The Annual Report covers the year ending September 1989. Its main purpose is for internal use in the Department, indicating what has been achieved and by whom. Hopefully it also gives new members of the Department some idea of the work programme in total and its objectives.
The Divisional Structure throughout the year was
It is interesting to note that this is the first year since the Department was formed that there has been the same structure and Division Heads throughout the year. This will not be the case in the year 1989/90 as the post of Head of Computational Modelling has now been filled by D R S Boyd and R W Witty will be leaving to take up a post with the European Commission at ISPRA in Italy.
The year has seen the completion of the Alvey Programme with the Department putting in a small amount of effort to wind the programme down gracefully and pay the remaining bills. As a result, the major funding source this year has been the Computing Facilities Committee of the Engineering Board and the major function has been the development of the Engineering Applications Support Environment (EASE) with its emphasis on providing Awareness to the engineering community. The Transputer Initiative continues to flourish with a well attended first Annual Conference.
The Department had hoped to play a significant role in providing the infrastructure for the post-Alvey joint SERC-DTI collaborative research programme. However, it was decided that a much leaner infrastructure would be provided to support the new programme. Consequently, the level of support is constrained to providing some effort to monitor the programme and assist DTI with managing the programme. The Department bid for research funding from the new initiative and had hoped that this would be both substantial and available in April 1989. At the end of September 1989, there was still no definitive view of what projects would be funded and when they would start. This has caused a great deal of uncertainty during the year and, as a result, it has been difficult to provide a coherent plan of activities, particularly in the second half of the year.
In Europe, the ESPRIT I Projects that the Department has been involved with are drawing to a close. To replace these, the Department was successful in being awarded four ESPRIT II contracts. These have now all started and are beginning to shape the future work programme.
This will be the last year when the annual report covers the period September to September. The advent of the Forward Job Plan for staff requires a mid-year review of staff activities at the end of June. Consequently, it has been decided that completing the report year at the end of June also will synchronise the two activities. The next Annual Report will, therefore, cover the period from September 1989 to the end of June 1990.
Last year's Annual Report gave the Department organograms at September 1987, September 1988 and December 1988.
Appendix A reproduces the December 1988 organogram, while Appendix B gives the Organogram at September 1989.
Design Division consists of three groups - Applications Integration, Systems Interface and User Interface - together with the Transputer Coordination Unit and the London and South-West Regional Support Centre Unit.
The objective of Design Division is two-fold:
Highlights of the past year, which are detailed in the sections below, are:
The Division expects that its main objectives in the coming year will be:
In addition to these broad targets, each Group and the Transputer Units have identified their own specific objectives for 1989-90. These are detailed, together with a review of 1988-89 activities in the sections that follow.
Staff in post.
The Group is divided into two Sections, one (led by JVM) is mainly concerned with data exchange, and the other (led by DT) with databases; however, DT has also been involved in data exchange.
The objectives for the Data Exchange Section over the year were to contribute to the CAD*I project, to contribute to the development of standards in engineering data exchange, and to increase the awareness of these standards in the academic community.
The main objective of the database section was to look into how databases may be used to integrate engineering applications. The initial work has used relational databases, the SQL query language and the Express data modelling language.
Specific objectives for the year were:
We were heavily involved in the CFTAG Data Exchange Workshop held in April 1989. Following initial approaches made there, it was decided to submit a bid for a Data Exchange Community Club to CFC. JVM has been responsible for discussing this proposal with interested persons in universities and elsewhere, and has written the proposal for submission to CFTAG.
A number of articles describing our work and achievements have been written, and will be published in forthcoming issues of Engineering Computing Newsletter.
At the beginning of the reporting period, the Group's activities were directed mainly to the CAD*I project. This involved the definition of a neutral file interface for finite element data for structural analysis purposes. The neutral file definitions were tested by several pilot implementations of a data transfer by means of a neutral file. Geometry data was transferred into a finite element mesh generator, and then on to a finite element analysis package. We were involved in a successful demonstration of the CAD*I project at Copenhagen in October 1988. The final report on product analysis data exchange was written during the year; this will be published by Springer- Verlag as a book in a series describing the whole project. JVM, MM and DT acted as editors, collating contributions from other European partners, in addition to making major contributions of their own.
As part of the CAD*I project we were required to participate in STEP, which is being developed under the auspices of the International Standards Organisation. After the end of our involvement in CAD*I (October 1988) we were encouraged to continue this work by CFC (following So workshop on data exchange). JVM and DT have attended meetings of relevant STEP committees. A large task has been our involvement in the critical review of the first ISO Draft Proposal for STEP. DT and JVM attended meetings of the BSI panel which is monitoring STEP, and organised a BSI panel to review the FEM (Finite Element Modelling) portions of the Draft Proposal. JVM was involved in a panel to review the STEP Physical File proposals.
A program for reading CAD*I files has been converted to enable it to read the STEP physical file. Work has begun on two pilot STEP implementations: from the STEP physical file to MSC/Nastran, and from physical file to the post processing program FAMresult.
STEP relies on the data modelling language Express. A compiler for this language is being developed, and it produced its first useful results during the year. A pretty print program for Express is under development.
Some work had previously been done in this area by McDonnell Douglas in the USA. MM and DT studied this work and found it deficient in many ways. They proposed an alternative solution and published an internal report which was then used as a basis for the project described below. Feedback from the project has resulted in changes in the original proposal and a revised report will be issued.
The mapping described in the report is based on an ideal world and does not consider actual DBMS limitations. In particular, it does not take into account the fact that most systems have a limit on the number of characters used for table and column names. KPD has designed a name generation algorithm which generates unique table and column names of restricted length from the original Express entity and attribute names. This work will be published soon as a RAL report.
It should be possible to design a set of subroutines that allow an application package to access data according to a standard data schema with the actual data storage mechanism (eg database or neutral file) being transparent to the program. We already had a set of subroutines to access data from a neutral file so PFC designed a similar set to access a relational database based on the mapping described above. He will publish an internal report describing his findings and the coding of the routines.
A simple neutral file for electromagnetics has been designed previously and is in use for simple data transfer between several research codes. It was decided to take this neutral file and formalise it into the data modelling language Express to enable wider use to be made of the technical content. This was then used to derive a database schema for more tightly integrating the codes together. KPD successfully completed this project and published the results in a RAL report.
BC has attended Project Board and Review meetings for CAD*I. He participated in the development and writing of a Brite/Euram proposal named AIDA (an environment integrating artificial intelligence, design and analysis for advanced magnetic devices). The proposal was submitted; its fate is not yet known. He has also worked on an individual merit interest - fully automatic meshing - which has made considerable progress this year. He has further developed a graphical interface package originally written by him to enable the porting of large electromagnetics finite element programs to the SUN (these were originally developed at RAL, and are now marketed by Vector Fields Limited, Oxford).
External Publications
RAL Reports
A main objective for the next year is to make all aspects of the expertise of the group available to the academic community. We intend to disseminate our results and information as widely as possible, in the fields of data exchange and engineering databases.
Effort will be devoted to participation in STEP and similar activities relating to data exchange with the aim of keeping abreast of new developments in this area, so that we can continue to inform the academic community.
A significant amount of effort will be devoted to support of the Automotive Design Project headed by Birmingham University, which involves other academic institutions and automotive industry partners.
We shall continue work on coding subroutines to read and write to databases. Using these, an actual electromagnetics package which currently uses a neutral file will be modified for use with a database. The results from this work will be fed back to STEP in order to influence the programming interface.
Staff in post during the year were:
The group's major interests are in providing tools, techniques and environments for the engineering community. The tools and techniques are directed at support for highly interactive user interface design; the environments provided are aimed at providing productive, integrated software for engineering applications development and execution. To this end, the Group undertakes a range of activities, from research (with a practical flavour), through advanced and conventional development, to support. Each activity is intended to act as a source of ideas and inputs for the others.
The X Window system is now a de facto standard for window management in the IT industry. Sun Microsystems will provide, eventually, a version (merged with NeWS, their own WMS) as a product. This project is now well overdue - at least 18 months late. As a consequence, given the number of SUN systems in the SERC community, it was decided to provide a version of the vanilla MIT X software tuned for the SUN. CMC and TAW, with input from CAAG, investigated the performance bottlenecks and applied a number of speed-ups to the MIT XI1R2 distribution, making this available on the Janet Access Machine (JAM). TAW then took the vanilla XllR3 release and applied both the RAL and Purdue fixes; this has also been made available on JAM, and the speedup information fed back to both MIT, and to the community at large via Usenet.
For the first part of the year TAW worked on providing a driver for the RAL GKS to work under X. Initially this work was done in cooperation with colleagues at the University of Manchester, but subsequently effort demands at Manchester led to TAW completing the work herself.
MMM's ww toolkit has been available, and under development, for some time. A limited amount of development has been undertaken this year by MMM mainly on behalf of internal users, and that mainly on behalf of CAAG. AVS since her arrival in July, has been familiarising herself with ww.
One of the CFC workshops suggested that an investigation be undertaken of the toolkits available on X. Work on this has just begun, with TAW and KR mainly thinking so far about things like assessment criteria. The work itself will be undertaken in cooperation with DCSD staff.
The RAL object-oriented toolkit development, mainly by CMC, was brought to a basic working state and then frozen, as the effort available to work on it was far less than that required to give a satisfactory product.
This ESPRIT II project aims to develop state of the art hardware and software for a European contender in the Stardent/Silicon Graphics market. The project is in several parts, and that with RAL involvement is to develop a board and associated software to enable quality 3D graphics interface to be developed. This part of the project is know as GRACE (GRaphics ACcelerator Engine), and RAL's role is to provide an Interaction Framework which will act as a vehicle for high-quality user interface design. There is also a work item for assessment of novel input devices.
The first part of the year was spent in revising the project to meet the reduced funding (some 38% of that requested) provided by the Commission. This effectively meant that the first half of the project was undertaken, with some concomitant impacts on the project deliverables and timescales. The situation was compounded by the Commission requiring a further 9% cut to accommodate contributions from a Spanish partner they required us to take on board!
The project began on I January, although contract signature was much later. Initial recruitment was not successful, and the effort on the project had to be found from existing staff, with MMM doing most of the work until he left, but with CMC, CAAG, TC and KR also pitching in. VAB joined the project on her arrival in July, and IW a few weeks later. Towards the end of August it was agreed that Chris Wadsworth, David Johnston, and Brian Henderson would be sub-contracted to work on the project also.
So far work has fallen into three main areas:
A specification document has been prepared as a first deliverable and provided to the Commission.
This project, funded by the Control and Instrumentation Sub-Committee via the CDTCE Management Committee, is aimed at providing a common environment for control engineers, for both standard package access and control system algorithm development. This has not been the smoothest of projects, for a variety of reasons, most of which are outside RAL's direct control. The basic environment is developed at UMIST'S Control Systems Centre under an SERC development contract, with other developments taking place in other HEIs. RAL's role is to provide management and technical support. This has meant making arrangements for beta testing, developing an interface between ECSTASY and the TSIM simulation package, and providing advice on general software and project management issues as needed. DMR and KGD have done most of the technical and detailed support work, while KR has handled most of the project management end, as well as attending the CDTCE Management Committee meetings and numerous other meetings.
During the first part of the year support was given for the first phase of beta testing, with sites at Warwick and Sussex principally involved. The results were fed back to a RAL-organised user community on 17 April. The comments made at that meeting caused somewhat of a rethink and the phase 2 beta testing was postponed to September as a consequence. On 18 April an ECSTASY course was run at RAL which was fairly successful, at least from the viewpoint of the course attendants the RAL people had a rather frantic time caused by various hardware and software installation problems, exacerbated by only limited access to the course room.
The TSIM software has been interfaced to ECSTASY, but further developments have been badly delayed, mainly due to problems caused by the unreliability of Fortran 77 under OS4, which should be resolved by the end of August.
In parallel with all this activity, Oxford University's Engineering Science's Department has been attempting to port the ECSTASY environment to VMS. This has been a long-drawn out saga, primarily because the small amount of effort available (6 man weeks) has been spread out over some 12 months, and recently compounded by the resignation of the programmer.
This is a SERC-funded project to investigate ways with which extensibility can be built-in to a graphical programming environment. Most of the detailed work is being done by CAAG, with occasional contributions from KR & TC. During the year, work has been built on the experience gained with a graphical spread-sheet and terminal emulator. A program has been written which allows the user to construct user interfaces on the fly. It is possible to layout a user interface as the interface is running, and attach actions and menus to the buttons and text areas. A simple language has been designed and implemented within this program for describing the dynamics of user interfaces, though this work is not yet complete.
The User Interface editing tool has been merged with the terminal emulator to facilitate rapid generation of front ends to existing text-based tools. Some work has been done to evaluate other user interface toolkits form the mutability and extensibility points of view.
Another experimental tool written during this year has been a data presenter, which is able to implement (currently only most of) the functionality of several existing tools, such as the MON process monitor and the FS Kent tool for file system browsing.
A seminar has been give on the subject of extensibility, which will form the basis of a future paper. Much planning has been made for future directions, including implementation of a graphical programming module to the User Interface module.
Absence of the planned object-oriented toolkit (due to lack of staff) has hindered progress this year.
This SERC-funded Intelligent Front End (IFE) project is DMR's project, and is collaborative with the University of Strathclyde. It finished in July. It addressed the difficulties that end users experience when attempting to use existing engineering appraisal packages. There were two main issues, firstly the poor quality of current user interfaces for existing engineering appraisal packages, and secondly the high level of skill/knowledge required to drive an appraisal package.
The IFE consists of a central blackboard, accessed by several knowledge bases - for dialogue handling, user modelling, and building the user conceptualization. The dialogue knowledge base uses a dynamic form-fill package, together with the other supporting packages such as a geometric modelling system, a map utility, etc, to obtain raw user input. The use of a knowledge-driven forms package enables a happy medium to be steered between application-driven question/answer input and unstructured command language input, and permits the intelligent use of inferred (ie context sensitive) defaults.
The raw user input is interpreted, checked and expanded by the other knowledge bases to produce a problem/building specification as conceptualized by the user. A further knowledge base then maps the building specification into the format/structures required by the application package and generates the commands necessary to drive the application package to produce the information required by the user.
A demonstrator system, based around Strathclyde's ESP program, is now available and the final report for SERC is in production. The complete infrastructure is in place, though due to the limited scale of the project, the knowledge bases supplied with the system are fairly rudimentary. The demonstrator is currently being distributed to more than 10 sites worldwide that have expressed an interest in seeing it. Several of these groups are interested in using the empty infrastructure as UIMS for developing specific front ends for their own packages.
The Energy Kernel System is an object-oriented programming (OOP) environment, funded by the Building Sub-Committee and CFC, for building thermodynamic analysis software. Its main objective is to enable research teams currently investigating various aspects of building simulation, especially thermodynamic performance to use software easily and safely from other groups for those parts of the overall appraisal system that are not directly concerned with their research. A second objective was to enable a collection of validated components/sub-systems to be built up, so that the robustness and integrity of appraisal software could be improved. This leads naturally to OOP and object-oriented databases (OODB). However, as the OOP approach to thermodynamic simulation is still unproven, it is important to show that the use of the EKS and its OOP approach will bring substantial benefits even in the construction of conventional simulation packages.
In July 88, CFTAG commissioned a report on extant OODBs from Leeds University. DMR was involved in this, partly as a representative of CFTAG and partly because of the interest in OODBs as a major piece of EKS infrastructure. In December, the report to CFTAG was submitted, and the following month saw the start of the EKS project. The RAL end is run by DMR. Two other groups are funded by the SERC Building Subcommittee, Strathclyde University - looking at the objects required - and Newcastle University - looking at the simulation process as a whole. RAL is funded by CFC, as part of the EASE programme, to provide the infrastructure necessary for the above groups.
After a slow start, an OODB, OB2, has been obtained and installed. A survey of existing simulation models has been completed and a suitable class hierarchy is currently being designed.
As usual there are some items which do not fit cleanly into any of the above categories. Here they are, in no particular order.
This includes the support given to the Alvey MMI Directorate, mainly by CAAG (the last MMI Mailshot before handing over to John Smith), and KGD, who took over from A Conway as the secretary to the Human Interface Club Committee. This work was essentially complete by the end of January.
KR has continued to act as the monitoring officer for the Programmable User Models Alvey grant. He has also participated in discussions on RAL'S possible role in IED-funded activities. The proposals prepared for possible funding under the joint SERC/DTI IEATP initiative attracted interest, but no funding.
So far this last year has seen the situation improve with Contracts at RAL actually generating (in May) a draft form of agreement enabling RAL to sell software it develops. The final form has not yet been seen, however and so far as KR is aware ww, ten etc have still to be sold despite numerous requests for copies.
The first copy of the licence for the RALpage PostScript interpreter has at last been sold, and many companies and individuals continue to express interest in the software.
DEC offered (originally 100%, then 75%) discount on DEC equipment as part of their EERP. As a result, two of the 12 MIPS DECStation 3100s (not to be confused with the inferior VaxStation 3100s) were delivered and are being used on the Extensibility project (qv).
KR and CAAG particularly have made some contributions to this ESPRIT II project, more detail on which can be found in User Interface Group's report.
KR attends ECFE. He also acts as the project officer for the UKC EMR on Software Tools.
CAAG is secretary to this. KR also attends.
External
Staffing during the Year September 1988 - August 1989.
Although the Group has two sections (Human Factors and User Interface Management Systems), staffing considerations have forced a degree of flexibility into peoples' work: for example MJP has spent much of the year working on UIMS and MMM spent much of 1989 on the ESPRIT-II EuroWorkStation (EWS) Project. Cooperation with other Informatics Groups and with CCD is needed in the Graphics area.
The Group has responsibility for work under the EASE Programme on UIMS Evaluation (with UMIST), PHIGS, Human Factors relating to User Interface Design. With Systems Interface Group (SysIG) we are also involved in two ESPRIT-II Projects: EWS and the ELusive Office (ELO). It is likely that work on Standards for the new programme will be carried out in conjunction with DTI/NPL. It is possible that the group will provide some input into the ARGOSI Project.
(Extensibility Dr A S Williams + TC grant application, but work is being carried out in SysIG).
This has involved MJP and TC (together with David Hicks from Applications Integration Group before his departure, and with UMIST). The evaluation of three DIMS systems - PASet, BLOX and Tiger - has been completed and the Report submitted to CFC via ECFE and CFTAG. Overall it seems that UIMS are at an early stage of development as products: it is too early to recommend one particular system to the community or to populate the EASE environment. MJ1P's UIMS survey paper has been updated and issued as a RAL report. This has been used as a basis for discussions with UIMS suppliers (commercial and research) at CHI'90 and MJP has been invited to a UIMS Panel Session at Eurographics '89 in Hamburg. A version is also being prepared for journal publication. A seminar has been organised at UMIST to be held in September 1989 to report back to the EASE Community on the results of the UIMS exercise. Further work on DIMS (and alternatives such as toolkits) available under X and making use of graphics standards (GKS and PHIGS) will be evaluated in conjunction with UMIST in the course of the next year.
Attention also needs to be paid to rapid prototyping techniques. It is hoped to hold a Technical Workshop in 1990 with supplier input, and to make a UIMS available for extended trial in a suitable HEI Engineering Department. LS has recently joined the Group and will, after a period of training and familiarization, be taking over work in this area.
