution point. The system has been distributed to about a dozen industrial and university sites.
The section has actively followed the development of the new Standard ML. C P Wadsworth attended the Edinburgh meeting in May 1985 to finalise the language. The standard is a consolidation of tried-and-trusted developments in the ML community since the original language was designed for the Edinburgh LCF project. Principal additions are: the data constructors and pattern matching facilities of HOPE, exception values, a richer and more systematic set of definition constructs, a module facility based on the latter, and I/O handling. Final documentation from Edinburgh is awaited. The language has Alvey backing and SEG expects to use it for their theorem proving work.
P M Hedlund joined the group in January 85 and has been working on converting ML/LCF to Standard ML. The new parser and type-checker are complete and he is now working on the rest of the system (code generator and run-time support). It will remain then to transliterate the 20 or so LCF sources files written in old ML to Standard ML. The aim is to ensure that LCF, as the largest application of ML to date, is fully available for Standard ML. (The modules facility is not needed for LCF and is not being implemented in the first instance.)
The plan for SEG infrastructure envisages a variety of Unix systems linked by Newcastle Connection. It was felt that SEG could achieve this quickest by using the Ring since SEG had one in place and common interest and a common machine range with Newcastle. A lot of effort has been consumed in the attempt.
PERQ BBP was at last completed, the VAX driver converted to BSD 4.2 (by Unix Systems section), and the LSI11 BBP driver has been upgraded to CR82 standard (by D R Gibson). A simple file transfer program that runs directly over BBP was quickly implemented (by D R Gibson) and tested successfully between PERQ, VAX, PDP11, and LSI11. The program provides file transfer similar to PERQ PUFTP, with a five-fold increase in throughput (despite the slowness of PERQ BBP!).
Release 1.1 of the Newcastle Connection for PDP11 was finally received after months of contractual wrangling and installed on the PDP11/34 and LSI11/23. D R Gibson did sterling work assessing the installation and feeding the results back to Newcastle.
The PERQ, VAX and LSI11 drivers were supplied to Newcastle to enable them to upgrade the Connection where necessary for interworking between dissimilar hosts. This has not had the expected results! Hopes of a distribution for interworking PERQ, VAX and PDP11 over the Ring have evaporated into the Ether (to coin a phrase). So near and yet so far.
The overall SEG research theme is Quality Certification of Software Products. Within this theme, the main research interest of the Group is in Formal Specification, the development and application of machine-assisted proof systems and the role of formal proof in software development.
D A Duce and E V C Fielding hold a research grant under the Alvey Software Engineering Programme entitled Specification of the Graphical Kernel System (GKS).
GKS became an ISO standard on 15 August 1985, and has been a British Standard for some time. The document describing GKS is some 245 pages in length; the style of the description is mainly English narrative. It is difficult for a potential implementer to get to grips with the standard from this document. There is a major effort in Computer Science research to develop formal techniques for the description of system designs; the aim of this project is to apply such techniques to GKS. (D A Duce and D C Sutcliffe (CCD) were the editors of the standards document and produced the camera-ready copy from which the standard was typeset.)
Initial work has been carried out with the Vienna Development Method (VDM), and more recently comparative studies with other formal techniques have been started. This work has resulted in a number of publications.
A complementary activity, looking at the emerging 3D Graphics Standards, GKS 3D and PHIGS, has been approved by SERC's Computer Science Committee.
To get feedback on the approaches adopted and to disseminate the work done, seminars have been given at the universities of Oxford, Cambridge, Salford, Lancaster, East Anglia, Leicester and Imperial College, and at SRI and IBM Yorktown Heights in the USA.
D A Duce and E V C Fielding participated in the ISO computer graphics working group meeting in July 1985. Both are involved in the formal specification subgroup. The main work of the meeting was to draft the structure of an ISO Technical report which will look at the feasibility of using formal specification techniques in the development of graphics standards.
D A Duce is secretary of the BSI computer graphics panel.
Good typesetting facilities are essential to the smooth progress of the group's research projects, especially the Formal Specification and ERIL projects. E V C Fielding did an excellent job in developing software to couple the Unix titroff text formatter and the IBM 4250 electro-erosion printer. The help of C D Osland of CCD and H K F Yeung (DIC Group) is gratefully acknowledged.
A J J Dick's research concerns theorem-proving with equations based on a technique called the Knuth-Bendix completion algorithm. In this approach, equations are considered as rewrite-rules, and can actually be used to perform computations. There are many important applications of such work in computer science, especially in proving properties of program specifications, modelling the execution of functional programs, transforming programs into more efficient but equivalent ones, and compiling techniques.
During the last twelve months, A J J Dick has built on theoretical and practical work achieved as a PhD student at Imperial College. His main practical goal has been to develop further a rewrite-rule laboratory (mechanical theorem prover) called ERIL (Equational Reasoning: an Interactive Laboratory) to the point where it can be made available as a tool on the Alvey infrastructure machines; this goal is almost complete. At the same time, he has been able to considerably clarify much of the theoretical basis for ERIL, particularly with regard to a special form of polymorphism involving hierarchical types and overloaded operators. Experimenting with this has represented original research effort.
The following useful visits and contacts have been made:
A J J Dick holds the Atlas Fellowship which is in association with St Cross College, Oxford. This means that he is a member of the College as well as RAL and spends part of his time there.
The DCS Programme formally terminated on 6 September 1984. The occasion was marked by a major conference at the University of Sussex. D A Duce was very heavily involved in the organisation of this, including the production of two books concerned with the tutorials and presented papers.
After the conference there were the inevitable tidying up activities, the major ones being the production of the DCS Final Report, disposal of the equipment pool and rationalisation of the associated maintenance contracts.
SEG welcomed the following new members during the year:
| Sept 84 | D C Findley | Alvey Management |
| Oct 84 | A J J Dick | Atlas Fellow with St Cross College |
| Jan 85 | P M Hedlund | Research Section |
| July 85 | A D B Cox | Research Section (transferred from RAL/ID/IKBS) |
Alan Kinroy left us in December 84 to return home to his native land.
In June 1985 G P Jones left us to be promoted to the post of ID/DAO (Department Admin Officer).
The staff structure at the end of September 1985 is given in the organogram in Appendix A.
The Alvey Programme has almost completed its first phase during which strategy has been developed and research projects established and funded to implement the Strategy.
SE Management is now changing course to concentrate on the management and direction of this portfolio of research projects to guide them to success and to prepare the way for transferring new technology into industrial use.
During 1986 an Alvey-wide committee will be revising the Programme's infrastructure policy and preparing advice for their successors in the After Alvey period.
To help the above activities the SE team will be developing a new strategy document to explain what the programme is aiming to achieve during its second phase.