PK has been working on the development of an evaluation procedure and test suite for PHIGS. The outline of this activity has been approved by ECFE and is described in Software Development Notes 57 and 60. It is hoped to actively involve engineers in the conversation of existing routines to run using PHIGS. PK has also been actively involved in the Joint CCD/Informatics Graphics activities. One expansion of activity in the standards area may come in the course of the next year with PK being involved with IED/NPL Standards activity for the new programme. There is some potential involvement with ARGOS I in relation to work on PHIGS.
Experimental Work:
This was on hold for the first part of the reporting year owing to lack of effort. In the latter part of the year this activity has involved MJP and ZM with some input from TC.
The Digitry system has been set up to allow psychometric studies to be carried out (chiefly where timing down to millisecond accuracy is required: which is not possible under UNIX on the SUNs). This runs on a PC system which now has an associated CD-ROM set up and running. ZM and MJP have carried out a study on the effects of different menu styles on selection performance. Currently this has been written up as an internal note, but should form the basis of a publishable conference paper.
Work using Swinvel-3D for DataGlove assessment and development of interactive techniques using DataG10ve have been started. (Some of this work is under the ESPRIT-II EWS Project.)
HF Awareness:
Some preliminary planning work in this area has taken place. This has included an approach to Aston University over the DTI Instructional Video Scheme and investigating other ways (such as Hypertext and PC-based tools) for presenting HF information in a usable way for engineers.
In relation to this the Course Human Factors Aspects of User Interface Design developed in part under the COMETT initiative has relevance. This has involved most of the members of SysIG and UIG together with others such as David Johnston (interactive colour editor _ Red): major effort has been from MMM, Crispin Goswell, MJP, Ken Robinson and TC. Additionally it is likely that a number of HEI engineering sites may be able to make use of HF guidelines and/or active help from UIG: especially if the extended UIMS evaluation scheme goes ahead.
These both started in January 1989 and are jointly run by SysIG and UIG. The primary role of RAL in both projects is user interface design and associated software provision together with some evaluation work.
EWS is concerned with the production of a high-end 3D graphics workstation. We are involved with the design of the GRACE graphics board and software which also is responsible for user interaction with the system. TC and MJP have provided input to the Technical Papers for this project and work is currently in hand on assessment of and techniques for the use of the VPL DataGlove.
ELO is concerned with the provision of the next generation of portable workstations. As well as providing good communications with central sites via conventional telecomms and Cellnet the system will provide an integration between the applications and an associated help and training system. RAL's major role is on the (software) Integration Framework. TC and KR have provided most of the input to the Technical papers and some initial specification and selection work. The effort on the project plan is minimal during the current year ramping up in the last quarter of 1989.
This was a result of a successful bid for funding under the CEC COMETT (COMmunity action programme for Education and Technical Training) Programme. The consortium was headed by ZGDV, Darmstadt. Only partial funding was made available for the first phase of this project. RAL's contribution was a Course on Human Factors Aspects of User Interface with innovative software for cooperative working and demonstration. Contributions to the project were made by TC, KR, CAAG, DJJ, MJP and MMM. The Course is intended to have a large hands-on component. This was an activity jointly undertaken by SysIG and UIG.
EASE start-up including assistance with questionnaire and visit design (TC). Grant bids prepared for IED and Joint Research Council Initiative (TC). Some involvement with external bodies including the DECUS AISIG and the (Alvey) Human Interface Club (TC).
(Excluding one-day meetings: covered by internal reports.)
(Excluding local Safety Courses etc.)
(Excluding ELO and EWS documents.)
Internal:
Staff in post at end of August 1989.
TPM joined the Unit in August 1988, followed by RJF in December. LJR joined in March replacing Fran Childs who returned to Space!
To carry out the Initiative programme in accordance with the Report of the CFC Working Party on Engineering Applications of Transputers, October 1986. In 1989-90, to carry the programme forward as may be modified in light of a Market Survey currently under way, and to target Continental Europe for the Awareness Programme.
The Coordination function in respect of this Initiative has a number of identifiable components reflected in the job responsibilities above. These can be regarded as customer or client oriented. There are in addition various activities which are Programme oriented including servicing of Committees and Meetings, and organising Conferences, Workshops and Seminars. A recent innovation has been in setting up and servicing a small number of Transputer Applications Community Clubs.
CB started the period as Secretary to the Transputer Applications Management Group (TAMG), and of the Directors Meeting, and was also Chairman of the Managers Meeting. CB also liaised between the DTI and the Support Centres in drawing up a Project Monitoring scheme in line with requirements for DTI partial funding of the project. CB also arranged purchase of equipment for the Loan Pool and the Transputer Support Centres from budgeted and windfall funds.
Since February this year, CB has been primarily responsible for liaison with Suppliers of Transputer hardware and software, encouraging them to assist and benefit from the facilities offered by the Initiative. Typical activities are in negotiating sponsorship or special deals to have new equipment available in Centres at the lowest possible cost; support of the Initiative Training Courses through companies sales promotion etc and encouraging companies to submit technical material to the Mailshot for publication. The present level of sponsorship is in excess of £700K.
CB is Secretary to the Transputer Applications Management Group (TAMG) and is a member of the Directors and Managers Meetings.
RJF joined the Unit in December 1988 and has taken over Liaison with the Support Centres, and more recently coordination of the Loan Pool. The first of these activities is aimed at encouraging the sharing of ideas and resources between Centres, monitoring the performance of the Centres under the terms of the Contract, and obtaining data for the DTI project monitoring system. RJF is now Chairman of the Managers Meeting and Secretary to the Directors Meeting.
The Loan Pool coordination activity (RJF) is primarily the pre- and post-loan phase wherein loan applications are collected and assessed against guidelines, and submitted to TAMG for approval or rejection, and also in the obtaining of deliverables in the form of reports for publication, and software for provision to the Software Exchange Library.
The major job greeting LJR on joining the Unit was in the cataloguing and distribution to Support Centres of equipment purchased late in the Financial Year with a special DTI funding. Since then, the Loan Pool records have been updated and developed so that more and better information is available for assessing loan requirements. This activity involves checking and testing of kit before dispatch and return, and having procedures to cope with user problems in hardware or software supplied on loan.
TPM has been responsible for putting the Initiative in the Industrial and Academic Shop-Window. The Initiative Display Stand is to be seen at Exhibitions and Conferences around the UK and elsewhere. There have been two major events demanding much effort and support from the Laboratory and the Support Centres, (a) the PC User Show in June and (b) the Initiative Conference and Exhibition at Liverpool in August. Heavy demands are made on the Laboratory Stores and Transport services in dispatching the Stand to all parts, and to everyone's credit, it remains by and large unscathed. TPM carries the onerous task of Editor of the Initiative Mailshot. This is established as a regular reliable reference to activities and products within the Transputer community and is eagerly sought after.
TPM is planning new ventures called Transputer Fairs aimed at taking awareness of Transputers into the more outlying areas of the Regional Centres.
The Centre is the Industrial contact point for London and South East England for the SERC/DTI Initiative in Engineering Applications of Transputers.
In addition, a number of members of the Department act as advisers and provide support for Centre activities. Chris Wadsworth (CPW), David Johnson (DJJ) and Cyril Balderson (CB) give lectures supporting Awareness; and technical seminars. CPY and DJJ together with Francis Yeung (HKFY) and Brian Henderson (BWH) offer technical support for Centre activities. Linda Reed (LJR) provides backup for hardware problems on Centre equipment.
PTPS joined the Centre on 31 July as an understudy for RP who has since left the Laboratory.
The principal function of the Centre is to assist with the Industrial Awareness programme run by the Initiative by offering appropriate education, pump-priming and evaluation facilities to companies in the Region. The Centre also has a mandate to offer consultancy in the design and support of Transputer-based applications.
A monthly meeting of all Departmental staff involved in the operation and use of the Centre is held under the Chairmanship of MRJ. This meeting advises the manager on matters of interest and agrees the programme of the Centre with attendant commitments.
A total well in excess of 135 (the number of names in the Visitors Book!) have been through the Centre since it was opened in April 1988. Many of these have been associated with training and awareness seminars, and others as single visitors, either as users of facilities or during visits for consultancy etc.
The course programme was established early in 1989 and to date the number of courses and total participants is as follows:
Detailed discussions about specific projects have been held with over 15 companies and six of these have generated specific follow-up activity.
More details of this work are to be found in the Parallel Processing . Group submission for this document.
The Centre has a range of hardware and software to use and demonstrate the effectiveness of Transputers. The main activity is based on 6 IBM PC-AT style machines, with various Transputer boards added. The Centre also has Meiko equipment and access to IBM 6150 and Sun Workstations able to host Transputer systems. To date, although in use fairly regularly by members of the Department and the Laboratory, take up by Industry has in fact been disappointing. It is expected that this will be successfully addressed in the near future.
The Centre would expect to benefit from new activities arising out of recommendations from a market Survey which has just been completed. The survey was on behalf of the Initiative and was seeking to identify the perceived needs of Industry from the Initiatives centres.
On the Courses front, the Centre is about to start offering courses for Fortran users of Transputer systems, and the course programme in general is now being considered.
The major problem has been the lack of a Division Head. Consequently, F R A Hopgood has attempted to act as Associate Director of Engineering, Department Head of Informatics and Computational Modelling Division Head. This, inevitably, has resulted in more of the Division's management being devolved to the Group Leaders. A Division Head, Dr D R S Boyd, has been appointed from 2 October, 1989, which should ensure a better organisationa1 structure in the next year.
The Division has continued to have difficulty matching staff with projects. The Mathematical Software Group has had particular problems attempting to complete two Alvey projects, provide the infrastructure for the EVEREST ESPRIT I project and keep the CFC applications software programme running despite the loss of staff during the period.
The Engineering Applications Group has provided the main thrust in putting genuine engineering software on the new workstation architectures. Many problems have been encountered and we have provided detailed critical information back to the suppliers. Even so, the potential of the new systems has been amply demonstrated.
The Parallel Processing Group continues to focus its activities on the transputer but with an eye to seeing how general particular approaches can be. In particular, Chris Wadsworth is trying to define a generic architecture for parallel processing akin to virtual memory in single processor systems.
The progress in the Computational Fluid Dynamics area has been disappointing. After a highly successful workshop in Abingdon, Swindon Office has been very slow in appointing a CFD Co-ordinator. This has slowed down the related activities.
The Visualisation Group continues to focus on both image processing and the output of multi-dimensional data. A major achievement has been the production of the IPAL image processing library for NAG. Research into object recognition continues in association with Surrey University. Hopefully next year will see the production of a set of tools for the rendering of complex multi-dimensional data.
The main purpose of the group is to provide a focus of computational techniques and software from which the research community can request advice and assistance in solving computational problems in mathematical physics.
The major objectives of the group for this year have been:
Coupled with these specific objectives has been the continuing process of securing new funding for research and development programming from SERC, DTI and EC.
During this year the group has been active in the planning and monitoring of the EASE Programme. CG is a member of the ECFE (Engineering Computing Facilities Executive) and has contributed to the discussion on the EASE Environment.
Over the past year some consideration has been given to the Engineering Applications Support Environment. TAPDANCE has been seen as a possible type of environment required by engineering users and a number of Divisional presentations have been given. (CG, JVA)
Over the past year the Library has only progressed slowly. Although Release 3 of the Library was completed and sent to NAG Ltd for assessment in October, detailed amendments requested by NAG have delayed the general release. The amount of effort available for the project has been very small (a small part of CG), in the new year CJC started to work on the project.
During the last 6 months CJC has familiarised himself with the Library by firstly proof-reading the current User Documentation and making suggestions for improvement, and secondly by implementing the Library on the SUN UNIX System. This has involved the production of a SUN specific Machine Constants Library and suitable Make files. As part of this exercise a number of timing trials have been comparing the SUN and Prime versions of the Library.
Attention has now turned to Release 4 of the Library and versions for the Stellar and Ardent. A number of bugs have been encountered in the Stellar Fortran compiler and now that these have been overcome work has started on using the vectorising and parallelising features of the system.
The major objectives of the project this year have been to develop and implement algorithms for the three-dimensional transient analysis of semiconductor device structures. A secondary objective has been to plan the development of the fully adaptive version of the EVEREST Suite.
During this period the EVEREST Project has entered its fourth and final year. Difficulties in recruiting suitable staff have hindered the progress of this project but CG and DG have been able to ensure that the major milestones have been met, although RAL have not been able to contribute fully to the programme of the project. As a result RAL have sub-contracted some of the programming tasks associated with the data management to staff at UMIST. CG continues to direct the software programme within the project coordinating the effort of five partners and the effort at UMIST.
DG has continued the implementation of the Solver Module of the suite and the three-dimensional transient code has been completed. With the completion of the basic on-state program, a comprehensive testing programme has been undertaken by DG and CG to ensure that the suite is both robust and produces the correct results. Some twenty benchmark problems have been defined for this purpose.
In the second half of the period JVA and RFF started to work on the project. As a result RAL is now able to contribute to the full programme of the project. JVA is now contributing to the activity on mesh adaption using the Alvey TAPDANCE Suite as a test bed, and also to the project software suite in the area of geometric modelling. RFF has concerned himself with implementing the whole suite (four large programs with three associated libraries) on the Stellar. A secondary activity for RFF has been the development of the physical models library used by the Solver Module and the Post-Processor.
The ACCORD Parallel Processing Engineering Applications Library (APPEAL) has continued development, with CG providing the project with detailed assessments of the library on design, portability and functionality.
The library has been implemented on the Prime, Vax and Stellar and a number of test programs have been prepared to perform timing comparisons.
CG has transferred the role of software co-ordinator of the serial library to another partner, so that co-ordination of the software and documentation of Release 2 of the library can be preformed by the same site.
The Alvey funded project on device simulation ended on 31 March 1989 and although a follow-on project was proposed, RAL was eventually not included. The role of RAL in the project has been to develop a two-dimensional device simulator based on the results of the EVEREST Project.
Throughout the year EMA and RJF have been developing and testing the two-dimensional device simulator DEVMOD which is one of the kernels populating the TAPDANCE Environment.
Early in the year, the off-state version of the program was made available to the partners and work then continued to complete the fully coupled on-state solver. There-has been a significant development of the physical models used within DEVMOD and the program has been tested using a sub-set of the EVEREST Benchmark set. DEVMOD is now able to solve all the problems in this set and a number of reports have been written on the results.
As part of the project the User Documentation and the Technical Manual have been produced as reports.
Although DEVMOD is not currently being developed as part of a major project, it is being successfully used as a test-bed for physical models and adaption methods within the EVEREST Project.
As with the Alvey Device Modelling Project the Process Modelling Project came to an end in March. RAL's role in the project has been to develop an integrating environment for the four other software modules being developed within the project.
This has resulted in CG, JVA and RJF producing TAPDANCE.
Over this year JVA has had the repsonsibi1ity for the day-to-day running of the software integration and has also helped the other partners integrate their kernels within TAPDANCE.
RJF has been responsible for the development of the ion implantation kernel SIMPLANT, and for developments of the geometric modeller to include simple lithography processes.
In the first half of the period ADI was a member of the group and his work associated with building performance and model assessment has continued. ADI has continued work on non-linear performance analysis and developed the methods in collaboration with Bristol University. He has also continued work with CJC on the time series analysis techniques with building model assessment.
In April ADI was transferred into the Energy Research Unit of Technology Department which was thought a more suitable environment for this type of work.
In the next year the group will continue its development of mathematical algorithms and software and provide a general advice service in these areas.
Specific objectives for the year will be:
Various meetings were attended at IEE London by Group staff.
The Parallel Processing Group (PPG) undertakes projects and development work on the systems aspects and programming techniques of parallel processing, as well as monitoring and assessing developments in the field generally. Advice and assistance is also given on porting applications to parallel systems. PPG also supports the programme of the SERC/DTI Initiative on the Engineering Applications of Transputers and of its London and South East Regional Support Centre located at RAL, both of which are coordinated in separate Units in ID.
D J Johnston and B W Henderson were successful in obtaining promotion in June 1989 to SSO and HSO, respectively.
Current interest is focussed mainly on scalably parallel architectures such as ones with distributed memory, eg the transputer and hypercube machines. A medium-term goal (12 to 24 months) is the integration of the transputer and/or other parallel systems into the EASE programme. The ease with which this can be achieved may well depend on the rate at which the community develops the ability to exploit parallel processing without the need for explicitly parallel programming. The long-term goal is to provide an effective environment for developing applications on and for parallel systems.
Much of the Group's work concerns the porting of applications to parallel systems, both by ourselves directly and through advice and assistance to others. The benefits gained serve both to widen knowledge of effective parallelisation techniques and to consolidate an understanding of user needs for an effective applications environment for parallel systems.
Porting an application is typically undertaken jointly with an applications team, eg internally or via Transputer Centre contacts, in one of two ways:
The aim in Type B projects particularly is to make PPG's expertise accessible to others (and to increase our own knowledge and expertise further!). Proposals for more Type B projects are particularly sought.
Specific objectives for 1988/89 have been as follows (in no particular order):
Logically this may be divided into support for the Transputer Centre and support/advice to the Transputer Initiative Co-ordination Team, though in practice it is sometimes difficult to distinguish the two roles (eg hot-line service to a loan-pool beneficiary who may not, strictly speaking, lie within the region served by the RAL Centre.)
The year has seen a strong build-up in most parts of the Centre's programme (booked usage by Industry of the Centre facilities is still proving a disappointment) and hence in the technical support provided by PPG.
3.4.3.1.1 Courses
PPG regularly contributes to two courses offered by the Centre: a half-day seminar Transputer Awareness for Managers (CPW - 90 mins) and a one-day course Exploiting the Transputer (all PPG staff). These courses now run regularly at monthly and two-monthly intervals respectively. Each is normally concluded with a visit to the RAL Centre and demonstrations by PPG staff (except when held off-site). Number of times held during the year were:
| Monthly Awareness | 6 |
| Special Awareness | 2 |
| Exploiting Transputer | 3 |
In addition, a special two-day course was designed for a joint audience from Cambridge Instruments and the EBL group at RAL.
3.4.3.1.2 Discussions/Consultancy
CPW and/or DJJ have frequently participated in discussions with prospective industrial customers of the Centre. HKFY and BWH have also been involved in several of these meetings. Usually a member of the Co-ordination Team, or the Centre Manager, is also involved, with the technical lead corning from PPG (and the customer!). Among the companies with whom discussions have taken place are:
(The MOD pair did not say (were not allowed to ?) which office/establishment they were from! The asterisks above indicate those companies with whom there has been specific follow-up to date, either in the form of consultancy/ sponsorship, work done for or with the company, or development of a special or on-site course.
3.4.3.1.3 User Support
DJJ has provided nearly all the day-to-day support to users of the Centre: getting users started; hand-holding; tutorial assistance; hot-line service for telephone/email/postal queries from users.
This area of support has seen a sizeable increase in activity recently, particularly since the exhibit at the PC User Show in May 1989. With changing priorities for PPG (from pressure on CFC booking), it may become necessary to cut back in this area, eg users outside the South East Region may be referred to their own Centre.
CPW is frequently asked by the Co-ordinator for advice on technical matters pertaining to the Initiative, in particular: advice to loan and grant applicants; appraisal of loan applications; comments on miscellaneous papers or proposals received; technical negotiation of EMR developments contracts. CPW also provided technical liaison to the QUB/Brunel/Liverpool EMR survey of language and compiler facilities for parallelism (this has close connections to a collaborative project PPG hopes to be involved in - see 4.7 below). CPW is a member of the Transputer Applications Management Group (TAMG) which oversees the operation of the SERC/DTI Initiative.
DJJ has been involved with the evaluation of all new hardware and software products being considered by the Initiative, both new products (eg 3L Parallel Fortran, Motherboards, TRAM Modules) and new releases (eg product releases of TDS and TDT), and of initial versions of software generated by EMRs (eg port of GRAIL to a PC).
PPG staff have provided effort to man the Transputer Initiative stand and show demonstrations at the BAAS Exhibition, Oxford, September 1988 (CPW, DJJ, and Rajka Popovic) and at the PC User Show, May 1989 (DJJ).
CPW chaired the Applications Workshop held by the Transputer Initiative at Cosener's House, Abingdon, on 14-15 February 1989. As with the previous Workshop, this took the form of invited presentations followed by parallel sessions (twice) interleaved with reports back to a general session. The three parallel sub-groups were: Control; Modelling; and Inputs, Data and Output.
DJJ wrote most of the Workshop Report with inputs from CPW, MHR, MRJ, Terry Mawby, and Raymond Fawcett who acted as notetakers in the parallel sessions. The major recommendation (now implemented) was the establishment of three Transputer Applications Community Clubs (TACCs) in Image Processing and Pattern Recognition, Molecular Modelling, and Real Time Control and Simulation.
With the popular appeal and attention focussed on the transputer, demonstrations have become a seemingly ever-growing part of PPG activities for a variety of arranged and casual visitors. At times it has seemed that hardly a day has gone by without one!
Among those to whom presentations have been given, in addition to the courses and companies mentioned above, are:
The range of demonstrations available has been steadily expanded and consolidated during the year. The standard set that has been regularly shown now consists of:
All of these run on the Meiko Computing Surface, except for BIM2D which is hosted by a 6150 (developed under IBM Joint Study - see 4.6 below) .
Demo (2) was written by DJJ in response to an observation (by CPW) that the standard Meiko demo was unimpressively slow - DJJ's version, now optimised, achieves 12 frames/sec against Meiko's 3 frames/sec! Programming effort was about 2 man-weeks; this illustrates a common experience with developing software for transputers that a relatively small effort can produce considerable performance benefits once relevant optimisation principles are understood. In programming the demo, DJJ also developed a geometric harness, a communication structure for a transputer network reflecting the geometry of a generic type of problem exemplified by wave propagation. This and the other demos on the Meiko system also illustrate the benefits of electronic reconfiguration of the transputer network in the Meiko system (the demos all run on a different topology).
Other demos available are:
Demos (a) and (b) were developed by BWH; (a) shows a visual comparison of the performance of a 6150 alone (with FPA) versus that of a 6150 plus 4 transputers (see 4.6). Demo (c) simulates an explosion in a pressure vessel, eg a nuclear rector, and was developed during a Transputer Initiative loan to Swansea. Demos (d), (e) and (f) were developed by DJJ; a special feature of (f) is that the world rotates backwards!
As part of the ESPRIT Accord project Vector Fields approached PPG to undertake a port of the host server for 3L Parallel Fortran from a PC to a Research Machines 386i running Unix. An extension to the Department's Accord sub-contract from Vector Fields was agreed in June and the work is now nearing completion. The work is based on the similar port for the 6150 host and has been done by BWH.
BWH developed a simple monitor for a transputer network which provides statistics on the run-time performance of each transputer in the network. This is interfaced to a graphics display program to plot the data and enhance the debugging and optimisation capabilities of the monitor. The present version of the graphics plotter allows plots of up to 32 transputers to be viewed concurrently.
DJJ wrote a worm program that explores a transputer network to discover the number of processors present and how their links are connected. In contrast to most such software, including the standard worm in Inmos' software distributions, this worm explores many branches of the network in parallel. This enhancement becomes particularly significant when the network is large, for which there are analogous considerations in the boot software for loading a multi-transputer application onto a large network.
Despite Inmos' best efforts to convince the world that Occam was the assembler language of the transputer, there is a more traditional assembler and much performance advantage can be gained by programming critical routines selectively in assembler. DJJ has investigated the relative efficiency of programming the transputer in assembler, compared both to Occam and to traditional sequential languages (Fortran, C). The results have been documented in PPG Note 12 for an example set of graphics primitives. An abridged version is to appear in the Transputer Initiative Mailshot.
Following discussion with Wootton Jeffreys, DJJ has ported one of their traffic modelling applications called ASPECT first on to a single transputer and then on to a multi-transputer network. ASPECT is a near ideal candidate for parallelisation, involving independent calculation of shortest paths for all possible pairs of starting and finishing sites in a traffic network (a bit of cleverness/thought may of course reveal a much "better" parallel algorithm!). DJJ investigated the possible choices for porting the original Fortran code to multiple transputers in order to determine the tradeoffs involved.
This work, along with that on the worm and on assembler, have been written up as PPG Notes. All three contain work worthy of outside publication and routes for this are being pursued.
DJJ has ported the spy editor developed at RAL to the Atari Transputer Workstation (ATW).
DTI have recently launched a scheme to place six (initially) different transputer software products for use at about twenty sites each. DTI chose to call it Parallel Software Evaluation. The scheme is operated by the University of Southampton.
PPG bid for and have now received Helios (PC-version) and the 3L Parallel Languages (Fortran, C, Pascal) for evaluation. The evaluation will require a report to be produced by the end of the first year (Summer 1990). Others wishing to try out the products and contribute to the evaluation which PPG will be doing are welcome.
There is by now a considerable variety of approaches, and substantial differences of detail within approaches, to programming transputer systems. The variety becomes even greater when one looks at other parallel systems, eg hypercubes or various shared memory multiprocessors.
There is a growing realisation, at least within one half of the parallel processing community, that this state of affairs cannot continue if general parallel processing is to deliver its promised benefits. (For a suitable consideration CPW might be persuaded to disclose which half of the community is starting to think right!).
CPW has spent much time cogitating these matters, with several lengthy and helpful discussions with Stuart Robinson particularly. Initial conclusions were presented in an invited paper at the Unicorn Seminar on Software for Parallel Computers.
In mid-April CPW was contacted by Space Science Department with an appeal for urgent help with programming in C on one of their projects. IDHM was immediately keen for ID to be seen to be helping others in RAL. After a rapid but thorough survey by IDHM of who could do the job, involving it seems all in PPG as well as others, BWH was selected as the lucky one to bail out Space Science (sorry: to make Informatics expertise available to other Departments!).
BWH was on loan to Space Science for two months (May and June), with occasional days since. Stuart Robinson has been the Informatics link for the loan and a source of expert help about C.
The work consisted of developing a system to allow testing of a space-borne instrument to be carried out in an automatic way. A 386 PC was used to control the test bed and the instrument. A test language was developed to allow long and laborious tests to be carried out with results logged and errors flagged.
A side-benefit of the loan was to provide a further link with the transputer activities in Space Science and at the Mullard Space Science Laboratory (MSSL) with whom RAL collaborates on the Solar A Project. As a result a Special Awareness Seminar (half-day) was given at MSSL (Dorking).
This Joint Study with IBM involves the IBM 6150 workstation and the transputer. IBM have provided RAL with three 6150s on loan for the duration of the project (two years). The Joint Study started officially on 1 November 1988 with the completion of delivery of the loan machines. The general aim is to explore the potential for enhancing the capabilities of an advanced graphics workstation (6150) with a reconfigurable parallel computer server (transputers).
With a jump start on the official start date, initial work has been completed well ahead of the target dates. The software to enable the 6150 to serve as a front-end to a transputer system has been in a stable state since the end of 1988. This consists of an AIX device driver and the afserver program which serves the transputer network with file-handling and screen and keyboard access via the 6150. The port of afserver to the 6150 has also been extended to provide IPC facilities between the server process and other processes running on the 6150. Later, BWH has revised the server program to use the AIX device driver for accessing the (first) transputer board.
Porting a first application, an electromagnetics package BIM2D, was successfully completed by MHR in collaboration with Vector Fields. BIM2D is a 40000-line Fortran program and the port was achieved with only 700 lines needing to be ported (and parallelised) to the attached transputer system. The results of this work are being presented at the Compumag Conference in Japan in September 1989. A demonstration of an interim version was given by MHR and Vector Fields as part of the latter's contribution to a presentation by the ESPRIT Accord project. Later, BWH has ported a PHIGS implementation of the GINO simulator from Vector Fields onto the 6150, and extended the BIM2D program to allow it to use the new simulator.
With the loss of Mark Roberts at the end of February 1989 and the loan of BWH to the Space Department and Vector Fields, the project has been for a time moving in a single track rather than spearheading in parallel. Though ideally one would like to advance both the system and application aspects of the project in tandem, progress has been made on both fronts throughout the year, albeit one at a time.
On the system side, one of the major developments is the implementation of the LAN based transputer server. This allows transputer systems to be accessed remotely and would be useful to some of the workstations here which are not expandable. The transputer server is effectively a compute server in the conventional sense which happens to be using the transputers for its operations. In practice, the server runs as a daemon process on the host machine which communicates with the clients using the TCP/IP protocol. However, by using the socket utilities, the implementation is not much exposed to the underlying protocol. The software is further protected from the various schemes, data is stored on different machines by adopting the XDR format in exchanging data between the server and the clients.
On the application side, a number of packages have been investigated which includes TOSCA, SPICE (a circuit simulator package), SCNAPT (an electronic circuit analysis package), APPLE (a finite element package), and ONDA (a river modelling package). With the effort presently available, it is unlikely that more than one application from the list would be embarked upon, and this makes the decision even harder to arrive at. In general one would prefer a real application to a toy program and ONDA scores handsomely on this count as it is currently used by both the National River and Thames Water Authorities. If realism is the main criteria, ONDA will probably be chosen in due course and the project will be carried out in collaboration with Sir William Halcrow & Partners. ONDA has already been ported successfully to the 6150 above.
BWH installed the new operating system for the 6150. The new release, AIX 2.2.1, included NFS and X-Windows (Version 11) which had been holding up the work on the project.
BWH developed a demo for the IBM 6150 that illustrates the speed difference between a 6l50+FPA and a 6150+4 Transputers. The demo consists of calculating the field around a set of electrical potentials in an area of 200×200 points. The demo uses X-windows to allow two windows, one from the IBM 6l50+FPA and the other from the IBM+4 Transputers, to be watched at the same time.
BWH tried to port the circuit design package SPICE onto the 6150. It was envisaged that this program could be altered to use Transputers to speed up its processing time. The available source code of this program was a few releases old. Unfortunately the program would have required a massive amount of re-writing to enable it to run. The program was thus abandoned for the time being.
BWH spent time porting an X11 version of the WW windowing library onto the 6150. This was successful. An attempt to port the graphical editor program ten (which runs on WW) to the 6150, was then made. It was found that this was 90% effective. A problem in the byte-ordering of the directory structure on the IBM has prevented this task from being completed.
PPG joined three consortia in collaborative IED/IEATP proposals whose progress has spanned the full year of this annual report. One of these, as expected, failed to get past the outline proposal stage. A second one, on graphical monitoring tools for transputer systems, was invited to submit a full proposal but the lead partner (GEC Marconi) took a dislike to the LINK funding rules which were introduced at that stage. An attempt to resurrect the project in a proposal to DARPA in the US, with much more generous funding rules, was made but came to nothing.
The third proposal entitled Fortran for Scalably Parallel Systems with Intercept Systems, Meiko, and Brunel University was technically approved in February 1989 (ranked third out of 12 approved from a total of 108 outline proposals originally in Parallel and Novel Architectures). The project will aim to develop the techniques needed to detect and extract parallelism from Fortran programs for multi-transputer target systems. As yet, however, the expected formal offer/announcement to the project consortium has not been issued from IED. Estimates of when this might happen are deferred to next year's report!
Discussions have also been held with IBM UK Scientific Centre, Winchester, on common interests and possible joint activities, but are currently dormant following the transfer of the IBM leader, Tom Heywood, from Winchester to Hursley.
DJJ designed and implemented a graphical colour editor called red on SUN workstations for the Human Factors Group in Design Division.
CPW has participated in two weekend meetings at Cosener's House as part of a project that has been running for several years to produce a three-volume Handbook of Logic in Computer Science, edited by Profs Gabbey, Abramsky & Maibaum of Imperial College. Each volume will have 10 to 12 chapters of about 50-80 pages. Volume 1 is nearing completion and is now expected to be published in 1990. CPW is principally involved with two chapters, on Denotational Semantics and Domains, in Volume 2 in the curiously named role of Second Reader (first reader might be more accurate). The main connection in this activity with parallel processing is that it is being done in parallel with everything else! (Other chapters in the Handbook will cover work on concurrency a la Milner, Hoare, Winskel, Plotkin, etc.)
None during this period.
The general thrust of the Group's work on parallel processing as outlined in 4.2 will continue with an increasing emphasis on addressing the problems associated with providing machine independent parallel software. This is seen as a key step towards the integration of parallel systems into EASE (at least for a range of parallel architectures wider than one machine).
Specific objectives for 1989/90 are as follows:
Previous plans for future work in parallel processing have been thrown into some confusion by the current difficulties in booking to CFC/EASE. As a result, three of the Group (DJJ, BWH and CPW in varying proportions) are now providing additional effort for the ESPRIT European Workstation project run by Ken Robinson in Design Division.
Further objectives in parallel processing are therefore tempered by the fact that at the time of writing there is no effort spare to address these laudable aims, which include:
Some of (a) to (d) may become feasible should the IED proposal (1) eventually fall through.
The group consists of Alan Bryden (ADB) and Manjit Boparai (MKB). MKB joined the Laboratory on 16 January 1989.
The main objective is to provide the computing infrastructure for the Engineering Board CFD community. This breaks down into two sub-objectives in the current year:
The CFD Programme has its basis in an Advisory Group report to the Electromechanical Engineering Committee in May 1988. Following discussions between Swindon Office and RAL, ADB organised a Workshop in Abingdon in January 1989 attended by leading practitioners in CFD from industry and academia. The recommendations of the Workshop were presented first to EMEC, which agreed to fund a Co-ordinator in CFD, and then to CFC which agreed to allocate 3MY of the effort within EASE to CFD.
Since then ADB has been following through some of the recommendations of the Workshop. He is currently involved in discussions with CFD users prior to setting up a Community Club later this year, and in determining the problems it will address. He is also discussing with Swindon Office details of the proposed Link Programme in CFD.
MKB is undergoing a period of training in use of numerical software at RAL. She transferred the FE Library to the Stellar computer and is currently extending one of the Library modules from 2D to 3D and using it to solve a benchmark problem from the Team Workshops, viz a coupled problem to find the vibrational modes of a bar oscillating in a time-dependent electromagnetic field.
ADB also spent a week in the USA from 6-10 November, for discussions at NASA Ames Research Center, Moffat Field, California. He also visited Stellar Computers, the Concurrent Computational Group at Caltech, Pasadena, and San Diego Computer Center.
The timetable for the future programme will be influenced by the date of appointment of a CFD co-ordinator by EMEC, the setting up of a LINK programme, and the appointment of a new member of staff experienced in CFD.
The Community Club will determine an initial set of test problems across a range of applications in CFD. This is already being done by ERCOFTAC (the European Research Community on Flow Turbulence and Combustion) in some areas, mainly in problems associated with aeronautics. Within the next year a version of the NASA Ames data on Turbulent flow will be mounted on SERC computers and a few guinea pig users will start to use the system. An initial design of a CFD software system will be carried out in collaboration with the Mathematical Software Group. A start will be made in integrating this activity into the overall software environment and in applying it to the solution of test problems.
Throughout emphasis will be directed towards use of graphics supercomputers like Stellar and Ardent Titan and parallel processing systems.
The need for visualisation arises from the increasing complexity of data and models which computers manipulate, to an extent that printouts or traditional plots are unable to convey the information. Often the data is in the form of images and these too need to be manipulated and displayed in order to be understood. In the opposite direction, the higher-level information needs to be extracted from images usually derived from an external source. The work of the group is responsible for research/ assessment/ development/ integration/ enhancement/ publicity in this area.
During the year, the staff in the group have been
During the year, Martin Carter and Rajka Popovic gained promotion to a HSO. The Group also had the services of trainee, Sue Davidson, from October 1988 to April 1989.
The group was formed with a number of projects that were due to come to an end during the period of this report. Furthermore, it was also formed with a brief that was not previously explicit in the department - to be responsible for the visualisation of complex data and models. The main objectives therefore were to determine
RM has continued his role of coordinating the RAL effort with NAG. However this effort has dropped during the year and is now simply RM plus a small fraction of MKC. Technical meetings have continued and are written up as PROFS documents with key-words IPAL MEETING. At the current level of effort the meetings have become less formal.
Part of the work that the group will do for EASE is to integrate the IPAL library into the environment. Planning has started and some commercial examples of image processing environments are being studied.
Sue Davidson wrote interfaces to GKS for both the Fortran and C libraries. This work was done as a six month trainee project under the supervision of RM. For the purpose of demonstrating the code, RAL-GKS was used for the Fortran library and SunGKS for the C library.
The library was demonstrated at ITEX at the Barbican in November 1988. The demonstration was written and set-up by RM but run by NAG.
A limited trial release of the Fortran library started in January 1989. Prior to this the library was subjected to testing at NAG by Mike Hooper. This trial release code has been mounted and tested at RAL by RM but is of limited interest since there are major gaps. NAG have issued the software to a very selected set of trial sites and will issue it to a further set soon.
Work to fill the gaps in the library has continued. Specifically E Golton, from Space Science Department, has supplied a comprehensive set of transform routines and Robin Oldfield (NAG) has added routines for texture. This additional material is not in the trial release.
RM has concentrated his efforts onto the C library. Agreed changes to the structure definitions have been implemented. The major effort has been to extend the library to handle float data-types for pixels.
Effort is going into the production of initial documentation for this library.
Over the last eighteen months we have been involved in a collaboration between RGO and the space science department at RAL developing an image photon counting system (IPCS) for both ground based and space applications. Our main contribution to the project has been in the design and development of a system to perform the real-time processing of data from the image intensifier. This has involved specifying the computation required, coding algorithms and testing their performance on a transputer based system and producing analysis software to study the behaviour of the intensifier. The electronic hardware for the first prototype of the system has been built and we are about to begin testing of the individual components. As part of the study for the next stage of the system we have looked at the real-time implementation of connected component analysis using a transputer based system and the theoretical implications of the sampling theorem for separation of overlapping events.
Other projects worked upon include a study of image compression for use in space applications and the use of a Vax based transputer system running parallel Fortran.
Research on an Alvey funded project concerned with the identification of objects from 2-D scenes continued this year. This project is a collaborative venture involving British Aerospace, Bristol University, Marconi, Reading University, RSRE, STL and Surrey University.
Work at RAL has centred on various aspects of the use of contextual information in pattern recognition and in particular the study of relaxation processes. The collaboration is now preparing its final report.