The current two specification projects will continue their comparison of the efficacy of specification notations and theories as part of the overall search for viable techniques to specify the forthcoming graphics standards such as GKS-3D.
SEG is considering joining the IPSE 2.5 project. IPSE 2.5 is a major Alvey SE industrial-academic collaborative project to build an advanced support environment for formal methods. SEG is proposing to concentrate on the development of the theorem-proving capability for IPSE 2.5 allied to research into new approaches to proof construction designed for advanced interactive use on large, high-resolution displays.
The Programming Research Group of Oxford University is undertaking a project, funded by Alvey VLSI/CAD, entitled the Development of Advanced System Description Language Transformation and Verification Tools. Inmos Ltd are the industrial Uncle. PRG has requested that A D B Cox of RAL/SEG should work on this project under the direction of Dr Roscoe of PRG. Authorisation of this SEG involvement is imminent.
Prof Miguel Bertran-Salvans from Barcelona University will be spending a sabbatical year (1986) with SEG. Prof Bertran has a long standing interest in the Dimensional Design idea developed by D A Duce and R W Witty of SEG. Prof Bertran will be working with SEG and DEC/IOSG of Reading in a small, collaborative project to build an interactive, graphical editor for Dimensional Design and evaluate its effectiveness in an industrial context.
The section supports the SERC/Alvey programme in IKBS and is starting its own research programme. Funding is mainly from the SERC contribution to the Alvey Directorate, and partly from the Engineering Board's ICF Applications which support the Special Interest Group, Artificial Intelligence (SIGAI). The Alvey activities are similar to those of SE Group except that the IKBS Director, D B Thomas, is partially resident at RAL in the R1 Link and, consequently, the interaction with the IKBS Directorate staff requires less frequent access to Millbank.
The Alvey IKBS programme has seen substantial changes over the period. At the beginning the research programme was still being planned and major initiatives were being taken in certain areas (e.g. Logic Programming, Declarative system architectures). Now, nearly all the funds are committed and the effort is being put into getting the projects under way, supplying equipment, monitoring etc.
The issuing of research grants is one major aspect of the Alvey IKBS programme, and it is the support of this, and the associated computing infrastructure, which has occupied most of the group effort. However there is also an embryo in-house research programme. It is hoped that during the next year, and given more stable staffing, the support work can be consolidated, and more effort can be given to research and development.
This covers that work which is driven directly by the requirements of the Alvey programme.
The IKBS research programme was divided into themes - Expert Systems, Intelligent Front Ends, Intelligent Computer Aided Instruction, Inference, Natural Language, Image Interpretation and Declarative System Architectures. In each of these, industrial and academic coordinators were appointed and charged with the formation of research communities around the topics. This is done by organising workshops, etc. W P Sharpe set up this organisation and played a very active part in its running, issuing theme newsletters, identified research requirements of Alvey demonstrators etc.
Recognising the UK lead in Logic Programming, the IKBS Directorate at the beginning of 1985 created a special initiative in Logic Programming with a series of workshops aimed at creating a single balanced programme of relevant research projects. W P Sharpe launched and it was in an advanced state when he left and R Ennals (Imperial College) took over. The programme now consists of a range of project investigating Prolog development environments, better logic programming languages etc. It relates to the IKBS architecture programme part of which is developing hardware appropriate to logic programming paradigms.
Apart from the VAX and GEC infrastructure run by DIC group the section has itself contributed to the development of the infrastructure.
The major initiative was the introduction of the SUN single user computer system in large numbers (now more 60) in the community, together with arrangements for central support etc. This has involved (P Kent, A B Smith):
The SUN responsibility is now moving to DIC group (with P Kent).
The group supports Whitechapel to some degree - 22 were purchased for assessment by the IKBS Director and the group manages the allocation and distribution of these (A J Lucas). It also handles central purchase of Orion (P Kent) and latterly ICL Series 39 for large IKBS architecture projects (C J Pavelin).
Within IKBS, SIGAI is responsible for setting the software standards and insuring their implementation infrastructure equipment. The group (C J Pavelin, A D B Cox) is responsible for running SIGAI and associated development contracts, establishing the standards, benchmarking software, etc.
Contracts are run at Edinburgh for general AI support, and for development of NIP, a portable Prolog which can be optimised for specific systems. At Sussex, work on POP LOG (a widely used Lisp/Prolog/POP11 programming environment) is supported. There is substantial work in technically monitoring these and helping to prepare new proposals (C J Pavelin, A J Lucas).
(A B Smith) The languages Cprolog, XLISP, and Quintus Prolog (SUN Only) have been ported to the VAX, SUN and Whitechapel. OPS5 has also been implemented on the VAX (because it is currently the only machine running Franz Lisp). The Prolog demonstration programs CHAT and PRESS are running on all three machines.
A D B Cox has worked on the BSI Prolog standards working group.
A J Lucas represents IKBS on the Alvey/GEC Implementation Group which monitors the development of various GEC software packages, on the technical side.
Negotiations have or are taking place with software suppliers to give favourable terms to Alvey workers - in particular POPLOG and Quintus Prolog.
SIGAI has proposed that the research community have better access to advanced IKBS development environments (e.g. ART, KEE) which exist on contemporary LISP machines (e.g. Symbolics). The hardware and software are typically very expensive, and C J Pavelin has been investigating means of providing such facilities.
The group is responsible for organising monitoring officers a requirement for all significant projects with DTI involvement. A procedure has been defined, and the first monitoring officers (NCC) are being appointed. A J Lucas has been offered as MO for one of the Vision projects.
The group is responsible for setting up research area Clubs. Now that most grants are committed, Clubs are being formed to enable projects to interact with each other, define infrastructure requirements etc. The themes will become Special Interest Groups within the Clubs. New Clubs in IKBS will be Knowledge Based Systems, Logic Based Environments and Declarative Architectures (Vision and Speech & Language already exist and are shared with MMI). The KBS Club has just been formed and the steering committee has had its first meeting.
The group runs the IKBS mailshot (sent bi-monthly to more than 450 researchers in industry and academe). An electronic Bulletin Board is planned, in order to reduce the paper mail sent out. The group also maintains an on-line version of an AI Tools Catalogue whose production was an Alvey IKBS Initiative (Editor A Bundy) and which is published as a book by Springer-Verlag.
An additional task in this area was the setting up by A B Smith of an interface to large Expert system packages (such as EMYCIN) on the Imperial Cancer Research Fund DEC10. This is available to authorised Alvey users for pump-priming purposes. It has involved system setup, and administration and implementation of a command environment on the DEC20 which allows the user to access only the AI packages.
C J Pavelin attends Alvey IKBS advisory group, Alvey infrastructure steering group and introduced and organises Alvey IKBS management meetings.
The group thus becomes involved in general Alvey tasks - e.g. providing rapporteurs at the Alvey Edinburgh conference, organising Alvey contribution to Expert Systems '85 etc.