This year research has concentrated on a new methodology for discrete relaxation. The aim has been to develop better methods for the consistent labelling of arrangements of image entities. The result has been a labelling procedure that can be realised by the iterative updating of symbolic assignments. This method has several attractive conceptual and computational features. In particular, it demonstrates how consistent labelling problems can draw on Waltz's dictionary representation of constraints through a congruency concept. This is a novel approach which enhances and extends the usefulness of the dictionary idea. It means that discrete relaxation algorithms can be applied to highly structured labelling situations without degrading their representational capacity.
A full account of the method has been prepared and submitted for publication in the Pattern Recognition Journal. Several shorter papers on the conceptual basis of the method, its applications and its control have been accepted for publication in refereed conferences.
In order to draw together some of the work on the application of relaxation processes, a comparative study of different edge postprocessing studies has been undertaken. The conclusion of this work is that the objective quantification of contextual information using the relaxation framework has some important performance advantages over more heuristic approaches. In particular relaxation is superior to the popular edge postprocessing methods of Canny and Spacek. A paper describing this work is to be published.
A high point of the year has been the acceptance of a paper by the highly selective journal IEEE Transactions on Pattern Analysis and Machine Intelligence.
An IED proposal to extend the work on relaxation within the context of artificial neural nets has now been 90 percent approved. The partners in the project are RAL, RSRE, Surrey University, and York University, with BAe and another as yet undetermined partner acting as exploitive uncles.
The project is concerned with developing methodology for subsymbolic processing of uncertain image data. Research at RAL will concentrate on extending relaxation methods beyond horizontal information processing through the use of hierarchical constraints and vertical evidence combining procedures.
Some work towards the assessment of superworkstations has been undertaken. Some applications are being mounted by others in the division. The group is responsible for coordinating the assessment and for the overall report.
As part of this, JH is responsible for system management and temporarily acts as system administrator for the Stellar, initially model GSIOOO and now the GS2000, and Ardent's Titan superworkstation. This involves setting up the systems to fit transparently into the Informatics network, installing and checking out new software, setting up accounts and assisting users. JH has also administered a colour thermal printer attached to the Stellar plus a few other Informatics colour machines. JH will shortly pass day to day administration of these systems over to the operations section along with details of maintenance though she will continue to offer technical support.
The group is installing and gaining expertise in a number of image generation systems - UNIRAS, AVS, Dore (including portable Dore pronounced Doray), Renderman (not obtained at present) and PHIGS+. Technical support can be offered to users within the department and (probably after some package selections) to external users.
UNIRAS has been installed on the colour Sun 4/260 in Lab 11 (JH). Since the set of manuals is very large, a simple users' guide will be needed, either written here (JH) or elsewhere if suitable (Central Computing Department or other UNIRAS users).
Dore is installed on the Titan 1. RP is providing technical expertise for Dore, having been on Ardent's Dore course (with JH). Portable Dore has been installed on the colour Sun 4/260 (JH) and a Portable Dore launch was hosted at RAL on behalf of Ardent on 20 July (JRG, RH, JH and several members of Distributed Computing Systems Division (RT, ACD, IV).
RP is responsible for technical support of PHIGS+. Version 1.5 of PHIGS+ graphics library arrived with the Stellar GS1000. After RP had done the first test, numerous errors were found and reported back to Stellar. The new 1.6 version was installed at the beginning of 1989.
Since then, RP has provided continuous technical expertise for Stellar users, and has written an internal report Introduction to Stellar PHIGS+ Library.
Some work on comparing these systems has been done. JH is comparing AVS (Application Visualization System) from Stellar and Dore (Dynamic Object Rendering Environment) from Ardent. AVS provides both an interactive system and a toolkit for programmers to handle graphics input and output, whereas Dore is a graphics toolkit to handle graphics output using simple rendering methods such as wire frame up to ray traced images. A report comparing PHIGS+ and Dore libraries is due to be issued (RP).
As a part of this project criteria for selecting the package were established. RP attended the CAD/CAM exhibition at NEC Birmingham in March 1989.
Although the transputers are not the brand-new hardware, there is no proper and complete catalogue of available graphics software. RP had the task of collecting and systemising data for such a catalogue. The aim is to produce the first version and make it available through the National Transputer Centre, and via the computer network. The catalogue will be regularly updated. At this point, collecting is in progress and the initial data is put together and will be sent to companies for approval.
The first phase of the project is completed (ie a simple tree can be loaded, processed and rendered on the screen). The paper is written and will be presented at the Parallel Computing 89 conference at Leiden in Holland, September 1989.
JRG has participated in 3 ISO meetings on PHIGS PLUS. The work is derived from the de facto document PHIGS+ which was produced by a number of USA suppliers. The ISO group itself includes USA suppliers, Sun, HP/Apollo. Some major concerns are to ensure that PHIGS PLUS is an adequate basis for interactive display of complex data and of product data.
JRG is also the document editor for the ISO Fortran binding of GKS-3D. The DIS (Draft International Standard) was produced in late 1988 and the meeting to resolve the votes and comments was held in June 1989. The final task is to produce the international standard based on the decisions of that meeting.
The group participates in ARGOSI, which contractually started on 7 March, but technically on 1 April this year. JRG participated in most of the meetings which established the project and now leads the Classification Workpackage. Since the project started, RM has also taken part. (RAD from DCSD and DAD from SED are also heavily involved, with RAD being the RAL representative on the management board.)
ERR continued to act as secretary until April 1989. The Club held its final meeting at Worcester College, Oxford in December. This meeting had a twofold purpose. Firstly, it provided a forum for a retrospective review of the Alvey speech and natural language programme. Its second function was prospective, providing the UK speech and language research communities the chance to express their views concerning the new SERC-DTI joint framework to its administrators and to discuss the possibility of a new Club within this framework.
A new Club (the UK SALT Club) has been established. ERR acted as interim secretary and was involved in the initial organisation of the Club's inaugural meeting at Keele University in April. The organisational responsibilities have now been handed over to the DTI.
ERR has joined the academic staff in the Electrical Engineering Department at Surrey University as an Associate Lecturer. This involves teaching various aspects of electrical engineering to first and second year undergraduates.
JH has continued to advise and assist people within the Laboratory in the use of the Mac+ and Mac II and purchase of new software.
JH completed the examinable section of a part-time M.Sc. course in December having passed all the exams gives her the temporary title of Post Graduate Diploma. A dissertation remains to be submitted, of which the original topic was to be a visualisation toolkit using PHIGS+ and X11. However this will be altered and delayed since Stellar have themselves implemented a system, in the form of AVS.
For EASE, a number of objectives are being defined:
To provide the experience and basis for this work and as a service in itself, further collaborative projects within CMD will be defined.
The group will also be able to provide expertise in graphics and image processing on transputers.
We hope that the SERC/DTI JFIT project Vision by Associative Reasoning will finally secure its funding and be able to start. The ESPRIT project ARGOSI will complete the Classification Work Package, but the group will be participating in other work packages.
Other funding and clients will be sought to complement the work being done for the EASE programme.
Systems Engineering Division consists of two groups, Knowledge Engineering and Software Engineering. Both groups are small in RAL terms. Each has a dual role of direct participation in the Department's support of the Engineering Board and a research programme which generates knowledge, skill and technology to be fed, indirectly, into EB support. Whilst the two groups' foci are distinct, they are united by the common threads of representation and reasoning.
This year has seen both groups working hard to obtain funding for their research programmes from external sponsors such as ESPRIT and IEATP. Both groups have also enjoyed stable staffing and successful recruitment exercises. With the appointment, this year, of a new group leader for SE, the Division has achieved a stable management structure.
KE group has continued its expansion through successful recruitment exercises and research project bids. Much hard work this year will be rewarded next year when two excellent new staff members will join the group's projects sponsored by ESPRIT, IEATP and direct industrial sponsorship.
This year has seen the termination of one of KE's most successful projects, Paralfex. This Alvey funded work had two main themes, graphical explanation and knowledge based architectures for the financial domain. Paralfex's Source of Finance Adviser was ported to the Compaq 386 and interfaced to Lotus 1-2-3 by a one year Sandwich Student.
After much hard work KE Group was awarded, with 6 partners, the MMI2 project by ESPRIT in January 1989. MMI2, Multi-Modal Interface for Man-Machine Interaction with knowledge based systems, aims to build a human-machine interface which will incorporate multiple modes of interaction; adapt to different classes of users; and provide intelligent, cooperative dialogue facilities. MMI2 will use the design of computer communications networks as its problem domain.
Links to other parts of RAL were boosted by the Lyman Space Mission project. A feasibility study was undertaken on the role of a knowledge-based system acting as an Operator's Adviser, which would capture some of the knowledge of the resident astronomer.
KE Group has achieved direct industrial sponsorship of its work from the Water Authorities and British Telecom. These projects will begin in the next reporting period. So, too, will projects sponsored by JFIT's IEATP programme. Considerable effort has been expended by KE Group this year in research proposal preparation and negotiation.
The KE Group has contributed directly to the support of the SERC's Engineering Board. In particular, KE Group organised the AI for Engineers EASE workshop, has helped with the AIAI contract, monitored an Alvey project, and acted as secretary of SIGAI.
The Group's research proposal and recruitment successes should make the next few years work rewarding and exciting.
The staffing and management of SE Group has been considerably strengthened this year. Dr S K Robinson was recruited to lead the Group. This has allowed Dr D A Duce to resume his 1M duties full time. The recruitment of Dr J M Spivey of Oxford University as SE Group's second Atlas Fellow, together with the part time appointment of Dr C M P Reade of Brunel, has greatly increased the group's academic links and intellectual horsepower.
SE Group staff are to be congratulated on achieving a PhD, MSc and an MBCS between them this year. Education, training and membership of professional bodies are strongly encouraged in this Division.
SE Group continues to collaborate closely with Manchester University on the IPSE 2.5 project, one of the largest Alvey projects. SE Group's role is concerned with the mechanised support of formal methods of software development and in particular with aiding formal reasoning itself. In the first half of the year the first prototype, FRIPSE, was constructed: Manchester implemented a generic proof assistant whilst RAL constructed a prototype structure editor for BSI VDM. FRIPSE has since undergone a major redesign and changed its name to Mural. SE Group's effort has improved the user interface and developed a mechanism for translating specifications and designs into theories in the generic proof assistant. SE Group has continued to work on the BSI standard for VDM.
SE Group has completed a project to produce a yacc-like parser generator system, written in Standard ML, which produces output code in Standard ML. This work was undertaken as part of the Alvey FORSITE project.
Formal specification and verification are the cornerstones of SE Group's research programme. Both Atlas Fellows and the part time academic visitor are extremely active in these areas, publishing profusely. With the Atlas Fellows linked to Oxford's Programming Research Group, their work has a Z, Occam and CSP flavour. This fits well with SE Group's work on the formal specification of computer graphics and in particular, ISO Standards for GKS and the development of a Reference Model for Computer Graphics.
SE Group, with DCS Division, is a partner in a new ESPRIT project called ARGOSI (Applications Related Graphics and OSI Standards Integration) which seeks advances in our capability to transfer graphical information across wide area networks. This year saw a major effort to bid, negotiate and start up this project.
SE Group completed work on the ERIL project, funded by SERC's Computing Science Committee. ERIL is a theorem prover for experimenting with specifications and for prototyping abstract data types, and is used by several research groups in the UK and abroad. ERIL is based on the use of rewrite rules for computing and for reasoning with equations. As well as developing ERIL, the group has organised its distribution and support and run tutorials and special interest group meetings which have helped to form an active UK research community in the equational area. The ERIL work has led to the award of an IEATP research grant on the Verification Tools for LOTOS Specifications.
SE Group has been working directly for the EASE programme. This has primarily been on the definition of the overall EASE strategy, the planning of the EASE IPSE-type environment, and the assessment of current IPSE technology such as PCTE, ISTAR and Maestro. The Division has also supported the Environment Committee's IT Advisory Group.
In spite of the considerable effort which the Division has had to expend on the pursuit of external research funds, it is pleasant to see how much excellent technical progress has been made. With over 30 external publications this year the Division has made a major contribution to IT research in the UK; has formed lasting collaborations throughout Europe and the USA; and has helped shape the future of the EASE programme.
The Alvey-funded Paralfex project came to an end in March 1989. The two main themes of the project, graphical explanation and knowledge base architectures for the financial domain, had both been brought to a successful conclusion.
The Source of Finance Adviser, the project's testbed system, had its graphical interface considerably enhanced. One aim was to enable a mode of consultation whereby the user directed progress. This was done by the user volunteering the information which would lead to determining whether a source of finance was feasible or not. The second aim was to allow the user to hypothesize about different situations. It is now possible for the user to change earlier inputs and see in a consistent way the effects on the viability of the sources.
In addition, the explanation mechanism was made more flexible by allowing the user to look at the explanation for any entity in the currently visible network and allowing the network to be expanded at any time with mouse clicks on the finance options. However, a bug still exists which causes the answer/input text in the command window to be printed over the questions being asked. This required a lot of effort to track down and report to Ferranti, who have not supplied a fix.
A general knowledge base architecture has been developed using the Source of Finance Adviser as a reference to stimulate and test ideas. The architecture has two elements: an inheritance network of entities in the domain, and what we have called problem solving contexts. The inheritance network achieves both abstraction and explicitness of knowledge, which have been shown to be vital for high-quality explanation of reasoning and conclusions. Entities in the knowledge base stand for domain concepts, general and particular, and represent their stereotypical form. Specialised types and instances of concepts inherit their properties by default (e.g. control knowledge of how to obtain a value for the entity). Any differences from the default must be explicitly specified. This makes for a better understanding of what reasoning is being done and why. In addition, the representation of control knowledge is a simple and uniform one which enforces explicitness about the ordering of the investigation, something which in rule-based expert systems is too often encoded implicitly in the rule order.
Problem solving contexts are a means of representing situations where, as a result of making an assumption or reaching a conclusion, a whole mass of new information becomes available to the problem solver. An example is the knowledge of expected financial ratios which becomes available when a company's market sector is known. Such knowledge should be represented declaratively; it is integrated with the inheritance network by having a context declaration associated with an entity in the knowledge base specifying what new values are acquired when that entity is being or has been evaluated.
The architecture has implications not only for explanation but also for reuse of knowledge. There are a number of ways in which this desirable goal is made possible. Firstly, the inheritance network obviously lends itself to reuse by its very nature: domain-independent entities such as Derived Number and domain- specific ones such as Financial Ratio will be needed in many different applications. Secondly, there is the idea of 'packages of knowledge' standing for such groupings of domain entities as the company's balance sheet. This is integrated by permitting the context declarations to refer to such packages bodily when appropriate. Finally, the idea of 'perspectives on knowledge', in which different user roles have different views of a situation but still have much in common (like lender and borrower), are handled as a special case of the problem solving context in which the nature of relationships between entities are being redefined in the new context.
The Knowledge Engineering Group's sandwich student from Teesside Polytechnic (LNW) spent some time on the Paralfex project. The Source of Finance Adviser was successfully ported onto Goldworks, an AI toolkit running on a Compaq 386 machine, and some experimentation was performed with forms of explanation made possible by Goldworks' facilities. An interface to Lotus 1-2-3 was implemented which allowed values to be read from a spreadsheet without querying the user. Not only did this work give us another comparative version of the Adviser, it was also valuable input to the evaluation of toolkits.
MMI2 is a five year research project drawing on 60 man years of effort that started in January 1989 with funding from the CEC under the Esprit initiative. The consortium undertaking the project consists of two software companies, two universities and three research laboratories: that is BIM (Belgium) as prime contractor, with Intelligent Software Solutions (Spain), University of Leeds (UK), Ecole des Mines de Saint-Etienne (France), SERC/RAL (UK), ADR/CRISS (France), INRIA (France).
The objective of the project is to develop a highly interactive multi-modal interface for human-machine interaction with knowledge based systems. More specifically, the project aims to build a human-machine interface which will:
The interface will provide simultaneous interactive use of modes suitable to support the developed skills of professional users, and natural communication modes well suited for naive users. These modes include: Natural Language typed at the keyboard (English, French and Spanish), Graphics with Direct Manipulation, Mouse Gesture, and Command Language.
The interface will initially be developed to accommodate a Prolog based knowledge based system for local and wide area computer network design, which provides the requirement for intelligent dialogue, different classes of users, and the integration of multiple modes of representation and communication. The interface will be designed to be portable across a range of potential applications of knowledge based systems. Tools for the rapid adaptation of the interface will be developed in the later stages of the project.
In the first phase of the project, the general architecture of the overall system has been described, the common meaning representation (CMR) which will be used to communicate between the modules in the interface has been outlined, and work has begun to specify individual modules of the interface and demonstrator application.
The Knowledge Engineering group at RAL have been involved in the specification of the general architecture and the common meaning representation, and have a specific involvement in the modules which address user modelling and the graphical interface mode.
The overall architecture of the project can be viewed as a three layer structure with the different interface modes represented by modules (or experts) in the upper layer, the central layer contains 6 experts to manage dialogue, and the lower layer represents the application program. The six experts in the dialogue management layer may be imagined as being placed on the points of a pentagram with one in the centre. The central expert acts as a switching facility to direct information packets between the other 5 dialogue management experts; the second communicates with the application; the third stores information about the state of the screen and user interface; the fourth constructs and manages a model of the user; the fifth manages the dialogue context; and the sixth manages the semantic information required to communicate between the user and the interface.
User modelling in MMI2 is performed by one of the experts in the dialogue management layer of the architecture. This component records information about the level of ability of the user, what type of user it is, the preferences of the user, the knowledge that the user possesses and what the user is trying to do. Using this information, the system can be more cooperative in its response because it can adapt its behaviour according to who is using the system.
Work on the user modelling component has been carried out by RAL during the first six months of the project and has included collaboration with partners in the project consortium who are involved in the knowledge acquisition for the project. During this time, an extensive literature review on user modelling was carried out for the deliverable dl part 1, and the architecture of the user modelling expert and its role were defined in the deliverable dl part 2.
Work at RAL on the user model is continuing in the second six months of the project. During this time, the basic framework of the user modelling process will be programmed in BIM_prolog. Collaboration will continue with partners involved in knowledge acquisition to discover the way in which real experts in the domain create a model of their clients during interaction with them. This will become the content of the user model. Meanwhile theoretical work will continue at RAL to investigate ways of exploiting the full potential of natural language for user modelling.
The CMR is the language which will be used to communicate between the various experts in the interface. This requires the CMR to represent not only information derived from English, French or Spanish natural languages, but also from graphic, gestural and command language input from the user and the output from the application program.
The details of the representation have not been completed yet, although several features of it have been agreed. A major review of representations used for similar systems has been undertaken by RAL and delivered to the CEC. Since it is essential that all interface experts can use the CMR, one of the major problems with establishing it is to manage the agreement of all the partners in the project on the requirements it must meet and its structure. The initial specification of the CMR is planned to be developed by the end of 1989.
The CMR will use a reified typed first order logic with annotations to describe speech acts. Temporal information about the dialogue event, language specific details to enable the resolution of deictic and inter-sentential anaphora and other non-common information will also have to be added to the CMR packet sent between experts, although they will not be incorporated in the CMR itself.