As well as support tasks generated directly by the Alvey programme, the group has been involved in minor developments, is developing a research programme and is planning a local IKBS service.
This is a research proposal in association with the Alvey IKBS Community Finance Club (Alfex). One of the goals is the investigation of Knowledge Segmentation on the functional transformation (Expert Advisor to Novice Tutor), modifiability and maintainability of Knowledge Based Systems.
This proposal has been supported by the IKBS Director and is going ahead. It has involved acquiring and using the Savoir Expert System package for initial prototype work. Subsequent work will be based on Knowledge Systems with modifiable control structures.
ESP is a total building environment simulation package produced by the ABACUS group at the University of Strathclyde. The group is interested in making the package easier for a designer to use. D McFarlane is investigating the possibility of an Intelligent Interface to the Building Plant suite in the package. A proposal is being produced.
An Expert System has been built that builds UNIX kernels for the SUN microcomputer. It is written in the OPS5 production system language and runs under Franz Lisp on the VAX 11/750 and Sun computers.
As an introduction to UNIX and RLVD, A J Lucas helped DIC group set up a data base (using INGRES) to hold information about faults on Infrastructure computers reported through RAL.
Various inexpensive packages have been purchased for an IBM PC-AT with a view to setting up a service to increase awareness of IKBS techniques at RAL. The Alvey Expert Systems Starter Pack (cut down versions of standard commercial Expert Systems), Prolog2, Golden Common Lisp, and Savoir are all mounted.
IKBS welcomed the following new members during the year:
| Feb 85 | A J Lucas | |
| Mar 85 | C J Pavelin | |
| June 85 | D F Mac Randal | |
| Aug 85 | C Y L Kwong |
Other staff who have contributed:
W P Sharpe left in March 1985 to take up an appointment with Hewlett Packard. A D B Cox left in September 1985 to join SE Group.
The list above gives an indication of the staffing changes which the IKBS section has had to endure. At the beginning of November P Kent joins DIC group and A B Smith leaves to do a PhD at Kings College. There will then be no one who was in this area at the beginning of the year.
G A Ringland is expected to join shortly; his research proposal is covered above. J W T Smith has already begun taking over the mailshot, and has become secretary to the Knowledge Based Systems Club. S Chadha has been investigating research possibilities in IKBS but has now been awarded an HEP Fellowship and thus may stay in physics.
The staff structure at the end of September 1985 is given in the organogram in Appendix A.
The Group had regular meetings during the year and talks were given as follows:
| MBASE, MECHO | S Shadha |
| ACME | A J Lucas |
| OPS-5 | A B Smith |
| ML, LCF | A D B Cox |
| 1st Order Logic | C J Pavelin |
Most projects have now begun or are at least committed, and the programme moves into a different phase. There will be less involvement with coordination of proposals, much more with monitoring, running research clubs etc. On the more technical side, an electronic Bulletin Board is being set up to complement the mailshot, and software standards, distribution and support are just beginning to be organised. It is hoped to begin an assessment of Knowledge Based toolkits.
The PARALFEX project should be well under way shortly, and we hope to have submitted a proposal for an IFE to ESP within the next six months. This should help to build up a nucleus of expertise in Expert Systems, and their application in the design field.
We expect to assess and publicise the availability of Expert System Shells etc within the laboratory with a view to establishing if any applications at RAL could gain from these techniques.
The section is new, and consists primarily of the SUS development team. Over the year, three new staff members have joined - C A A Goswell, A J Seaton and D F Mac Randal. (Damian Mac Randal is on loan from the IKBS section). Much of the year was spent on Common Base business; probably the major task being the Single User System evaluation which involved most of the section staff and not a few outside!
Three staff managed to get through the promotion system this year - A S Williams, J C Malone and R G Dancey.
Staff list:
The overall objective of the OOP project is to evaluate the methodology as a means of software production and user interface design for the highly interactive graphics environment. An investigation of available software has been undertaken by C M Crampton. Two aspects of the project are significant - Smalltalk and C++.
Smalltalk-80 is probably the best known OOP environment, and was developed at Xerox PARC. The first implementation of Smalltalk-80 on the PERQ (Pipsqueak) was initiated by E Miranda (a Summer Vacation student), and J C Malone got the first 1982 byte codes of the virtual machine working correctly. The implementation was very slow, due to lack of optimisation and poor virtual memory management on the PERQ. Janet Malone then spent some time porting the Berkeley SUN version, which has considerable optimisation, but VMM performance prevented effective use of the implementation. (The PNX5 general release should appear soon - this appears to have a reasonably effective VMM system.)
The other project relates to the C++ programming language, an extension to C containing classes, developed at Bell Labs by Bjarne Stroustrup. We have a pre-release version; this fact is confidential under our licencing agreement. A S Williams and C M Crampton have collaborated on the implementation of C++ on the PERQ; J C Malone did the implementation on rlvd. Experience is being gained with its use, and Stroustrup is being kept informed of progress/problems.
C M Crampton has investigated the Macintosh QuickDraw and the (RAL, see below) ww(3) systems. A set of simple C++ classes has been defined and implemented.
A Window Management Workshop, held from 29 April to 1 May, was organised. Most of the Division seemed to be involved in note-taking and so on! The outcome of the Workshop was felt to be successful enough for a real book to be produced; the five co-editors were F R A Hopgood, D A Duce, E V C Fielding, K Robinson and A S Williams. D A Duce and E V C Fielding did most of the editing in their (copious?) spare time. A S Williams presented a paper at the Workshop, comparing several different window managers. One outcome of the Workshop has been the request for a standard Application Program Interface to a range of window managers; A S Williams, K Robinson and F R A Hopgood have been involved in discussions with UK manufacturers and A S Williams is preparing a draft specification.
The Window Manager research proposal, originally submitted to SERC two years ago, is now to be funded as a development by Alvey, initially for a Portable Window Manager. Discussions have also taken place between RAL (A S Williams, K Robinson) and Dr Donald Broadbent (Oxford University) on a joint proposal concerned with user interaction with expert systems. RAL would be particularly concerned with the impact of multiple window systems.
Experiences with spy (inter alia) indicated the need for a set of graphics-oriented tools, implemented by M M Martin as ww. (No-one, as usual with UNIX naming, knows the reason for the name). At the lowest level, this offers portable means for basic functions such as rasterop, mouse input and so on. Interfaces are available for Pascal (J Haswell), Fortran (C A A Goswell) and C. As a way of exercising ww, and also to provide input on the next level of the toolkit hierarchy, a series of projects using ww have been undertaken. These include a file tree wanderer and a program to help in selecting command line options in UNIX (M M Martin); a dynamic, graphical version of ps (C M Crampton); and a tool to permit window description files to be created (C A A Goswell). The system is in use outside RAL. A tutorial relating to an early version was given before the PNX User Forum in 1985. D F Mac Randal has undertaken a study of the problems involved in implementing a highly interactive post-processing system for finite element work, and an outline specification of the software has been produced.
This was a major exercise, occupying most of the staff in the section for about four months, and K Robinson even longer. Prior to the period under review, over 120 manufacturers had been contacted with an outline operational requirement. Ten suppliers of those 80 replying to that requirement were identified as being likely to be able to supply equipment of the required power on the required timescales; the suppliers so identified were Apollo, DEC, Gould-SEL, Hewlett-Packard, High Level Hardware, ICL, Racal Electronics, Ridge, SUN, and Whitechapel. These suppliers received a detailed Operational Requirement (mainly the work of C J Prosser, now with ICL, and A S Williams). The detailed replies received indicated that Apollo, SUN and Whitechapel were in a position to supply evaluation equipment.
Most aspects of the systems were investigated, either qualitatively or quantitatively. A S Williams provided expert technical advice; J Haswell ran most of the virtual memory and AIM benchmarks, as well as looking after the Apollo DN550; K G Dancey took care of the SUN2/120 and investigated some of the more arcane aspects of VM performance; A D B Cox evaluated the compiler and libraries performance; K J Fermor and P Tempest-Mitchell waded through the Comms and IPC mire; T A Watson kept the Whitechapel MG-1 going as well as she could (it was a pre-production machine) as well as preparing and running the graphics benchmarks; M M Martin ported spy, once to the SUN, once to the Apollo, and several times to the Whitechapel's different versions of the operating system and latterly, its window manager. All produced reports and so on. K Robinson tried to make sense of it all (with help!), coordinated with Edinburgh University (who were looking at the systems from an IKBS viewpoint), Technology Division (who were assessing the equipment from a CAD viewpoint), and had innumerable discussions with the suppliers and then wrote it up and presented it to Alvey and SERC representatives.
At the end of the day, no clear picture emerged: the Whitechapel was too early in its development cycle to base a service on, although the basic hardware impressed in terms of performance, apart from the graphics (which were abysmally slow and have improved markedly since); the Apollo, while fast, is very expensive and has an attitude to window management which can best be described as quaint; while the SUN, at similar cost to the PERQ2, had a graphics performance slower by a factor of about 6, but with a basic hardware/system software that impressed. Finally, the relevant committees decided that the SUN2 should be added to the Common Base. Alvey have purchased a large number of SUNs mainly for the IKBS community.
K G Dancey is secretary to the Alvey MMI Human Interface Club Committee. This has involved organisation (at short notice) of a one day Forum at King's College, London, as well as regular meetings.
K Robinson has discussed MMI section's involvement with the MMI Directorate on support in the MMI area, and has been chivying to get decisions/action.
As usual, throughout the year there have been regular meetings of the Single User System Steering Group, involving paper preparation and presentation, finance, user discussions and meetings, talking to ICL, SUN etc.
Following the SUS Evaluation already discussed, the Alvey Board asked that a survey be done of the UK capacity in this area. K Robinson, together with an outside consultant (who has learned a lot about SUSs in the past year), has drafted (yet another) operational requirement, discussed it with various (about 20) UK manufacturers, reported to Alvey, and subsequently begun discussions with various Government Departments about their SUS purchasing policy.
A number of other developments have taken place which do not fit neatly into any single heading. These are collected here.
An implementation of PostScript, a language used to describe documents in a non-device specific way, is being undertaken on the PERQ. The language looks likely to be a standard of some sort in this area, at least until ISO catches up (1990?). postprocessors for troff, TeX and Scribe already exist and are available from the supplier, Adobe Systems. At present the basic language has been implemented, and work is proceeding on the graphics. When the work is complete, the PERQ will be available for use as a preview device before sending output to the laser printer. In addition, the implementation has been made object-oriented; not only does this ease the graphics implementation, but also experimentation will be possible in the area of user interfaces and graphics.
The PADL (Part and Assembly Description Language) implementation, received from Darmstadt, has been mounted (several times) on the PERQ, with varying degrees of success. Severe problems, not yet fully diagnosed, have been encountered with one or more of the paging systems, the compilation system, and the sheer size of the software: filestore corruption has been a constant problem.
Pascal Plus has been ported to PNX, the VAX (BSD4.2) and the Whitechapel (BSD4.1), together with as much of the associated development system as possible. A tape has been sent to QUB for onward distribution to interested parties. A utility to generate makefiles reliably has also been produced.
Later versions of PNX have more than one clock, each showing different times. In the short time he has been in the Division, A J Seaton has produced a utility to allow these different clock times to be reset to a single value.
K G Dancey, K Robinson and A S Williams had above-average problems in this area over the period in question, and have coped well in the distressing circumstances. Regretfully, the problems have not been totally cured and further treatment is likely to be necessary:-)
A Mac was purchased to enable assessment of the style of working and toolkit provided, with the object of providing input for future Graphical Toolkit developments. Various bits of software, some free (Mac to Unix window package, kermit file transfer, fonts, and definitely no games) and some not (Pascal, Music Works, Filevision etc) have been obtained.
As well as a low-level activity on keeping the Apollo working and up to date, a study of the Brown University Workstation Environment has been made. Severe difficulties were experienced in building a system that worked at all (on the SUN!), and the end result has proved to be fragile.
Over the next year, work is envisaged in the following areas:
Crispin Goswell, Damian Mac Randal and Arthur Seaton all joined MMI during the period.
The staff structure at the end of September 1985 is given in the organogram in Appendix A.
The oldest project being run by the Division is the Interactive Computing Facility which provides interactive facilities for Engineering Boards' research workers via a linked set of GEC and PRIME computers spread across the country, plus, until 31 March, a DEC10 KL.
The long-term philosophy is for interactive computing for researchers to be provided by integrated design facilities consisting of single user systems connected to servers on a local area network. It is not envisaged that this will happen quickly.
The current plan is to have phased out the GEC systems by the end of the five year forward look period. The PRIME systems will continue to run throughout the forward look period but with an aim of moving existing software that currently runs under PRIMOS to a UNIX environment. To achieve this initially, PRIMIX (UNIX) is being mounted as an alternative operating system on the PRIME systems and a Pyramid system has been installed for assessment.
A large number of hardware changes have taken place on the GEC 4000 machines in the last twelve months. This activity was coordinated by M R Jane and G A Lambert.
During August 1984 a Benson plotter was installed on the Southampton 4070 workstation.
During September 1984 the Heriot-Watt and Glasgow 4070 machines were upgraded to 4180s, which are owned by the Universities, but the software is still supported by RAL. These upgrades took place with no problems.