The graphics mode in MMI2 is required to support user input to the demonstrator application KBS in the form of diagrams of computer networks which the user is designing. It will also have to output graphic forms of networks, but also data about the network performance as graphs and tables. These in turn should be modifiable by the user by direct manipulation, but also through natural language or a combination of natural language and direct manipulation. These requirements have given rise to several interesting research questions. For example, what in the domain representation is being referred to if the user asks for the longest bar in a histogram to be extended. The representation of both the graphical description itself and the domain information being represented by the graphic at a level which will support the reasoning required to perform this request is currently absorbing much research attention.
A review of graphic interfaces has been performed and presented to the CEC. The specification of the graphical mode will continue into 1990, with the first program due for delivery in December 1990, although the more complex components required to solve problems such as that outlined above are not expected until mid 1992.
A feasibility study was undertaken on the role of a knowledge-based system in assisting in the Lyman Space Mission (the Far Ultraviolet Spectroscopic Explorer). The role examined was that of an Operator's Adviser, which would capture some of the knowledge of the resident astronomer. The benefits would be savings in cost and time, consistency of performance and archiving of knowledge. The study resulted in a contribution to the proposal for UK participation in the Lyman mission. It is hoped that funding for the proposed work will follow.
The Knowledge Engineering Group has just begun work on a project for a consortium of Water Authorities and Water Companies under the coordination of the Water Research Centre, Swindon. The aim is to develop a pilot knowledge-based system for assisting in the operations of water network management and control. The project plan calls for a definition phase to put some limits on what can be achieved and to outline some essential requirements, followed by a phase of prototyping and the building of a pilot system for one particular site. The pilot will be such that it will then be possible to create a platform - a site-independent but domain-specific knowledge base on which further site-specific applications may be built.
Not only will the results of the work be useful to the consortium members, but they will involve some challenging research issues such as time-constrained reasoning, representation of network knowledge and truth maintenance.
The project is, at the time of writing, in the definition phase. The first knowledge acquisition meeting was held at the offices of Wessex Water in Poole in August.
Work is about to begin on a joint KBS application project with British Telecom (project ADAM) which they will fund on the planning of strategic management decisions (CKYL, SCL, GAR). Further application project research is planned with British Telecom which it is hoped will be started within the next year.
A proposal for a CASE studentship at the University of Buckingham for work on Knowledge-Based Systems in Strategic Planning was prepared by JRG and GAR. The proposal was successful and work related to the group's activities is expected to start in January 1989.
Following the feasibility study on the project for the Lyman Space Mission, a proposal for the full joint UK/US/Canadian research project, including the proposal for a Knowledge Based application to be undertaken by this group (HRC, SCL, GAR, LNW), has passed the Phase A review by the British National Space Centre and it is hoped that it will be funded in approximately two years.
Following proposals submitted to JFIT under the IEATP programme in August 1988 two research projects on KBS development methodologies are currently under contract negotiation and are expected to begin funded work in the near future. Logica will act as prime contractor on Project Gateway (CKYL, GAR, MDW) which will investigate metrics for large knowledge based system. British Aerospace will act as prime contractor on Project Rocket (CKYL, SCL, GAR, MDW) which will investigate and develop methods and tools for the development of closely coupled knowledge based systems and conventional software systems.
Members of the group (CKYL, GAR, MDW) are currently drafting proposals in response to the second call for the CEC Esprit IT initiative for further research on KBS development methodologies which will be submitted in 1990.
An initial outline research proposal has been submitted to the second call for proposals under the IEATP programme to investigate measures for assessing the impact of IT on user organisations, with STC as prime contractor (CKYL, GAR).
Several outline proposals have been submitted to the Joint Research Council Initiative on Cognitive Science and Human Computer Interaction by MDW and members of other groups in Informatics for interdisciplinary research.
In December 1988 members of the group (GAR, MDW) helped organise a workshop at RAL on behalf of the SERC CFC to review the current support offered to engineers for work involving Artificial Intelligence (AI) and to determine the engineering community's requirements for continued support.
Following this workshop, the CFC accepted proposals for the EASE programme which includes general support for AI use by engineers, and specifically it will be incorporated within the EASE software environment. Members of the group have been working with members of the software engineering group on proposals for the AI component of the EASE software environment.
Another aspect of EASE where members of the group have contributed is in developing the structure of a comprehensive survey of the current use of computers by engineers and their perceived future needs. This survey is expected to be conducted in the near future.
During the year MDW has monitored Alvey project IKBS 098, A Knowledge Acquisition Methodology for Research Associations, in which role he has attended project management and technical meetings, assessing both aspects of the project's progress.
Until March 1989 MDW undertook the secretaryship of the committee SIGAI. SIGAI has been an advisory committee on Artificial Intelligence to SERC and Alvey in the past. SIGAI is currently in abeyance while the SERC/DTI JFIT is being organised.
The working group for BSI Prolog Standardisation is IST/5/l7. This group now meets only to discuss and promote the UK view on contentious issues rather than defining the BSI Standard.
The effort of defining the standard has been taken into international levels where ISO working group SC22 WG17 is working to produce an ISO standard for Prolog. The BSI Standard will be the same as the ISO Standard. UK delegates to the ISO meetings are elected from the BSI group.
The ISO meetings are held in the spring and autumn every year. The next is scheduled to take place in Ottawa, Canada on October 11-13 1989 followed by Vienna, Austria (Spring 1990), Marseilles, France (Autumn 1990) and Budapest, Hungary (Spring 1991).
The ISO Standard Prolog currently exists as Draft 2.0 which will be discussed in Ottawa.
The Joint project with the Water Research Centre requires a toolkit on which to build the KBS. The evaluation of two contending systems Goldworks II and NEXPERT was started by LNW before she left and is now to be concluded by CKYL.
During the year, members of the group (HRC, SCL, CKYL, MDW) have contributed several book reviews to the AISB Quarterly, and acted as referees for other academic journals. MDW has continued as a member of the editorial board of the journal Interacting with Computers.
A series of Divisional colloquia has been organised by SCL with the objective of maintaining awareness and stimulating interchange between members of the Division. The colloquia are given by members of the Division on aspects of their work, on a particularly significant paper (the journal club idea) or on some general topic of interest, and take place every few weeks.
None
None
This year the Software Engineering Group has remained at about the same overall numbers, although the members of the group have changed. In November, Dr Jeremy Dick (AJJD) left to work at Racal Research Ltd, Reading. In October, Dr Mike Spivey (JMS), an Atlas research Fellow, joined the group and in April, Dr Stuart Robinson (SKR), who had been a part-time visitor from Brunel University, joined the group as Group Leader. Dr Chris Reade (CMPR), a lecturer from Brunel University, also joined in April, filling the part-time visitor role vacated by SKR. CMPR's interests in formal methods compliment the activities in the group in this area and go part way to fill the gap left by Jeremy. In July, Brian Matthews (BMM) obtained a permanent HSO post in the group and, finally, Dr Francese Cormellas (FC), our visitor from Universitat Politecnica de Catalunya, returned home to Spain in August having completed his visiting year.
BR received a PhD from the University of Edinburgh with his thesis entitled The Design and Implementation of an Interactive proof editor and became a MBCS in May 1989. BMM successfully completed a MSc in the Foundations of Advanced Information Technology at Imperial College London in October. This culminated in an original research project, which BMM undertook on the use of the ERIL equational reasoning system, and completed in October. JCB registered for a Ph.D (Algebraic Approaches to Model-Oriented Specification) at Manchester University in October 1988.
JMS is taking a leave of absence at the Computer Research Laboratory, Tektronix Inc., Beaverton, Oregon and at the Department of Computer Science, Cornell University starting in October 1989.
The current staffing of SEG is summarised below:
EASE
The current Engineering Board supported research community tends to be split along Committee lines. This does not take into account computing solutions that often straddle several communities' application areas. Therefore, CFC is taking an active role in facilitating cross-fertilisation between Committees, disciplines and industry in the computing area and, as part of this activity, the Engineering Applications Support Environment (EASE) programme has been initiated with the mission statement:
To stimulate and encourage engineers to use appropriate, state-of-the-art software and hardware to enhance their research
Specifically, CFC wishes to encourage engineers to migrate to state-of-the-art integrated project support environments (IPSEs). These will be available in the 1990s as the result of Alvey and Esprit funded research activities.
Integrated Project Support Environments
Until very recently, most software has been developed and maintained predominantly on large, centralised computer systems using ill-matched tools and methods. It is now recognised that better control and productivity gains can be achieved through increasing the amount of Computer Aided Software Engineering (CASE) support available.
A number of such tools have been brought to the market but have, in general, suffered from the lack of a unified approach. Users incur unnecessary costs and risk loss of control in manually converting results produced by one tool into a form suitable for another. Coherent, complete tool sets covering the whole range of tasks of system development and maintenance have not been available on a common basis. It is against this background that the concept of an Integrated Project Support Environment (IPSE) has emerged. An IPSE is a software development environment
Engineering Software
The software systems developed and supported by the engineering community are, in general, large and complex. Such activity, in the academic community, frequently involves an integration of commercial and academic software to produce ENVIRONMENTS suitable for interactive design. The need for tools to assist this integrative activity have already been identified by the CFC Workshop on Tools for Integration, and are actively being explored within the Software Engineering R&D community.
The software design process itself is one of exploration of alternatives and, consequently, most Engineering Board Committees see the need for the expert and intelligent retrieval of data and its interpretation. Such intelligent access and use of data is actively being researched within the Knowledge Engineering R&D community.
The major activities in software engineering (SE) and knowledge engineering (KE) supported by JFIT (Joint Framework for IT) can therefore be expected to deliver systems that are relevant to the engineering community as a whole. To assess this expectation, JFIT is anxious to have joint programmes with the other Engineering Board Committees to try out SE and KE research results in real environments.
In the context outlined above, the Systems Engineering Division needs to act as an interface between the (eg JFIT sponsored) IT research community, in particular the Software and Knowledge Engineering communities, and the traditional engineering communities which are supported by the EASE programme. To meet this requirement the Division needs to have a (well informed) foot in both camps, acting as a bridge (in the form of education and technology transfer activities) between the two.
Therefore, the major thrusts of the Software Engineering Group are:
This is all undertaken within the unifying research theme The Quality Certification of Software Products.
Introduction
A software product is a set of components such as
A software process is that which is done rather than that which is produced.
For example:-
process product
programming source code
compilation executable code
Any SE programme has two major goals
SEG does not tackle productivity issues, except where SEG's own productivity in achieving quality objectives is impeded. The concentration is on quality issues.
Software quality is a property of the software product. Software quality has nothing to say about the process, ie high quality of a software product is not necessarily guaranteed by a high quality software process.
The properties which make up software quality include performance, reliability, security, correctness, usability and cost. Certain quality properties may be directly measurable, for example CPU time, memory occupancy, lines of source code, etc.
However, other properties have a certain subjectivity that makes it difficult, if not impossible, to measure, for example usability, maintainability etc.
Certain quality properties are certifiable. Certification is taken to mean the independent (of the software producer) and repeatable demonstration that a software product possesses a specified property or set of properties. A simple example of certification would be the independent verification that a software product's executable code is less than 64 Kbytes; once verified a certificate would be issued indicating that the product has been examined and the truth of the claim established.
The SEG Research Theme
The idea of having a long term theme, which ties together various semi-independent activities within a group, has advantages eg: Reliability at Newcastle, Ada at York, Formal Methods at Edinburgh.
The SEG research theme is the search for techniques to demonstrate
The Quality Certification of Software Products
that is, techniques which will enable SEG to independently and repeatably
demonstrate that certain software products possess certain well specified
properties as outlined above.
Within the group, quality certification is approached from the formal methods angle. Two strands of interests are describable as:
IPSE 2.5 is a major Alvey project to research and develop an Integrated Project Support Environment based on advanced distributed systems and man-machine interfaces, and incorporating support for both the practice and organisation of design tasks. A major objective is the development of an integrated framework for supporting the use of formal methods in software development, including formal specification and theorem-proving techniques.
The project began in October 1985 with three initial collaborators (ICL, STC and Manchester University) and had a successful first review with the Alvey Software Engineering Directorate in July 1986. RAL joined the project in April 1986. Three additional industrial partners have also recently joined the collaboration.
At RAL, Juan Bicarregui (JCB) and Brian Ritchie (BR) are engaged in the project. DAD is the RAL representative on the project Review Board. The work at RAL is carried out in very close collaboration with the University of Manchester. JCB and BR spend a fair proportion of their time at Manchester.
The Manchester/RAL part of the IPSE 2.5 project (Theme C) is concerned with the support of formal methods of software development and in particular with aiding formal reasoning itself. The intention is to build tools which enable a user to construct proofs at the workstation; modern workstations such as the SUN3 should make it possible to design proof assistants which are much more usable than earlier tools developed around glass teletype interfaces.
Manchester has concentrated upon the specification and implementation of a generic proof assistant (known variously as FRIPSE, Magus and (currently) the Mural Proof Assistant); RAL has concentrated upon implementation of a "Mural VDM Support Tool" for both specifications and designs using BSI VDM. The intention is that specifications (and the refinement of one specification to another) in the latter should be "translated" into theories in the generic proof assistant, including those proof obligations which must be discharged in order to verify a specification or design step.
Originally, the VDM support tool was intended to be minimal, the main aim being to provide an existence proof that the generic proof assistant could be instantiated to support particular formal methods. However, in consideration of the present status of the IPSE 2.5 theme responsible for a full support system, RAL have been encouraged to make the design less minimal.
Over the period August December, the first prototype, FRIPSE (Formal Reasoning in an IPSE) was constructed: Manchester implemented a prototype generic proof assistant from a formal specification, whilst RAL constructed a prototype structure editor for BSI VDM. The latter included the ability to generate proof obligations from specifications and translate them to expressions in the generic logic of the proof assistant. This early version was demonstrated at a review of the IPSE 2.5 project on January 17th.
JCB and BR submitted and subsequently presented a paper on the VDM support work to the 1st International Conference on Systems Development Environments & Factories (Berlin, May 9 -11). They also gave demonstrations of the same prototype system at the conference.
Since this time, FRIPSE has undergone a major redesign (and been renamed as Mural). The largest changes are to the user interface, taking into account the experiences with the initial interface. The initial link between the VDM support tool and the proof assistant consisted solely of a means for translation VDM expressions (such as proof obligations) into proof assistant expressions. It was decided that a fuller translation mechanism was necessary (and feasible within the lifetime of the project). Thus BR and JCB's time on IPSE 2.5 since February has been divided between improving the interface to the VDM support tool and specifying a mechanism for translating both specifications and designs into theories in the generic proof assistant. Some initial ideas on such translation are documented in a project document. Both tasks are ongoing at present.
There are also plans to produce a Book of Mural, documenting the system and the designers' experiences in implementing in Smalltalk from formal specifications. BR and JCB have produced a first draft of their chapters.
BR and JCB are currently searching for funding for further work with Mural beyond the lifetime of the IPSE 2.5 project. Future proposals include collaborative work with Tim Clement at Manchester University relating specifications and program transformation.
BR and JCB have continued to sit on the VDM Specification Language standardisation panel (IST/5/50) with the intention of making the Mural VDM support tool follow the standardised language as closely as possible. During the year the major components of the language, abstract syntax, ASCII and 'mathematical' versions of the concrete syntax, context conditions and semantics have gradually stabilised and been brought together into the first version of the proto-standard which was released in May. There are also moves afoot to set up an ISO working group for VDM SL.
The aim of this project was to produce a yacc-like parser generator system written in SML which produces output code in SML. This project was undertaken as part of the Alvey FORSITE project. The SML-Yacc Parser-Generator System was completed by BMM at the end of November 1988.
There were two major problems to be overcome in the final testing of the system. One was associated with a problem in the Edinburgh ML system which meant that large parsers would not be loaded into the system. BMM tackled this problem by breaking the output down into smaller components, which are then rebuilt when loaded into ML. The other more major problem was a problem of an underspecified case where no action is associated with a empty production in the grammar. In this case a value is needed but none is known to the system. BMM resolved this problem by requiring the user to declare a dummy value in such cases.
As a case study in the use of the parser-generator, an implementation of a lexical-analyser-generator was produced by SKR and BMM. This tool was designed to be compatible with SML- Yacc. Although a full implementation was not achieved in the time available, this work demonstrated the ease of use and general applicability of the Parser-Generator system.
To complete the work on the parser-generator, a prototype parser for the Z-specification language was produced by BMM. This used the grammar for Z as used in the for the FORSITE project and supplied by RAGAL Research Ltd. This parser is the subject of a group report.
The parser-generator has been distributed to Oxford University, RACAL Research Ltd, and Brunel University. Imperial Software Technology has also expressed an interest.
The User Manual for SML-Yacc has been produced by BMM as a Laboratory Report. Further publications on this project are in preparation.
For some years now JGPW has been an investigator on the CICS project being carried out jointly by IBM UK laboratories at Hursley, and the Programming Research Group (PRG) at Oxford. The research objectives are to demonstrate the applicability of mathematics to the development of large industrial software products. These objectives are being met, since the Z notation, which is based firmly on elementary set theory, has been used by IBM to develop the latest release of the CICS transaction processing system. Furthermore, IBM have stated their intention to continue this research.
Working with his D. Phil student, Jim Davies, JGPW completed a specification and implementation of the Ethernet protocol in Timed CSP. This is the first major use of the theory, and it has thrown up many research topics. At present, work is proceeding on a set of proof rules to make proofs easier. It will not be possible to complete the proof that the implementation satisfies the specification without a probabilistic model for CSP.
JMS has been working on a rigorous description and standardization of the Z notation which has resulted in the publication of his book The Z notation: a reference manual.
JMS has undertaken an investigation of tool support for Z specifications and has studied two case studies in formal specification:
JMS has been working on consistency theorems for free type definitions in Z.
Funded by the Austrian Ministry of Science and the British Council, JCPW spent two weeks as Gastprofessur at the University of Klagebfurt in Austria. He taught a course on formal specification in Z.
Working in collaboration with Hewlett-Packard research Laboratories at Bristol, JCPW has discovered some strategies for substantially reducing the burden of proof in the implementation of occam programs from their specifications. The main idea is to prove that there is a normal-form specification which corresponds exactly to a normal-form process. The specifications that are being worked on are then transformed into their normal forms; these then corresponded to normal-form processes, which are then transformed into the final implementation. None of the usual inductive proofs that recursive programs satisfy certain predicates are necessary, the task of implementation being reduced to one of transformations by rewrite-rules.
JCPW has recently started work with IBM UK Laboratories at Hursley on large-scale proofs of parts of the CICS transaction processing system. JCPW has developed proof techniques for Z specifications and refinements, and shown how these can lead to routine calculations using Abrial' s B proof assistant. In order to prove large specifications correct, it is necessary to structure the proofs in a careful way, deliberately minimising the amount of effort required. It turns out that a well-structured specification or design in Z also has a well-structured proof. That is, the proofs of theorems about individual pieces may be composed, just as the pieces themselves are composed. Thus, instead of large proofs, one need carry out only small proofs, thereby keeping the process manageable.
CMPR is beginning a survey of existing theorem proving systems in UK in October.
JMS has been studying the application of the Boyer-Moore theorem-prover to reasoning about specifications in the Z style.