Also during September the Bradford 4082 was closed down. The peripherals on the machine were used to enhance an already existing 4090 which was owned by the University. This machine also had a memory upgrade at the same time, which caused a number of problems, as the new memory was faulty.
At the end of September 1984 the Cranfield 4085 was closed down and the machine was brought back to RAL.
Towards the end of 1984 the London PSS Gateway was brought into service, as well as the JNT News Machine, both machines running RAL supported code.
In February and March this year further hardware changes took place. RLGB had an extra two 70MB disc drives installed plus another 112MB of memory installed. The Cardiff 4090 had two of its 70MB disc drives replaced by two 275 MB disc drives and an additional 1MB of memory installed. At the same time the Cambridge 4090 had an additional 1MB of memory installed as well as two 70MB disc drives. Unfortunately the changes at Cambridge did not go smoothly as problems occurred which were eventually traced to the hardware.
At the end of March this year the Queen Mary College 4070 was closed down and brought back to RAL. Also at the end of June this year one of our oldest machines, the Appleton 4070, was closed down.
With the lack of effort available for GEC Systems Support the development work for the systems has been greatly reduced, with the majority of the available effort going into general day to day system support. Staff involved have included S A Wood, R Poole and N Davidson.
A facility to allow the dumping system to read a command list and tell the operators what discs to mount was installed on the GECs at Rutherford.
Due to increased numbers of Hackers abusing the network, network authorisation was installed on most GECs to stop the unauthorised use of the network from these machines.
The Pad Print Server was declared ready for service early in 1985. This service will provide the queue for all output waiting to go to SERC provided PADS and printers at remote sites. This facility should allow is to start replacing some of our workstations with PADS.
The latest NAG library has been distributed to all our MUMS. We have also distributed the latest version of the Pascal and Fortran 77 compilers from GEC (bugs and all).
During the summer of 1984 the HASP function on the 4000 machines was replaced with a process called FILECARD which emulated the HASP functions. This allowed a job output from the IBM to be FTPed to the 4000 machines. Problems are still being experienced at some sites with this facility.
Special two year contracts have been arranged by G A Lambert to provide special facilities for GEC Systems. These are at:-
The last User Group and the last Managers Meeting held earlier this year decided that in future the two meetings would be combined to produce a GEC Manager/User Group Meeting.
Support activities continue at a consistent level with most enquiries being responded to within a few hours. Considerable progress has been made in updating the HELP files and the provision of additional utilities, eg KERMIT has proved especially useful.
Support has been provided by B A Alson and M J Loach.
| Site | Machine | AUs |
|---|---|---|
| Bristol University | GEC 4090 | 2482 |
| Cambridge University | GEC 4090 | 1097 |
| Cardiff University | GEC 4090 | 971 |
| RAL (RLGB) | GEC 4090 | 2358 |
| RAL (RLGK) | GEC 4090 | 837 |
| Birmingham University | GEC 4085 | 386 |
All machines have been upgraded during the year with additional memory on all systems. The 9950 at UMIST now has 6 Mbytes, all the 750s have 4 Mbytes and the 550s 2 Mbytes. Extra disk capacity was added to every system although the introduction of the 600 Mbyte fixed disk proved difficult, especially at UMIST which experienced problems extending over 6 months and required the disk unit to be replaced 3 times.
Prime 2250s were installed at Hatfield, Imperial College, Bath and Middlesex polytechnic and these are all in the process of being networked.
All this activity was arranged by M R Jane and G A Lambert.
Development activities have been restricted due to a severe shortage of trained staff. However, responsibilities were divided between RAL and UMIST and all PRIMES have been upgraded to Revision 19.3 of PRIMOS. Identification and removal of systems bugs has been a significant activity and it has been estimated that up to 80% of Systems Groups effort is taken up in this activity.
Benchmarking of PRIMIX, in association with Surrey University, has taken place and RAL will be involved during the Autumn in the final Beta testing.
Other software development work has involved additions to the Archiver software to support the 6250 BPI tape drive installed at RAL, extensions to the Mail Server to circumvent nameserver problems and additions to LPOST. Extensive modifications have also been made to the Software Distribution System such that this is now largely an automatic process.
Staff involved have included M E Claringbold, L C Peckover and R Harris.
The majority of requests for help from users continue to be handled by the local MUM managers with assistance from RAL and UMIST staff as necessary. This was provided by S Nightingale and P C Phillips.
Work continues in the provision of additional HELP facilities, ie GKS has been added, and the NAG On-line supplement has been made available at UMIST. Work has also commenced on a Network Help System.
The full Applications Manual has been distributed to all Users, with the Archive section being added during the year.
Two new user courses were held during the year, plus a special course for the new RAL staff and for new Managers. In addition Support staff have been involved in giving lectures at the SERC Summer School.
| Site | Machine | AUs |
|---|---|---|
| RAL (RLPA) | P750 | 5888 |
| RAL (RLPB) | P750 | 4791 |
| UMIST | P9950 | 15441 |
| Warwick | P550 | 3855 |
| East Anglia | P550 | 1919 |
| UCL | P550 | 1965 |
| City University | P550 | 2112 |
| Surrey | P550 | 2646 |
| Sussex | P550 | 845 |
Software support provided for Prime systems at Bath, Hatfield, Imperial College, Middlesex Poly, UMIST, Nottingham, and 7 systems at RAL.
A Pyramid 90X machine was purchased in August to provide UNIX cycles to enable ICF Applications Software to be transferred from the PRIME to a UNIX environment. The Pyramid is a novel architecture (Reduced Instruction Set Computer - RISC) machine with a unique dual port of UNIX (Berkeley 4.2 and System V). The major effort on this machine has come from I J Gunn, S T Frost, J R Aitken, P J W Randell and L Sheather). I J Gunn has performed the main role as System Manager.
The DEC10KL service run under contract by the Edinburgh Regional Computer Centre, closed down for ICF users on 31 March 1985. The service continued for the Alvey community until it closed altogether on 30 September 1985.
B G Loach and M R Jane have been closely involved with the DEC10 support contract for many years. The formal closing ceremony of the service was performed by B G Loach on 30 September.
In August the ICF Systime 8750 (VAX 11/750 equivalent) was transferred to the ABACUS design group in Strathclyde University. This group was one of the largest users of the DEC10 service at ERCC and the only way to provide sufficient resource for their needs was to allocate this Systime 8750 to them. This machine runs UNIX and is the first external ICF supported UNIX facility.
C Balderson, I J Gunn, J R Aitken and S T Frost were all involved in the transfer and installation of this machine.
The main staffing effort has come from P D Athawes and B G Loach, with N B M Calton providing programming effort in the development of the new automated procedures. R Parkes joined the section taking over as local MUM Manager and he will be taking over many of the routine Resource Management activities for the RAL Machines, releasing Pat Athawes for other activities.
The level of activity has remained constant with 80-100 applications being processed each round.
Development of the allocation of usernames and creation and distribution of Pinks is now complete and full scale testing has commenced. It is intended that in the future these functions will be undertaken by the remote local MUM Managers with RM at RAL only acting in an advisory capacity.
The last twelve months have seen a number of changes in the GEC Section. The section moved in May this year from Central Computing Division to Informatics Division. S G Birtles left the Laboratory in September 1984. B Alston and M J Loach joined the GEC section towards the end of last year to work in the User Support Technology area. K Duffey left the GEC section at the beginning of 1985 to work in Technology Division. The beginning of the year also saw the arrival of K Poole to work on the systems side. During August and September C D Rust and N Davidson joined the systems team. The only member of the Section throughout the year was S A Wood.
The year has been one of change, both for staff and the configuration of machines. With two exceptions, M E Claringbold and L E Peckover, all staff are new, to those which were in post at the beginning of the year. In addition to the two mentioned the following have been involved with the Primes systems activities, R Harris, M H Roberts, User Support, S Nightingale and P C Phillips. In addition, I J Gunn, L Sheather and P J W Randall have been involved with preparations for the introduction of a UNIX service in the future.
The staff structure at the end of September 1985 is given in the organogram in Appendix A.
Work is in progress to release a new version of the GEC Operating System 4.15, expected to be made available by mid 1986. This will be the final version and no further development work will be undertaken.
The major effort on the Prime systems will be the development and introduction of the Prime UNIX (PRIMIX) facility, expected to be available early in 1986.
Further rundown of the GEC systems is planned, Newcastle at the end of October 1985 and Birmingham at the end of the year.
R Harris, Final Report on PRIMIX Beta Test at University of Surrey, April to October 1985.
The Single User System project started as an Engineering Board initiative to provide a Common Base of hardware and software for the use of EB researchers. The Engineering Board's view was that this would be a major area of development in the future and it was important that some measure of standardisation took place to ensure that resources were not wasted in duplicating activities on a number of systems. Later the project was taken over by the Central Computing Committee as a Council-wide programme. More recently, it has been returned to the Engineering Board as a result of the Computing Review Working Party's recommendations.
The past year has seen the distribution to the user population of the full range of Common Base hardware and software - PERQ, PNX.2, Fortran 77, Pascal, GKS, NAG, and 20 sites now have X25 software. The same period has also seen the acceptance of the SUN2 range into the Common Base, and the beginning of the establishment of the SUN Support mechanism.
Many of the activities this year have been reported in the MMI report (see Section 5). This section concentrates on the support activities associated with the Common Base programme. P J Hemmings ran the Support Service, with T A Watson and M J Prime as the main software support staff. K M Lewis and D C Frith provided hardware and documentation and supplied software.
Testing of PNX4 has been completed and distribution will commence shortly. This release provides, for PERQ2s, the fast ERCC Fortran compiler, shared code and data, fork available for processes over 128KB, dual disk support, swap area bad block handling, and Newcastle Connection (over Ethernet) as well as the proprietary ICL C03 product.
The hardware upgrades for the PERQ1s, ie the addition of the 16K Writeable Control Store, have been added to 92 machines as an essential prerequisite to PNX/SR.
P J Hemmings has had a number of discussions have been held on various aspects of Customer service, mainly concerned with ordering equipment and hardware maintenance. Overall agreement has been reached to enable faults to be reported directly to the customer engineers.
The 7 January meeting of CFSC approved SUSSG's recommendation that the SUN2 be added to the Common Base. Since then negotiations have taken place with SUN involving P Kent, F R A Hopgood, K Robinson and M R Jane on a whole range of activities (central purchase, software supply, hardware and software maintenance and so on), and agreement has been reached on most of these items.
The prototype Cambridge Ring implementation by C P Wadsworth (see section 3) has been tested at Newcastle University. The Z80 upgrade programme has been completed but only slow progress has been made in producing a viable Cambridge Ring connection; in fact there is a fundamental problem due to major interference with tablet operation on PERQ1 which makes the Cambridge Ring connection unusable.
In spite of a serious shortage of staff the usual round of support activities continues. All routine enquiries are handled via the support office which has been manned by rostering members of staff from DIC Group and MMI Section for a fixed period each day. When a member of staff cannot be present queries can be handled via an answer-phone facility. GKS sets - some 90 of them - have been distributed and performance problems been tackled with the co-operation of the CCD graphics section.
User Notes on the NAG Library have been issued to all users.
A User Forum organised by P J Hemmings, attended by some 80 users, was held at RAL on 14 May. The opportunity was taken to hold a Graphics Toolkit Tutorial given by M M Martin the day before.
Contact with the SUN UK User Group has been established and a SERC SUN User Group has also been set up by P Kent. It is intended that meetings will be held 3-4 times a year.
Further meetings involving RAt and QMC have taken place when matters of common interest have been discussed. K Robinson and P J Hemmings attend these.
Resignation: P Tempest-Mitchell
The staff structure at the end of September 1985 is given in the organogram in Appendix A.
Responsibility for central support and software development on the SUN workstation is moving into the SUS section of the DIC group and as a result the section will be responsible for both the Common Base machine ranges and will be doing this on behalf of the Engineering Board's computing programme and the Alvey programme. The support arrangements will be reorganised, using the opportunity of the addition of SUN to the Common Base, the reorganisation of DIC group, and the move to R1.
Assessment of upgraded versions of machines already in the Common Base will be carried out, but a full-scale assessment on the scale of the 1984/85 exercise, would be entered into with extreme caution! This would be necessary when the market-place for high-performance workstations has altered.
Following the reorganisation of the DIC group, the section has a software development team and there are some major areas of work:
The Project initially involved the purchase of 10 GEC Series 63/30s and 5 Systime VAX 8750s for installation at sites selected by the Software Engineering and IKBS Directorates. Software development on these machines was the subject of a Memorandum of Understanding between Alvey and GEC (MOU. The GECs were intended to provide a UNIX service, once the Operating System had been mounted, and the VAXes (running BSD 4.2) were given to those sites whose function was to generate software, in order that they might have a reference machine, and also be able to import and export software. The sites were:
Those marked with * had a VAX as well. Imperial, with active departments in both SE and IKBS, were assigned two GECs as well as a VAX.
Subsequently, it was decided to add Newcastle to the set of sites. They had obtained a GEC for the ASPECT project, but are now treated as a standard Infrastructure site.
Each site had a Site Manager appointed to look after the machine and to contribute to the Programme in other ways. These Managers varied considerably in their level of expertise and their willingness to participate.
During the first few months, the remaining machines were installed and Contracts signed. However, as will be seen from the subsequent report, the delays in software production on the Series 63 have affected everything. Those with VAXes have been able to run a service on them, but some other sites have yet to get started.