Further, JMS has been developing a theory of exceptions in higher-order lazy functional programming with applications to term-rewriting and implementing a theorem-prover for higher-order equational logic.
JMS has developed and presented courses in the area of Dijkstra's calculus of weakest pre-conditions.
DAD has continued to work on the formal specification of graphics standards. This project started in 1985 and ended in April 1989. The relationship between GKS and PHIGS has received considerable attention. This work culminated in a paper at the Eurographics '89 conference. Essentially PHIGS can be described as a database on top of GKS or GKS-3D. Traversal of this database generates graphical output, which can be described in terms of GKS functions. This simple model does not completely define the relationship between GKS and PHIGS; there are complications arising from differences between coordinate systems and clipping in the two systems. The control of operator attributes, visibility, highlighting and detectability, is also different in the two standards, which limits the analogy.
The relationship between GKS-3D and PHIGS is one aspect of a broader issue, namely, a Reference Model for Computer Graphics. Insights from the specification work formed the basis of an outline reference model. This work is discussed in the section on standards activities below.
A review of GKS is now starting within ISO, and the work of the completed formal specification project discussed above is being fed into this process. Papers have been written for the March 1989 meeting which describe a simpler structure for GKS and propose the removal of extraneous functionality. Central to this proposal is the explicit idea of a picture in NDC space which the application is constructing, and the notion of a workstation viewing this picture through a filter mechanism based on the PHIGS name set mechanism. Some of the ideas in this originated in the original specification project. The description of GKS produced by this project has also been simplified extensively by using more abstract data-types and functionality. A formal specification of the simplified model has been given which has a particularly simple structure.
DAD spent a week working with Drs P J W ten Hagen and R van Liere at CWI, Amsterdam, on formal specification of the GKS input model and extensions thereof, at their invitation. A description of the GKS input model using Hoare's CSP notation, was produced as a result of this collaboration. This paper has been accepted for the Eurographics '89 conference. The description shows clearly the differences between the different operating modes for GKS logical input devices and gave the authors some new insights into the input model.
It has long been felt that it should be possible to allow user-configurable input devices and allow hierarchically structured devices. During a visit to CWI in February 1989, it was found that the CSP input description could be extended easily to describe hierarchically structured devices. Some examples were worked out and the first draft of a paper was produced.
DAD spent a week at INRIA, Paris, teaching part of a course on distributed graphics.
DAD is Vice Chairman of Eurographics and Programme Co-Chairman for the Eurographics '90 conference.
DAD has participated in the BSI Computer Graphics Committee IST/3l and in the international work on reference models and revision of GKS. DAD and FRAH are currently joint editors of these two documents.
Both activities have built on insights arising from the GKS formal specification project. The Reference Model work has deviated significantly from the component/framework model mentioned in the last Annual Report. The work has reverted to a layered structure similar to that proposed in February 1988, but now recognizing four layers: application, virtual, logical and physical. This approach was developed at a meeting in January 1989 and refined at a meeting in May 1989. Since that time effort has been expended on improving symmetry between input and output in the model.
An outline proposal for a revised GKS was developed by FRAH and DAD in conjunction with Dr K W Brodlie at the University of Leeds. The direction of the work has been to develop a cleaner structure than that in GKS ISO 7942:1985. The initial proposals were subject to international review in March 1989 and a fuller document has been completed which takes into account the comments made. This document will be subject to international review in October 1989.
DAD also participated in a meeting in February 1989 which aimed to produce an improved input model for graphics standards. Again, work from the formal specification project provided input to this meeting. A report has been produced which will be considered by ISO SC24 Working Group 1 in October 1989. The results of this work have also been partially incorporated in the revised GKS draft.
DAD has just completed a book with Dr D B Arnold at the University of East Anglia on the first generation of standards for computer graphics.
DAD is involved in the Esprit II ARGOSI Project which began in April and finishes in March 1992. Project ARGOSI (Applications Related Graphics and OSI Standards Integration) is a multi-national European project funded under the Esprit II Programme which aims to advance the state of the art in the transfer of graphical information across wide area networks. Two of the specific objectives of the project are:
The first stage of the project is looking at the classification of applications in terms of their graphics requirements and the networking services required for those graphics. It is important to emphasize that the project is not about graphics in isolation, nor networking in isolation, but about the particular problems and requirements arising from the combination of graphics and networking.
Participation in standardization activities is another major theme within the project, and the consortium members are already well-known for their contributions to either graphics or networking standards in ISO/IEC and CCITT. This reference model, or classification scheme, will also be used to guide their input to the standards-making process to achieve the objectives of improving the quality and applicability of standards. Obviously, the reference model has to take into account a broad range of application areas and existing practical experience with graphics and networking. The consortium members wish therefore to gather information and requirements for the combination of graphics and networking services from as broad a range of organizations as possible, together with experience of using combinations of existing graphics and OSI standards.
RAL are leading the Classification Workpackage and the participants here are currently interviewing a large number of user organizations.
DAD and FRAH worked with the Esprit Directorate and Eurographics, to formulate a proposal for a Special Interest Group on computer graphics within the Esprit Programme. Three information exchange sessions on graphics in Esprit have been organized in conjunction with Professor Encarnacao and Dr Schonhut in Darmstadt, as part of the Eurographics '89 Conference Programme.
This work was funded by SERC's Computing Science Sub-Committee. The ability to reason with equations (for example, to deduce conclusions from a set of equations), is important in a number of areas, for example specification of abstract data types and their validation, program transformation, synthesis of programs, program optimisation and solving equations. Jeremy Dick (AJJD) has over a number of years developed a rich theoretical framework for equational reasoning and has developed a practical tool, ERIL (Equational Reasoning an Interactive Laboratory), based on this framework. ERIL is a useful tool for experimenting with specification and prototyping of abstract data types, and is used by several research groups in the UK and abroad. ERIL is based on the use of rewrite rules for computing and reasoning with equations.
Unfortunately, AJJD left the ERIL project in November 1988. This was a major set back to the progress on this project. However, before leaving, AJJD did complete the implementation of the Recursive Path Ordering and also the Unfailing Completion method within ERIL.
AJJD also commenced a new clean implementation of the ERIL system in the PROLOG language, which was to incorporate new features which seemed desirable in the light of the experience gained in using the ERIL system. Due to changing requirements (see below), this work was curtailed after the departure of AJJD.
During the first half of the year (Sep 88 - Feb 89) John Kalmus (JRK) was involved in the ERIL project. The ERIL Users' Manual was issued by JRK and Jeremy Dick (AJJD) as a Laboratory Report in September 1988.
A paper written jointly by AJJD, JRK and Dr. Ursula Martin of Royal Holloway and Bedford New College, University of London was submitted to Acta Informatica in January 1989. Entitled Automating the Knuth Bendix Ordering, it describes a new approach to dynamically orienting rewrite rules during Knuth Bendix completion which has been successfully implemented in ERIL.
JRK also helped run several meetings over this period. For example, he was the principal organiser for the BCS-FACS Term Rewriting Tutorial and Workshop, the organiser for the "ERIL collaboration meeting" and the co-organiser for the Term Rewriting Group meeting.
During this time, JRK was responsible for the distribution of the ERIL system and User's Manual, and for maintaining a mailing database of postal and e-mail addresses used for publicising meetings of the Term Rewriting Group. BMM took over these functions in April 89 when JRK moved to the EASE environment project.
Early in the year, BMM, in support of the ERIL project, undertook an investigation into the use of the ERIL system as a tool for theorem proving. He found that the tool proved flexible, and could be used for a variety of differing equational techniques, although the size of problem which could be undertaken within the current ERIL implementation was not great. This work became the subject of BMM's MSc dissertation. BMM is now following up this work by investigating the flexibility of the ERIL system for differing configurations of the Knuth-Bendix algorithm and the Unfailing Completion algorithm.
In order to continue activity on the ERIL project, BMM undertook (in February) to review the work undertaken under the CSSC grant on the ERIL project. From discussions at this time and later, within the department and -with informal collaborators at Royal Holloway and Bedford New College, it was decided that a sensible approach to further activity was to 'freeze' ERIL in its current implementation, and to concentrate on extracting essential features with the view to producing a 'toolkit' of useful functions for equational reasoning within a functional language such as Standard ML. BMM has been carrying out some preliminary studies into this, and is producing a formal algebraic specification of the toolkit. Contact has been made with the group at CRIN, Nancy, France, who use a toolkit produced in the functional language CAML. It is hoped that this work will continue under the IED LOTOS verification project which commences in October.
CMPR, DAD, and BMM will be involved in SERC IED project with RAL/RHBNC/Glasgow Verification tools for LOTOS Specifications starting in October 1989.
CFC wishes to encourage engineers to migrate to state-of-the-art integrated project support environments (IPSEs). These environments will be available in the 1990s as the result of Alvey and Esprit funded research activities. However, due to the wide technology gap between existing practises and IPSEs, CFC believes that an environment developed and offered now using this new technology would run the risk of being rejected by the community as requiring too much retraining.
CFC therefore recommended the approach taken in developing the prototype environment be as follows:
A major decision at the last CFC meeting was that Informatics should produce the initial software environment and undertake the assessment of IPSE technology. To encourage use of the initial environment by the community, it should be populated by a set of tools or modules with wide applicability. SEG have been given the task of planning this activity and assessing IPSE technology as it emerges.
In the second half of the year (Mar 89 - Aug 89) JRK began working in the area of Software Engineering Environments, under the direction of SKR.
As an initial assessment activity, a course on PCTE and Eclipse, run at University College Aberystwyth, was attended by SKR and JRK. in April 1989. It became clear during the course that the software provided by the Sapphire Project, which was the only SUN version of PCTE (In-Kernel) available at that time, was in a rather volatile state and would continue to be so for some time. Other candidate environment bases, such as Maestro and ISTAR, are being explored.
A major task for JRK during this period has been to gain a working knowledge of available IPSE (Integrated Project Support Environment) technologies, with the view of producing a survey document (as an internal report) by the end of the year. As part of this process, an Emeraude implementation of the Portable Common Tools Environment (PCTE) has been installed on a SUN 3 Workstation, and will be used as part of an assessment of PCTE as a Tool Support Interface (TSI).
In collaboration with He Jifeng, a colleague at the Oxford PRG, JCPW has been working on a combination of Z and CSP. They have developed several special refinement techniques that have proved to be useful. This has resulted in several case studies which show how concurrent systems may be developed in a state-based framework.
As an alternative to the approach using a combination of Z and CSP, JCPW has investigated Cliff Jones' work on the specification and development of concurrent systems using rely and guarantee conditions. The original work concentrated on operation decomposition in particular on parallel decomposition - and issues of data refinement were less well treated. JCPW has discovered hitherto unknown problems in the technique, and shown that they militate against hierarchical design. As a development proceeds, important sorts of refinements cannot be proved correct without knowledge of how the rest of the system is being developed. If such refinements are permitted, then the technique cannot be described as hierarchical; if, on the other hand, they are forbidden, then the technique has poor expressive power. JCPW has started on some collaborative work with Cliff Jones that promises to attack these problems.
CMPR has written a book Elements of Functional Programming which has been published by Addison Wesley.
JMS has been exploring the applications of category theory to functional programming and program transformation, and has prepared and presented courses in this area.
JMS has been studying the implementation techniques for lazy functional programming languages.
The distribution of the ERIL equational reasoning system continues, being carried out by BMM after the departure of AJJD. New sites include Twente University, the Netherlands, and RSRE, Great Malvern.
AJJD also set up a mailing list for the Term Rewriting Group, which BMM has now taken over.
BMM continues to distribute the Rutherford Standard ML/Cambridge LCF system, although requests for this system are declining.
Benchmarks for software engineering tools were requested in support of the EASE programme. However, due to the highly interactive nature of most such tools, this did not prove trivial. BMM attempted to produce a version of SML-Yacc suitable for automated benchmarking. This was not a success as Edinburgh Standard ML proved not portable enough for the purpose. BMM then produced a simplified version of the ERIL system suitable for this purpose.
Poly-ML from Imperial Software Technology was received and mounted by BMM.
The overall objective of the EASE environment project is to develop a national IPSE for EASE by the year 1995. The environment is to be based on de facto standards. The environment is to stimulate and encourage engineers to use appropriate, state-of-the-art software and hardware to enhance their research. Its use is to be encouraged by being designed so that it can be populated by a set of tools or modules with wide applicability.
As a sub-goal, the initial environment needs to be designed and built. This process is to be led by SEG. Existing 'conventional' environments (such as ECTASY and EKS) need to be examined and the results of this uti1ised in the design process. A suitable environment basis needs to be identified and the appropriate tools then integrated onto that basis.
Further, emerging IPSE technology needs to be monitored as part of the design process for the follow-up, IPSE based environment. This assessment activity is seen as an on-going commitment to SEG.
SEG research projects should lead towards increasing SEG's strength in
SEG needs to build up the right intellectual environment (the skills, enthusiasm and energy of the members of SEG) and the right physical environment (offices which are conducive to concentration but allow group interaction, computing facilities, laser printers, etc).
On the intellectual side SEG needs to build up its
Specifically, our research plan is
DCS is primarily a service Division. It serves the communities supported by the bodies which fund the Department. It also serves the other three applications Divisions within Informatics. Finally it makes its expertise available to the rest of RAL. Its main functions are:
For K F Hartley this was the year of the paper mountains. Firstly a start was made on archiving documents relating to the Department's involvement with the Alvey Programme. Having been responsible for helping set up a Departmental archive, KFH then set about trying to fill it and follow the guidelines. There are five filing cabinets full of papers and several cardboard boxes which were returned by Adrian Wheldon, having been given to him in March when the Alvey Unit evaporated.
Secondly, it was decided by ECFE that RAL should join the Open Software Foundation, which is the organisation of companies IBM, DEC, HP-Apollo, Apollo-HP, Siemens, Bull and 150 others - trying to create a portable rival to the Sun/AT&T flavour of Unix. KFH and Neil Calton attended a Members Meeting in Monte Carlo, added our names to various SIGs and have since been deluged in paper. Massive draft descriptions of the Operating Systems Component, copies of every response to Requests for Technology, project plans, discussion papers and so on. Much of this is sent out by special courier. It is certainly possible to see what we get for our subscription. If only we had the manpower to become actively involved in a couple of SIGs or even to take a serious look at all the papers.
The third mountain of paper arrives regularly from the library at Swindon. The Department Head decided that we ought to see copies of the Official Journal of the EEC. This turns out to be comparable to Hansard, reports from the High Court and more. Much more. One can only be impressed by the effort involved in typesetting so much material at high speed with so few errors. FRAH persevered for a week, then asked KFH to select relevant pieces. We are now trying to stop the flow altogether.
Fourthly, Helen Jenkins has succeeded in merging KFH's files with his Group Leader's files into one coherent system. She is also able to retrieve files on request, at impressive speed.
Finally, Geoff Lambert manages to obtain, organise and distribute a hill of paper which describes the Everest project in more detail than most of us care to know about.
Following the Departmental reorganisation for July 1988, this year has also seen many changes. The computer room has been drastically modified and improved. Lab 11 has been completely refurbished and again greatly improved. New offices in R1 have been allocated, occupied and fitted with appropriate communications, sometimes in that order. A major upgrade of Sun's operating systems, involving a hundred or so machines, was carried out with minimal disruption. At the end of the year we are about to embark on an experiment in doing without maintenance of our diskless Suns, as just one of the changes brought about by the Summer financial crisis.
Alvey has gone, Esprit I is evolving into Esprit II, and IEATP may arrive soon. Some things, however, remain constant. We are still short of text and graphics effort, could do with more in systems and communications, and have nobody to work on Argosi. In spite of that, progress has been made with GKS, local and remote access to Supercomputers and in Directory Services. Major visible contributions have been made to EASE through the work of the assessment team, improving the quality, quantity and distribution of the Engineering Computing Newsletter, and organising a succession of too successful (that is, over subscribed) education events.
This year the decision was taken to recognise at Divisional Meetings achievements over and above the call of duty. The following Achievers have been embarrassed in this way since the scheme started:
At the start of this period (September 1988), the group membership was:
Jeremy Isserlis left in November. Andy Jackson transferred to A & G Division in December. Mark Phillips returned to work in November, following his accident in the summer.
The July 1988 reorganisation placed Tony Lucas, Subodh Chanda and David Barlow in the Assessment team on a temporary basis. From September, Tony and Subodh decided to remain there; David moved to Support.
Linda Reed transferred to the Transputer Coordination Unit in February 1989, Karen Holloway joined Distributed Systems in January 1989, and Pat Athawes transferred to Systems Administration (taking over Linda's work on Security). Kevin Lewis was involved in a rugby accident which put him out of action from January to March 1989.
Mark Phillips left in June 1989. Karen changed her name half way through (just to confuse the mailing lists).
Ines Vollmer and Ian Johnson were promoted to HSO.
At the beginning of August, the group membership was:
In the middle of August, the Communications and Systems parts of the Distributed Systems section became a separate group. Since this took place at the end of the reporting period, it is not presented separately here.
The Group has three sections: Distributed Systems, Systems Admin and Text/Graphics. Distributed Systems has two main activities: systems/communications support and assessment. Systems Admin covers operations and support activities.
The main objectives are as follows:
RAD and APM have been involved with the Ethernet bridge evaluation work contracted to the Division by the JNT. Stage One of the work involved RAD in the construction of a questionnaire to around 20 suppliers, to find out what products were available, along with brief details. From the replies to the questionnaire, a shortlist was drawn up of products which should be subject to detailed testing in future.
Several remote bridges (ie bridges that can connect Ethernets that are not at the same site) were evaluated in detail. This proved to be a moderate hassle, particularly in getting evaluation machines on the dates promised, and with physical interfaces that we had suitable cabling for. However, a report was submitted to the JNT, and there are now a set of recommendations for this area.
During the summer, the cisco combined bridge and router was evaluated. This had a dual interest: the JNT were interested in its bridging capabilities, and the routing capabilities were of interest to EASE (with a view to use of high-speed links to the Cray etc).
Part of the work involves publishing summaries of what is available to the community. This turns out to be a logistic headache, as the JNT decided that they wanted extra reports in a form suitable for publication to the wider community, but omitted to say this in the original negotiations, specifying reporting to themselves alone. However, a compromise has now been found whereby suitable extra reports can be generated without too much extra work.
A significant change also occurred in the JNT's attitude to evaluations during this period. Whereas before they favoured producing a single recommendation for a product, and using this as a bargaining tool with suppliers, they have now adopted a policy of approving an arbitrary number of products, and letting individual sites do the haggling. This shift is very similar to that made by CFC with workstation evaluation. Although this means potentially more bridges being evaluated in detail than was first envisaged, it makes running the rolling evaluation a lot easier than would otherwise have been the case.
NBC has worked on implementing a Yellow Pages service on the Pyramid. A client-only service was established on the Pyramid Workcenter (nfs4) and utilities altered to permit YP access. APM has been helping with the Yellow Pages on the Pyramids by compiling and testing utilities that need the YP libraries.
Problems with a number of programs were fixed by NBC. A disk crash then resulted in a loss of much of the YP software and a lot of the work had to be repeated by both NBC and APM. The new system has been tested and has been installed on the Pyramid (pyr-a). Completion of the implementation awaits the delivery of some missing source by Pyramid UK Ltd. A list of required software has been compiled and sent to Pyramid. In return we have provided a Pyramid client in the US with details of our YP implementation. However, we have not yet obtained the required sources.
NBC has installed the Usenet News software on the Pyramid Workcenter (nfs4). This includes the latest version (2.11) of the News reading software. As part of this work he also installed and implemented UKUUCP on nfs4. With the official closure of the VAX at the end of March, NBC moved the spool directories from vax-d to nfs4. All the software and maintenance scripts have been tested and the system is running well. Versions of rn and inews have been installed on Suns and the Pyramid. NBC (with IV) reconfigured the kernel on nfs4 as the machine was becoming very slow (due to lack of memory) now that it had to run News.
The uucp link to Warwick has successfully been established on nfs4, but there are problems with outgoing calls. NBC has investigated this and it requires software from Pyramid to fix. However, since ukuucp is not a Pyramid product help has been limited to informal channels.
NBC has pursued the possibility of RAL getting the News directly from Kent. At one stage this looked a distinct possibility. However, Kent have decided that they do not need any more feed sites at the moment so were not prepared to send us News directly. Fortunately, the service from Warwick has now improved since they get news from Kent via ftp. NBC has been talking to the people concerned at Warwick about RAL also obtaining News over ftp, which would be much quicker. It will also be needed if RAL is to supply News to other sites.
RAD was involved with NBC and IJJ in the setting up of a project with researchers in the Mechanical Engineering Department of the University of Sheffield to provide fast access to the Cray. "Fast" in this case is a relative term - the CFC has agreed to the funding of a 64 Kbit/sec leased line from RAL to Sheffield for the project. This is still considerably slower than ideal, but should be much better than JANET access, particularly as the use of a leased line means that there is freedom to use protocols (such as NFS) which would not be permitted over JANET. The CFC have funded this work as a trial of the fast batch mode of working.
It is desired to run the TCP/IP protocols and possibly NFS between the Sun in Sheffield and the Sun in Atlas. One way of doing this is to use X.25 to carry the IP using a standard encapsulation. To test running IP over X.25, IJJ established IP routing between the Suns jam and juniper. Although there is only a slow-speed X.25 link between the two machines, it gave some idea of what can be achieved.
To simulate the KiloStream link from Sheffield to RAL before the line is installed, IJJ arranged the loan of a Sun from SE Group (willow) and moved it to Atlas to be in close proximity to sam. IJJ installed a MCP board (Sun's high-speed comms controller) in willow and tested it. This board is necessary to drive the Kilostream line at 64 K bps.
In April, the KiloStream line from Sheffield to RAL came into operation. The researcher at Sheffield (2 P Wang) is making some use of the line, and this will increase as his knowledge of UNIX improves. To assist in this, NBC and IJJ invited Mr Wang for some training when he was visiting RAL in May. This comprised general UNIX usage, UNIX administration and debugging (NBC), and usage of the Cray Station software and NFS (IJJ).
RT has managed the preparations for upgrading to release 4.0 of the Sun OS. NBC assisted with the introduction and assessment of Sun OS4.0 by gathering and comparing performance measurements of ND and diskless NFS.
IJJ mounted SunOS 4.0 on two machines (louis and redwood), making them available for trial use by interested software developers at RAL before the public trial of SunOS 4.0 started. In addition, IJJ took a member of the operations staff (IV) through the procedure of installing SunOS 4.0 on redwood, so that operations had a taste of what to expect.
NJW was responsible for testing some of the IDUS software on a trial system running Sun OS 4.0. This included installing and testing Fortran, the NAG Fortran Library and Adobe Transcript 2.1, and testing Unipress Emacs. Part of the testing of Transcript was delayed due to the absence of DWB Text Processing software on the trial system.
IJJ looked at a new feature of SunOS 4.0, the Automounter. This allows dynamic mounting of filesystems upon demand, multiple sources (servers) for filesystems, un-mounting of filesystems upon an inactivity timeout, and the use of YP maps to hold configuration information. IJJ took the Automounter configuration that M Martin had been using and ran it on redwood. The result of this led to IJJ making recommendations on the suitability of the Automounter for use on the Suns in IDUS.
Eventually, SunOS 4.0 was installed in the Department on all except a few stand-alone machines, where software needing the earlier version was being run.
CCD have purchased a box to enable TCP/IP connections to be made to the 3090. This enables a direct connection to the Department's Suns, and has allowed experiments with 3270 emulation software and NFS. The former was provided by Simware, and a beta test version appeared to provide the necessary functionality (3270 screen in a SUN window). It is expensive, so a decision to purchase it has been postponed. In the meantime a SUN hardware product will also be tested (when SUN can deliver).
IJJ assisted CCD in the evaluation of VM/NFS on the IBM 3090. This consisted of running a subset of the EASE benchmarks from a Sun, redwood, to the 3090. IJJ was only able to run some of the tests because VM/NFS does not support directories. The VM/NFS server was in an unreliable state at this time, so IJJ was only able to run the test for one pass. Nevertheless, IJJ obtained performance figures for the most important NFS operations and the level of VM/NFS's adherence to the NFS protocol.
NBC, APM and RAD have been involved with changes to the IP addresses of machines on the ID ethernet.
APM helped IV, ASJ and ACD in bringing up the ID village with the new address scheme (he provided the new Hosts table). Tests were carried out with a few machines before the whole network was switched over.
It had been hoped to move to the new IP broadcast convention (all ones) at the same time but unforeseen problems with booting some machines means that this move has had to be delayed. NBC and RAD carried out some tests on a private ethernet and found some problems with rwho and with Suns broadcasting on zeroes at boot time. It is hoped that this problem will soon be overcome.
NJW has installed Adobe Transcript 2.1 on a Sun 3/50 running SUNOS 3.4 and has tested it in order to allow a decision to be made in favour of Adobe Transcript or that supplied by Sun.
She has also ported Transcript 2.1 to a SUN4 under SUNOS 4.0. This was so that we would be able to provide Transcript to grant holders who have bought SUN4s, only to find that SUN-supplied Transcript is not yet available for this architecture. It has subsequently become apparent that there are licensing problems involved with this; it is not clear whether these will be resolved before SUN themselves get around to providing a SUN4 version. In order to fix a known fault in the Adobe Transcript, NJK incorporated another program into the Transcript suite, to allow Sun raster files to be printed.
APM has ordered, installed and tested PC-NFS from Sun. The testing proved satisfactory and the Transputer Centre have ordered 6 more copies. APM has written a report on PC-NFS which because of its wider applicability has been circulated beyond the normal audience for Comms technical notes.
Before departing, Danny Smith (DFS - our Aussie) installed the THORN X.500 code that he had been working with on nfs3, along with elmdir, a version of the mail interface program that can interrogate the X.500 Directory. He also wrote up his year's work. This was semi-completed in the usual hurry before he left; he and RAD then spent some time (courtesy of IPSS) sending versions of the report backwards and forwards between RAL and Brisbane in order to get it in a suitable shape for release to the outside world. (This proved to be a very effective mode of working, due to the 10-hour time difference a version worked upon at RAL and sent by RAD before leaving for the night would be received by DFS early in the Australian morning, giving him a day to work on his next contribution and mail it back for RAD to pick up on his return in the morning.) The finished report was submitted by DFS to his University as the first year of a Master's degree; a somewhat expanded version is being published as a RAL Report.
For two weeks, NJW assisted in rebuilding work lost on nfs4, after a disk crash. This included the THORN, ISODE and ELM software which needed to be configured and rebuilt from sources. She has provided some support for this software. In February NJW did some timing tests of the THORN software. During these, the software stopped working. The fault was found to be with the THORN database. NJW rebuilt the database and then retested the software. NJW then handed over the responsibility of the THORN software to KG. The new distribution has improved support for images in the directory, for display under X and SunView. This has been tested on the samples provided and possible formats for local examples investigated. KG has also set up basic monitoring of outside access to our data.
KG has been exploring the existing ISODE sources and documentation. The newest was obtained as soon as available, and ISODE-5.0 was built, tested and installed on nfs2. All five volumes of the User's Manual were coaxed from the LaTeX source. Study commenced, in conjunction with other system documents. Further copies of the User's Manual were made and dispatched to the JNT for distribution.
ISODE-5.0 provides QUIPU, an alternative (and genuine X.500) directory service to THORN. KG built our first QUIPU Directory Service Agent (DSA) and rebuilt THORN to be self-contained on nfs2 and to run over the new version of ISODE. Old versions of THORN and ISODE were archived. KG brought the Informatics Department data in both QUIPU and THORN directories up to date, by means of scripts written to process an ascii text version of the departmental Directory Card. (One day it will be the other way around!).
KG installed a transport-level bridge (part of the ISODE distribution) on nfs2. This bridges ISO TPO packets between ethernet and X.25, so allows a QUIPU interface running on any workstation on the network to make "direct" calls to a remote DSA anywhere on JANET. This is not an essential feature for QUIPU functionality, since our local DSA can make the calls instead, but is desirable. THORN theoretically could make far greater use of the bridge: this was explored, but when initial attempts were unsuccessful (with no documentation and little help from support people at UCL), efforts were shelved pending the arrival of the next release of THORN.
A further supplied interface for the QUIPU directory, widget, was built and tested. As an interface, this is horrible, but will be useful as a model for future developments, demonstrating how to use the directory access library directly.
NBC, KG and RAD have been involved in an X.500 pilot project to provide a site-wide directory service. A joint bid with CCD was submitted to the JNT and has been accepted. Sun equipment is being supplied for the project.
In his role as Network Manager APM has continued to allocate Internet Addresses to new machines and keep the Hosts database up to date. APM has applied for and obtained a Class B network number from the DDN Network Information Center (in the US) to use for the RAL IP network and is currently engaged in working out a new addressing scheme.
An HP Ethernet Monitor was loaned to Oxford University for half a week during January. APM took the monitor up and set it up and retrieved it when they had finished with it.
During February APM was involved in some detective work trying to figure out who was involved in logging onto the Stellar as root. Although the problem was not completely resolved, the major cause for concern (apparent logins early in the morning) was traced to a faulty time-stamp operation in the log mechanism. LJR actively assisted with this.
APM was asked to help with the planning of computer communications wiring in the first floor offices acquired from HEP. His suggestions included rewiring the Thin Ethernet already on that floor, three RS232 connections for all offices on one side and some on the other side and two connections for the remaining offices as well as one Profs coax for each office. He also asked for extra mains sockets in the plastic trunking as experience has shown that the number put in earlier offices was never enough. He was also involved in setting up comms for people as they were forced into offices before the comms wiring was finished.
In January, NJW installed and tested X.25 and the yellow book on nfs1. This was needed for spooling to the Linotron.
Sunlink X.25 was a major part of NJW's work during April-June.
X.25 version 5.2 was eventually received from Sun. NJW installed this on kiwi, a 3/50 with an X.25 line and tested it successfully. She then installed the current release of the Coloured Book software and tested it with X.25 5.2. When it proved to be running correctly, she installed X.25 version 5.2 on nfs2, the mailserver. This version of X.25 was required for future trials of MHS(X.400).
NJW received an advanced copy of X.25 version 6.0 (the version required for use with SunOS 4.0). She installed this on redwood, a disced sun, and, after moving the X.25 line from G5l to G50, she tested it. Then she recompiled the current version of the Coloured Book software (1.1) for SunOS 4.0 in order to test it with version 6.0 of X.25. It was found that, while all the Coloured Book source files could be recompiled, the yellow book listening daemon (for incoming call requests), which was supplied as a binary, would not run on SunOS 4.0. NJW is now waiting for the new version of Coloured Books (version 2.0).
Some time ago, the Department provided a spooling system for the Reprographics Linotron phototypesetter. Unfortunately, both authors left and, with little demand for the service, the facility fell into disuse. Recently, it was found that the system had never provided a reliable service.
NBC and KG have been involved with trying to sort out recurrent problems. Their efforts have been hampered by there occasionally being genuine hardware faults which are not always easy to distinguish from software problems with the RAL spooling system.
Tests by Telecomms. On the line were inconclusive, and NBC is studying the documentation with a view to rebuilding the system on another Sun.
It has been discovered that Kodak sell a turnkey system, including a SUN, to handle Linotron spooling. A preliminary meeting showed that it does indeed meet most of the requirements. We are waiting to see whether Kodak are prepared to make the necessary mods to allow the identification of networked jobs (and also whether RAL has any money to buy it anyway).
Without having to do anything much, RAD was able to offer members of the Department a major technological breakthrough. For many years people who have tried to use dial-up terminals from home have been frustrated by the appalling quality of BT's lines, and the uselessness of the modems supplied by CCD Telecomms in coping with line noise. Telecomms finally admitted defeat with these modems and started a trial of decent error-correcting modems. RAD organised the acquisition of some of these for certain key members of the Department; they were a revelation, and make computing from home a positive pleasure. The success of these has stimulated a lot of interest, and demand is steadily increasing from other members of the Department for modems. All we need now is to be able to run X-windows over a modem link and we needn't bother to come into work at all any more!
MEC, AJL, WJH and SKC have formed the backbone of the section throughout the period. PDA and DSB were original members, who have since been transferred to other work. RAD, NBC and RET have played an active role and the first assessment also involved NJW, APM, Julian Gallop, Peter Kent and Dale Sutcliffe.
The first evaluation was to identify possible successors for the SUN3 as a middle of the range workhorse. This eliminated the Intel 80386 machines at the lower end of the market and the specialist graphics engines at the other end. The plan of campaign was to produce a shortlist of suppliers, to ask for a questionnaire to be completed, and to benchmark the most suitable machine in the supplier's range. The majority of these would use the Motorola 68020 processor.
By using the knowledge of the interested parties in the Department, by scouring the advertising press, and by visiting a Trade exhibition, a shortlist of 9 suppliers was produced. These were asked to complete the very comprehensive questionnaire, based on questions asked when the SUN3 was selected.
The benchmark tests were designed to assess three areas: the compilers and application packages, X-windows, and NFS. NBC provided technical assistance with running the NFS evaluation test suite on the Stellar machine (since this gave access to a non-SUN implementation), and investigated some of the problems that arose. NJW provided programming support for the IBM benchmark for EASE. DSB assisted in the running of the benchmarks on these machines, and later ran the analysis programs on the data obtained, to produce the figures for the final report. APM provided assistance in installing equipment and attaching machines to the Departmental ethernet. The suppliers were asked either to loan RAL a machine for the Section to benchmark or to run the RAL benchmarking software at their own sites. Only one chose the latter. In several cases, others loaned machines on more than one occasion as they improved software.
With one supplier providing neither questionnaire nor benchmarking machine and another deciding for itself that it was not suitable, the shortlist was reduced to 7. A detailed paper of information and benchmarking results was presented to CFTAG of 5 September, revised and updated for CFTAG of 16 November. These meetings concluded that -
The 80386 machines had been excluded from the first assessment. The plan was to ask the suppliers to complete a questionnaire (modified in the light of experience), to discuss this and their machine(s) with RAL, and to benchmark a loaned machine at RAL.
With an initial list of about 40 suppliers, it was clear that there were many more in this field than in the "mid-range" of the previous exercise. It being impractical to assess such a large number, only those running some form of Unix provided by the supplier were considered. This procedure reduced the number to 19, of which 8 agreed to take part. Subsequently, 9 of the rejected suppliers requested questionnaires. The revised questionnaire was sent out at the end of January with the request that it be completed and returned by the end of February. Within this period, only 5 questionnaires were returned, and, even more discouragingly, only 1 machine made available for benchmarking. The period was therefore extended to the end of April, allowing another three questionnaires to be returned and another two machines tested.
Although a large number of 80386 suppliers run a version of Unix, it is apparent that most of them target users wishing to use desktop publishing and database software rather than compilers and engineering packages. MS-DOS is still considered to be the prime operating system in most cases. The suppliers, in general, are vendors of hardware and basic software only, and refer the purchaser to software houses for anything else.
With this background, the technique of using a questionnaire proved less effective than before, since the suppliers were able to provide much less information about their machines, particularly in any depth of detail. There was, especially, little information on what packages were available on their hardware. This contrasts with the suppliers of the larger workstations, who provide catalogues.
The overall conclusion is that there is insufficient reason, at present, to add 80386-based machines to the Approved list.
CFTAG requested that the SUN3 replacement exercise be repeated, since there had been a considerable change in the marketplace in 12 months. Work started at the beginning of May, expecting the 68020 workstations to be replaced by machines using the 68030 or RISC processor. All those participants invited in the first exercise were again invited, together with a number of others. Various improvements had been made to the benchmarks, particularly the NFS ones (to overcome caching). It was also considered necessary to change the baseline SUN3/60 machine from 4Mb to 8Mb. IBM and Apollo declined to provide machines on loan, so the ones already at RAL were used (and in some cases old data was re-analysed, as little had changed). 15 Machines were handled this time, and a report was produced for CFTAG's September meeting. The DECstation and SPARCstation performed well, and have been recommended for Approval. The SUN 3/80 had most tests run with software Floating Point, and the error was discovered after the loaned machine had been returned. Steps are being taken to redo the tests. Several other machines showed impressive raw speed, but lack relevant software. In general, the expected workstation power has increased by a factor of 5 since last summer.
SKC has been attempting to improve the applications benchmark set, concentrated in the applications area, notably in the FE field, and, in particular, with the BIM2D, DFT and Nastran packages. In the AI/SE field, the idea of using ML was abandoned after consultation with members of SED. Instead ERIL is being considered. Other additions include tests for the Ada compiler and for GKS. However, suppliers could not provide Ada or GKS for benchmarking during the evaluation, and the other benchmarks are not yet in a form that can be used.
NBC has been making additions and improvements to the NFS evaluation test suite. Some extra functionality tests have been written and all the basic tests have been changed to allow easy parameterisation and to increase the default values, so that the programs are a more rigorous test of the newer, more powerful machines. The read test has been altered so that random reads can be specified in an attempt to defeat the NFS caching mechanisms. Also, two Yellow Pages tests have been written. New scripts for running the tests, and a set of awk scripts for analysing the results have been written.
NBC has also produced some test results for running the new suite of tests between various combinations of Sun clients and servers. Some of these showed peculiar performance drop-offs which need further investigation.
The team is responsible for the operation of the Departmental Service, including dumping, filestore management, fault finding and liaising with engineers. New machines have been installed. Departmental mailing lists are maintained, and the Mail/News system supported.
Records are kept of the location and configuration of the machines. The Summer Financial crisis demonstrated that more effort is required to keep these records up-to-date. General support includes the JANET Access Machine, which provides the basis of software supply to the community.
During the period, it was decided to divide the Computer Room in two, with half going to EBL. This involved much liaison with EBW, and the reorganisation of the machines. Eventually everything came back to normal. The work included the provision of a security lock for the door, and the installation of a fire-proof safe for the dump tapes.
The upgrade from version 3.4 to 4.0 was planned and undertaken. It included setting up trial systems so that users could ensure that their software still worked under the new version. Then all the servers and the individual disced Suns were converted. Several have had to remain at the previous revision, because of particular software problems. One of the difficulties of the whole exercise concerned the lack of a suitable X.25 from Sun. The technical problems involved are discussed earlier under Distributed Systems (see 2.3.1.5).