The following have been involved in the project from the start:
M I Woods joined the UX63 team in August 84. A M Jackson joined the Management team in October. P J Overy and I Harding have recently joined (Communications and UNIX systems respectively). On the debit side, P J Smith left in August 84, A S Dunn in January and K J Fermor in April. All three took with them a considerable amount of expertise which was difficult to replace. S Harrod helped R A Day for a couple of months, and D Hicks has been with the Section for a year, first on UTS and later on Communications.
UTS (UNIX on the IBMs) has come to be regarded as part of the Infrastructure. The work has been done almost exclusively by H K F Yeung. There is considerable potential in providing powerful UNIX, especially to the Software Engineers and this has been borne out by experiments with Cambridge. H K F Yeung has recently been joined by N B M Calton.
Recently, the scope of the Project has been expanded to include SUN single user systems. Over 50 have been distributed to University users. Work is in hand to implement the Network File System and to connect these where appropriate to existing machines.
The first quarter saw the completion of the installation of most of the computers. Two Cifer terminals per machine were delivered to each site. Toshiba printers for the GECs were delivered to sites, although (because of the software problems) these sometimes migrated to VAXes. pragma Laser Printers have also been delivered to Cambridge, EdAI, ERCC, Sussex and RAL, following a survey by C Balderson. Printers for the remaining sites will be ordered when problems over the ability to support postscript (software provided by Adobe) have been resolved.
Wide Area communications (covered in more detail below) were provided initially by the York Box on both machine ranges. Subsequently, all the 63/30s have been upgraded to 63/40s. Provision was made for one Local Area connection to each machine. The only option on the Series 63 is an Ethernet, with software written by ERCC (LLCl). As yet, there is little that can talk to it (but VAX and SUN implementations are imminent). The VAXes have been provided with Interlan Ethernets, and most are running TCP/IP (the appearance of SUNs running BSD 4.2 has made this protocol popular).
The Management costs of the sites, including the salary of the manager, are met by SERC. This cost is based upon an initial Estimate for the coming year, followed by a reconciliation exercise at the end. As expected, the first reconciliation proved difficult. Various short cuts taken earlier (in order to speed up the agreement) led to problems (such as the assumption by some sites that they would receive the total amount of the estimate regardless of whether they had actually spent it!).
C Balderson and R E Thomas spent most of February sorting this out, and final agreement with the last site has still not occurred.
Until May, RAL was responsible for Systems support on both sets of machines. However, the difficulty in recruiting suitable staff, and other related problems, caused RAL to seek Series 63 support elsewhere. ERCC replied to the tender, and the support handed over from May, following agreement on the Contract. As a component of the transfer of support it has been necessary to move the support computer to ERCC, thus leaving RAL without a Series 63.
C Balderson, I Vollmer and A M Jackson have been involved with the general management of the equipment, with help from other members of the section as required. Work involved operation of the local GEC and VAX, preparation and distribution of Operating Systems (for both machine ranges), dealing with site queries and operating the GEC SIR database (set up by A S Dunn on one of the ICF GEC 4000 machines!). This involves the monitoring of submitted error reports (SIRs) for their suitability for submission to GEC, and updating the database after replies from GEC have been received. Operational documentation has been produced and distributed.
The state of software on the official service machines meant that it took some time to be able to declare a 'Start of Service' date. Eventually this was taken as May 1st. As part of the necessary provision of management software, work was undertaken to set up an acceptable accreditation procedure, accounting and extension of the GEC fault reporting system to cover other areas.
The Fault Reporting system is intended to cover software faults in all Infrastructure supported systems and application software. The system is known as the Fault Reporting And Monitoring Environment (FRAME) and provides for fault reports from Site Managers to be sent to a central coordinator using full-screen input forms and FTP. The coordinator distributes the faults to the relevant Support agency for the software involved, enters the fault report into a database, establishing a Status record and history for the fault.
Work on the GEC System has been undertaken by A S Dunn, R A Day and M I Woods, with S Harrod (a student) assisting for the first few months. A S Dunn left the group in January. The intention of the exercise was to assist GEC in providing UNIX on the Series 63, helping with the porting of utilities, providing accounting and distributing the software supplied by GEC to the sites. The work was covered by the MOU mentioned above.
Initially, GEC decided to mount System III UNIX. During the Autumn of 1984, GEC changed to System V, which became available in January. This had two effects: first, since porting utilities to System V was much easier, GEC no longer required RAL help in this area: second, it was necessary to undertake a major upgrade of all sites at the start of 1985.
Problems were experienced with the GEC releases (both their frequency and their inconsistency), which meant that sites saw a considerable delay between GEC announcing the release of a version and its subsequent appearance at the site. Much effort was expended in repeating jobs (such as sorting out the release tape) each month. There were also occasions when GEC informed us that they had mounted software themselves which either M I Woods or R A Day were in the process of mounting (as previously agreed). All this made the job of working on the Series 63 somewhat unhappy, and it was with considerable relief that the whole effort was passed to ERCC.
Turning to specific activities, R A Day and A S Dunn set up procedures for providing system releases to sites, including the extensive checking needed to prevent problems appearing remotely. A spooler and driver were written for the Toshiba high-quality printers, much of the work for this being done by S Harrod. R A Day spent some time investigating ways of organising the discs to suit the requirements of the users better, but this was hampered by the changing characteristics of different system releases. M I Woods began by porting system III utilities (such as "bfs" and "regcmp"). He also attempted to port the screen editor vi but stopped after GEC announced they had done it. He then worked on the journal logging utilities and on the accounting system (with S T Frost).
During the Autumn, GEC decided to hold a site audit. This involved going to each site and asking questions about problems, use etc. Almost all members of the section were involved. As mentioned earlier, the switch to System V meant a considerable amount of effort in converting the RAL software mods, and then visiting each site in the first quarter of 1985 to install the system and convert the filestore. Sussex proved impatient, so GEC installed a system for them without the RAL mods (which involved another visit later). Having installed System V, GEC's next release (the bug-fixed version) had also to be distributed, and this was completed in March.
Most of the work has been done by S T Frost, assisted by J R Aitken and K J Fermor. As well as providing accounting and journal facilities, systems were prepared for release to sites. Since the operating system was well-known, the VAXes have been heavily used, and some problems have occurred (eg, sites changing their configurations without informing RAL, so that subsequent releases fail to work). Various versions of the York software have been released, but it is still proving unreliable. Systems are therefore released on tape rather than over the network.
The version of emacs for 4.2 had some initial problems with windows and checkpointing files and these were sorted out by J R Aitken. S T Frost ported 'sees' to BSD 4.2. A tape of bug fixes for BSD 4.2 has been received from Mt Xinu. These have been implemented by S T Frost and P J Overy, and the new version distributed.