The role of Lab 11 was reviewed, and a new layout planned, which included the removal of old kit. New furniture has been purchased, and installed.
The team support both the users in-house and grant holders in the Universities. Queries are answered, and equipment moved. They are responsible for the transportation and setting up of the EASE Training set of Suns, a frequent occurrence which takes at least a day to perform. Advice on costings is supplied for Grant Applications. Liaison with Sun over orders and maintenance is undertaken. Documentation is also covered here, but this has had to take second place to all the other activities. The Introductory User Guide is in the process of being updated.
The changing nature of the Department's work has seen the need to dispose of a lot of kit. This includes VAXes, PERQs, SUN2s and the Orion. Most of the effort involves liaising with stores.
Since Jan Malone left in August 88, work on text processing has effectively been suspended until a replacement is found. However, RMK has kept track of reported problems, and has helped users whenever possible. She also incorporated user comments into Jan's WYSIWYG criteria list. Recently, RMK checked the code left by Jan Malone for positioning troff output on "new"-style LaserWriters, to try to solve the page position problem.
In August 88, RMK released RAL GKS 1.20 on the Primes. this included some new facilities to ease the conversion of existing programs from GINO-F to RAL GKS. This will be the last GKS release on the Primes. The corresponding release of the RAL GKS 1.20 Master source was done by RMK in September.
PLP has completed a GKS driver that generates PostScript output. The original driver was written by a Starlink VAX/VMS user. This was converted by Martin Prime to run on Unix. PLP first adapted this driver to run on the Prime and then went on to enhance the UNIX driver. New features like support for Landscape/Portrait page orientations, Encapsulated PostScript File Format and device-independent handling of hardware text were added. Existing code was in many cases expanded and sometimes completely rewritten in order to make fullest possible use of the graphic capabilities of the PostScript language. As a result, the driver now generates structured, efficient and compact PostScript code that can be either sent to a hardcopy device such as LaserWriter or (if it is in EPSF format) imported into another document.
RMK managed a collaboration between Trudy Watson and Tony Arnold (University of Manchester Computer Graphics Unit). The aim of this is to produce a RAL GKS driver for X on the SUN. A side-effect of this collaboration is that Tony Arnold has done much of the work of getting a working RAL GKS system under SunOS4.0. The driver works reasonably well on monochrome Suns, but is slow on colour workstations (because X is slow on colour machines). It will be included in a future release of RAL GKS.
RMK started planning the work needed for RAL GKS to be validated by the GKS Testing Service run by NCC. A number of enhancements and bug fixes are needed before the implementation will pass the tests.
PLP has put the RAL GKS Master Source under sccs (Unix's Source Code Control System). This was done in two stages. First, only a part of the Master Source was put under sccs. Experience of using sccs on such a large system was gained and remedies for sccs drawbacks were found before full implementation went ahead. Scripts for using sccs were written and a trial generation of RAL GKS release 1.15 for OS4.0 performed. Following this, PLP installed the updates for release 1.20 of the Master Source, and started work on 1.30 (adding in the PostScript and X drivers).
RMK has been working with CCD on updating the RAL GKS Guide. This has involved fairly major changes to the text of the manual, plus a lot of consistency checking.
RMK has produced NAG Graphics libraries for use on the SUN and the Pyramid. She has also written the necessary RAL-dependent user documentation to accompany the NAG Graphics Manual. Some changes had to be made to the GKS interface which was supplied by NAG: the interface used the Draft Standard of GKS, whereas RAL GKS is now an implementation of the International Standard; and some tailoring of defaults was needed so as to fit in well with RAL GKS. Details of these changes and suggestions for improvements have been passed back to NAG.
PDA is the Manager for the two Primes in the Department and also Resource Manager for the other four at RAL. The master machine, RL.PA, is scheduled to be closed by March 1990 and all the others must close before or at the same time since the software is controlled by RL.PA. The operating system has been frozen at Rev20.2 for some time and so, too, has the system software wherever possible. User support and system support, if necessary, are performed under contract by UMIST.
A regular meeting has been set up to highlight problem areas and report on the progress of closure procedures. PDA is heavily involved in the planning of the timesca1es to ensure that reasonable allowance is made for the tasks to be efficiently carried out. Tape decks have given cause for concern. After much pressure from PDA to Prime Area/District Managers on the subject, and many hours of our site engineer plus the field engineer working on the decks, both decks are now working efficiently.
The biggest problem of the closure seems to be the need to retrieve a lot of archive data in order to transfer it to other systems. There is no means of extracting this data other than perform a Retrieve, which is very labour intensive (files will be distributed over a number of tapes). Since there is a limited amount of disc space for the retrieved files to occupy, the potential problem is obvious. Discussions have been held with individuals to try to keep this activity down to a minimum.
PLP supervised a group of Trainee Programmers on their three-week C-project. This was a C fuzzy clock" program (presenting the time in the form almost noon). Preparing for this, PLP completed the project himself and also put together a brief introduction to Sun workstations. He also helped in setting-up the EASE training machines, which were at Trainees disposal (mercy?) for the entire project. PLP and RET conducted the acceptance tests of the Trainee code, which showed a number of problems. The exercise went very well, and enabled the trainees to gain experience with workstations (previously this training was carried out using UTS on the IBM).
During September and October 88, RMK and PLP worked on software for two demonstrations: a graphics demo to members of CFC and a text processing demo for the BCS Displays Group Meeting on Interactive Documents. RMK gave these demonstrations. IV was involved in the setting up of the SUN network and due to people being out at the last moment, was roped in to deal with the final panic stage of the installation.
Alvey funding for Infrastructure ended officially on March 31 1989, but the final clearup continued into the next financial year. The main effort has therefore been in bringing the contracts to a close. The GEC machines were closed in April 1988, and most found their way to Daresbury. A final Site Managers meeting was held in October. Agreement was reached with Contracts on the rules for disposal of equipment, and the VAXes have been offered to sites. So far, only EdAI have refused the offer!
RET and Peter Hemmings have been involved in the reconciliation of the contracts. Thanks to Peter's efforts, all the bills for the first 9 months of the year were received on time. The final quarter cannot be paid until after April 1 1989, and there are still three bills outstanding.
The Project Meeting has been formally terminated.
The Departmental Unix Service is managed by a small committee. This receives reports from the Service Meeting and from the various technical projects associated with the service. This year has seen the phasing out of the SUN2s, considerations of security and the refurbishment of Lab 11 and the Computer Room. Attempts were made to identify the money and effort used by the service, but the Laboratory finance system was not up to the need to redistribute this back to funded projects. A more labour intensive scheme had therefore had to be adopted. It is reported that MSA will solve this, and many other problems.
The major concerns of users have been the continued unreliability of pyr-a and the state of the laser printers. The major upgrade to SunOS was successfully completed. Improvements to the file servers continue, and a loan pool of shoebox discs made available for solving short-term problems.
Following the scare caused by the USA Unix virus, Division Heads decided that the Department should take a much more serious view of Security. RET was appointed Security Officer, and is charged with preparing and implementing a Policy. LJR helped initially, and PDA took over when LJR transferred.
RET prepared a paper for Division Heads, looking at the implications of the strategy set out in Ken Hartley's original paper. Part of the requirement centred on the production of a Code of Practice for the Department, which every member would be expected to abide by. LJR prepared the first few drafts of this, and the job is being completed by PDA. Meanwhile, Division Heads asked RET to carry out a proper Risk Analysis. This led to a considerable amount of work, including literature searches. The resulting paper attempted to quantify the threats and countermeasures for a whole range of activities, from fire to hacking. This led to a complete policy document which will affect all the Department. To date, this document is still being discussed.
Initial steps are being taken to improve the security of the Department. LJR has checked the root to root access on machines. Unknown users that have been added to local passwd files have also been checked. Sysdiag was passworded on each machine and then later removed on most machines. No entries are to be without passwords. The combination lock on the Machine Room is now operational. A fire-proof safe has been purchased to hold dump tapes. Offices will normally be kept locked outside prime shift. Lists of who knows what passwords will be maintained.
PDA has been on the first of the Civil Service courses on Security. The next course will cover the CRAMM Risk Analysis software, which has been ordered but has not yet arrived.
A project on Prime A, concerning a forecasting model for SERC grants has been extended. PDA works with Jim Hailstone, who is currently consulting for Swindon. She has been heavily involved in planning and preparing programs for use when the data is available. A lot of learning has taken place, in preparation for moving the prototype from Prime A to another machine. PDA also prepared to receive grant data for the Modelling project. Shortly, Swindon should be able to contribute data on a regular basis, and the model should begin to give relevant results.
With the closing down of the Prime service, UMIST's support contract has been changed to provide support for ECFE-related activities. GAL chairs the management meeting, with RET in attendance. As part of this work, UMIST have been preparing a number of documents on Introduction to SUN, Systems Administration, and Primos to SUNOS conversion. RT and KML have assisted in reviewing these.
RAD is involved with the ARGOSI Esprit project. The Commission decided to accept the proposal, but contingent upon a budget cut of about 60%. This was fairly standard for all proposals, and was caused by a gross underestimate on the part of the Commission as to how many projects would require funding from a fixed-size pot of money.
However, the Commission also required ARGOSI to absorb a project in a vaguely similar area, and restructure to accommodate the partners of the other project. RAD and DAD spent an entertaining, if somewhat gruelling, few days in Brussels and in Paris persuading the other partners that proposals for doing this worked out by RAD/DAD/JRG on the back of several envelopes would do the trick. The revised proposals seemed to satisfy the Commission.
A number of promises of a contract were made by the European Commission, but during January and February nothing materialised. Then, at the beginning of March, we were summoned to Brussels to re-re-re-discuss the proposal. Having negotiated this hurdle, we got a provisional go-ahead and were able to hold a project kick-off meeting at Cosener's House. As by that stage we hadn't got a written contract, nobody was too keen to really get started; since then the Contract has come through.
This led naturally to the next problem - finding someone to do the work. We went out to recruitment for an RA (in a joint effort with Ken Robinson who needs a small army to work on his projects). Although Ken was quite successful (at least in the sense of finding people he'd like to offer a job to), for ARGOSI we had no luck. RAD was on the Board for the interviewing, and it appeared that the problem is that the RA advertisement was (as always) geared to junior, relatively inexperienced people. What ARGOSI needs is people who have some experience plus an interest in ISO Standards (which in itself requires the sort of philosophical outlook that comes with experience!). We are now attempting to raise our sights a bit, which of course involves the usual battle with Personnel.
Technical work finally started on this project on April 1. Two strands of activity are underway at present. The first is to start classifying existing graphical applications in terms of their graphics and networking requirements. This involves going out and visiting providers and users of software, and asking them detailed questions. (This is a similar idea to the proposed HEI visits, except that ARGOSI got there first!). To date RAD has been involved in these visits along with David Duce, Julian Gallop and Bob Maybury.
The project held a one-day progress meeting in Milan in May to discuss progress so far. This was very useful, with some genuine work being done. Since then the RAL team managed to make 3 visits (organising them is not easy, a lesson to be taken on board for the HEI visits.) The most interesting was one to the Met Office in Bracknell, where they actually use GKS for some applications.
The second task is to set up basic X.25 connectivity between partner sites, with a view to its use for a demonstrator of ISO graphics + networking across Europe. RAD is task leader for this. Although PDNs will probably be used for the connectivity, there is interest in using the facilities promised by the COSINE project. RAD attended the first COSINE Users' Group meeting in Brussels to get an introduction to this.
The inability to recruit RA staff to the project makes progress very difficult, with RAD, JRG, RM and DAD having to reschedule some of their other work to devote time to ARGOSI.
RAD (with J R Gallop) spent a considerable amount of time advising Central Office on the computer procurement aspects of the proposed IRC at Glasgow. Christened the Engineering Design Research Centre (EDRC), this facility was proposing a $500,000 purchase of workstations, servers and fast networking to hold it all together. Apart from the relatively straightforward technical issues, the whole saga proved to be somewhat political in nature. The role of RAD and JRG was to satisfy themselves (and hence Central Office) that the equipment proposed was suitable for the job (although it was not clear at the time exactly what the job would be), and that the procurement process was "fair". All this proved to be a complex story; suffice it to say that after a number of visits, reports etc a mutually acceptable outcome seems to have been achieved.
The Group is involved in a considerable number of meetings, both departmental, laboratory-wide and external. The following indicates who goes to which.
EASE Workstation Assessment RAL-89-017: RET + 9
As the Division has a new structure these objectives are collected together for the two new Groups which had been TSG during the report period.
Fran M Childs (FMC) departed 31 March 1989
The objectives of this Group are to provide those services for Informatics Department which it is sensible to centralise. Some are provided on behalf of the whole Department, eg EASE services, Departmental Administration Officer, others for specific projects, ESPRIT project support.
The project formally ended in March 1989. The last six months of this project were fairly quiet. The Site Managers attended a final meeting as Cosener's House on 20 October 1988. There were no technical issues outstanding and it only remained to dispose of the equipment, with the sites being allowed the option of retention either as gifts or loans according to the age and value of the equipment.
All of the sites except Newcastle were visited and satisfactory reconcilliations were provisionally agreed pending the submission of final bills. All but two sites (Oxford and Imperial) have submitted final bills.
The Alvey Mail service closed on 31 December 1988 with much lamentation from a number of users. Central Computing Division kept the machine running on an informal basis for several weeks after the formal closure, forwarding incoming mail and sending notes about the closure back to the senders of such mail. PJH spent a lot of time chasing the DTI to give the machine to SERC and a letter was eventually received from them in August.
PJH has maintained the programs used by the DAO to produce the departmental finance paper. A certain amount of maintenance work is required, mainly when the financial year changes.
There had been a problem drawing the boxes around tables, which had persisted from the early days of the project. This was believed to be due to an obscure bug in the Unix utility tbl used in conjunction with troff. The output routines were changed to use a series of user defined Postscript macros. Some other improvements to the layout were made at the same time.
Generally this project requires hardly any changes to be made as a result of changes in the operating system CMS. However, when CMS 5.5 was introduced at the end of April it contained a bug, believed still present, which prevented the monthly updates from completing correctly. It was therefore necessary to circumvent this problem.
PJH acted as one of the course tutors at a Summer School held at the end of September 1988 in Cosener's House, being responsible for one of six tutor groups. PJH repeated this role at another Summer School in July 1989.
PJH has been developing a project to meet a wide range of Departmental requirements using mailing list data. The aim of this project is to maintain information about Departmental contacts. Subject interests, committee memberships and newsletter distribution requirements are recorded. The Department already has two large mailing lists which have been used to establish the initial data. Facilities provided by this project already include the ability to display lists of users and view details of any selected user, then define subsets within the list. It is possible to produce individually addressed copies of a standard letter to members of the subset and to produce typed address labels.
PJH was asked to investigate the usage of the SUN Fileservers installed by the ECFE at six universities (Lancaster, Leicester, Oxford, Swansea, UMIST and Warwick). Each site was visited; it was found that some sites already had extensive experience of UNIX and using SUNs and that there had been no major problems. Some sites, however, did not have such expertise locally and these visits helped to clarify the amount of support needed from SERC when such installations take place.
Essex, Leeds and Strathclyde are due to receive servers in a subsequent round and PJH and GAL have carried out preliminary visits to discuss aspects of the installation.
PJH has become the Informatics representative for the use of these facilities. Work has begun on tidying up the user accounts to correspond with current projects.
Sheila Davidson (SGD) assumed responsibility for the production of the Engineering Computing Newsletter, taking over from Fran Childs (FMC), upon her arrival at the beginning of April. The decision was taken to give priority to increasing the frequency of production from quarterly to monthly and to achieve a significant increase in the circulation.
To reach these objectives, a stable version of Framemaker had to be obtained, mounted and SGD trained in it's use. Production schedules had to be agreed with Reprographics, material had to be obtained from the community and a distribution network set up with the HEI community to supplement the standard mail list. Overall, these objectives have been met, the circulation nearly double what it was twelve months ago, and with the Newsletter attracting favourable comment from many diverse sources. These are viewed as initial objectives and it is intended to improve both the presentation and content of the Newsletter in the future.
Susan Hilton (SCH) joined the Group at the beginning of July and has assumed responsibility for the production of other publicity material about EASE. Initially the majority of her time has been taken up trying to understand the background to the formulation of the policy, prior to producing a glossy for general circulation. She is also attempting to ensure that the various Secretariats making up the Central Office at Swindon are kept fully aware of the various aspects of the EASE Programme and that such material as we have is readily available to the user community.
SCH has commenced revising the EASE definition document and the associated schedules. Work has also started on updating the Departmental handouts.
Education and Awareness has been defined as one of the high priority areas of the EASE Programme. Unfortunately staffing this activity has proved to be a major problem. Professor H Rosenbrock was appointed as Education Co-ordinator at the beginning of the year but resigned at the end of June 1989 on a matter of principle and interpretation of policy, a serious loss, especially as a replacement has not been identified.
John Smith (JWTS) joined the Section at the beginning of April and, with the assistance of David Lomas at UMIST, provides the main effort.
The following events have been organised during the year by GAL, with assistance from JWTS, SGD, HVJ and David Lomas (UMIST):
| Workshop | Graphics Standards | 18-19 October, 1988 | Manchester University | 55 |
| Course | PostScript | 2 November 1988 | RAL | 16 |
| Seminar | AIAI Tools | 13 December 1988 | RAL | 60 |
| Seminar | Network Services | 23 February 1989 | RAL | 58 |
| Seminar | X-Windows | 22 March 1989 | RAL | 100 |
| Display | Computer Graphics Metafile | 29 March 1989 | Manchester University | 30 |
| Workshop | Data Exchange | 11-12 April 1989 | Leeds University | 60 |
| Seminar | Kent Tools | 1 June 1989 | RAL | 50 |
| Course | Server-System Administration | 17 August 1989 | RAL | 6 |
The other major activity of the Education and Awareness team has been to initiate a programme of visits to the Engineering Departments of all HEIs in the UK. One objective of this activity is to attempt to identify the base level from which to measure the impact and effectiveness of the EASE policy. A number of pilot visits have been undertaken, to Strathclyde, Warwick, Bristol Polytechnic, UMIST, and from these a questionnaire has been developed. This will be used as the basis of the information gathering process.
Carol Barnes has continued to provide a day-to-day administrative service for all staff in the Department. In addition, she has been involved in major office accommodation re-planning, refurbishment and fitting of new offices, and the organisation of many office moves.
Another important task has been to provide Division Heads with regular expenditure and effort usage summaries to enable close monitoring and matching to allocations. This has been particularly important over the last year with the closing down of the Alvey Programme and the monitoring of expenditure at the end of the financial year and in the current financial climate.
The past year has seen a growth in the responsibilities of the DAO and it is now essential that a part-time AO be recruited to provide assistance.
CB is also Secretary of the Informatics Division Heads Meeting, and of the SERC Suggestions Scheme Local Awards Committee.
GAL has acted as Secretary to the Computing Facilities Technical Advisory Group (CFTAG), and as a consequence has been a member of the Engineering Computing Facilities Executive (ECFE) along with KFH and RET. This has resulted in a significant number of follow-up actions after the meetings.
GAL continued to provide management support to the project. This