BSD 4.2 continues to be the favoured operating system because of its Virtual Memory and the software available over Franzlisp. However, BSD 4.3 is due shortly, and the latest releases of System V have Virtual Memory. S T Frost is investigating these systems to determine which should be provided for the sites. There would be some merit in moving to System V (to be compatible with the GECs) once the required functionality is present.
H K F Yeung has continued to look after the system and enhance the facilities available. In July 84, UTS (at the level of UNIX Version 7) had been mounted on the IBM 3081, but there were few users and it had to compete with CMS. During the year, various enhancements have been made, more users put on (a number of them new to UNIX) and the whole system has been moved to the Atlas 10 (where it does not compete with another interactive system). While this last move has been beneficial in general, it is difficult for UTS to obtain enough store to operate to best advantage (MVS being the major batch system).
One of the enhancements provided by H K F Yeung has been the extension of utility programs to access the 4250 and the 8700 printers. Franzlisp and CPROLOG have been mounted, and the ML/LCF system was successfully implemented (tested by A J Kinroy and A D B Cox). Recently, a user from Cambridge succeeded in running a program on UTS (with 12Mbytes virtual memory) 18 times faster than on a VAX (with 6 Mbytes real memory).
H K F Yeung has been ably assisted by D Hicks, who has worked on several projects. These included modifications to allow UTS output to be spooled to the 8700, provision of hhmail to allow UTS to use FTP to communicate with remote users (including mapping CMS ids to UTS identifiers) and an interface to the CMS Archiver to allow UTS files to be stored.
H K F Yeung has just been joined by N B M Calton. The new version of UTS (to System V standard) should arrive shortly, and work will be needed to incorporate the RAL modifications. This version provides a full-duplex facility which should allow some of the UNIX editors (such as vi) to operate successfully. Work can then begin in earnest to encourage users and to enhance the text processing facilities (at present, the 4250 is the best device available for high quality text at RAL).
This report covers those aspects of the Communications work which relate to the Infrastructure. Most of the effort has been provided by K J Fermor and J R Aitken. P J Smith worked on the Cambridge Ring. When K J Fermor left, R A Day took over Management responsibility for the section. Recently, P J Overy has joined. Work done by the section for the Common Base Programme is described elsewhere.
While the wide-area needs of the Sites were well defined, few had plans for local area networks. Subsequently, the needs of most sites are being met, with the exception of Essex, who, almost alone among Universities, run an LLC2 Ethernet. Connection to this is some way off.
Most of K J Fermor and J R Aitken's effort has gone into bug-chasing in the York software, and distributing versions to sites. Debugging is made difficult by the non-repeatability of the faults, and the lack of debug aids in the front end. Problems ranged from call failures, security breaches, and the general reluctance of the user interface to provide what is expected (according to the documentation). The security holes were found by the VAXes in the distribution; it just goes to prove how useful universities are in security testing code! General problems of X25 user interfaces concerned i/o when network utilities were in background mode, or were defined in command pipes. All problems have been reported to York via the SPR mechanism.
A new version of the code is expected in December, and it is hoped that this will be more reliable. The GEC upgrade has meant that the Infrastructure is less dependent on this code, and steps are being taken to find alternatives for the VAXes.
Other mail interfaces are also being reviewed. A copy of mmdf (a US system in general use at Imperial and UCL) has been obtained, and P J Overy is mounting it (when a complete version is supplied!).
The Cambridge Ring has not featured very highly in the list of Infrastructure needs. GEC do not provide an interface, and have withdrawn the only hardware which might have been modified to provide this function. Only Oxford, Cambridge and Newcastle showed any interest. However, the appearance of SUNs (also without a ring interface) caused Cambridge to withdraw their request. Oxford decided to request an interface for the VAX only, and the same interface was provided for a (non-Infrastructure) machine at Newcastle.
A ring driver for BSD 4.1 existed on RLVC. P J Smith started to port this to BSD 4.2, and the work was completed by K J Fermor, with assistance from SE group. The UMCZ80 and access logic unibus cards were moved from the SE VAX to the IKBS VAX. Only BBP is provided. No TSBSP implementation is envisaged (Oxford have their own software, and Newcastle merely require Datagram for the Newcastle Connection).
This has been the more popular LAN, and has required much investigation (since there was little previous experience at RAL). BSD 4.2 comes with a driver for TCP/IP, and hence this protocol (non-standard) was available on the VAXes, and later on the SUNs. GEC provide an Ethernet interface for the Series 63, but their software (written by ERCC) supports LLC1 and Class 4 (so-called connectionless protocols).
Connecting various manufacturer's machines to the Ethernet has caused a few problems, which were made more difficult by the lack of suitable monitors. The problem has been partially solved by obtaining a Spider Monitor, but the best buy, an EXCELAN Nutcracker, proved too costly. Following the connection of SUNs and VAXes, it was found that a great many collision fragments were being generated by some (several?) interfaces. The development VAX (RLVF) was found to have a bad connection and after replacing the braid picks, most collision problems disappeared. VAX to VAX connections proved a problem until it was discovered that a fault in a transceiver was putting a voltage onto the Ethernet! This was discovered after much crawling under floors. Finally, one of the SUNs decided that almost every packet was colliding. This was traced to switch settings within the SUN.
As an aid, a portable Ethernet produced by DEC (a DELNI) has been obtained, and has proved so useful that another will be ordered. The first came on immediate delivery via Technology Division (who had ordered one by mistake!).
Following K J Fermor's attendance on the ad hoc Ethernet meeting which assisted the JNT in their decision to support a connection orientated network service over Ethernet, he became an active participant on the JNT Ethernet Advisory Group investigating the proposed service.
Future plans include the mounting of the ERCC code on the VAXes, (and possibly the SUNs if someone else does not do it first), and the implementation of the JNT-recommended connection-oriented LLC2 (with X25 above it).
There are three areas of interest:
Uucp is the UNIX protocol which provides access to USENET. The USENET facility was setup on RLVD by P J Smith, connected by an ACU to Quantime, with whom we had an agreement to receive the news. J R Aitken has mounted various versions of uucp since, together with associated software. The latest version works over X25, and so is much more reliable. Uucp has also proved useful in connecting to other UNIX systems which have not had any network connection provided.
New members:
| Aug 1984 | M I Woods (from CCD) |
| Sep 1984 | D G Hicks for 1 year |
| Nov 1984 | A M Jackson (from CCD) |
| Apr 1985 | P J Overy |
| Jun 1985 | I Harding |
| Jul 1985 | N B M Calton (from ICF) |
Resignations:
| Aug 1984 | P J Smith |
| Sep 1984 | S Harrod |
| Jan 1985 | A S Dunn |
| Apr 1985 | K J Fermor |
Our plans can be considered under the following headings:
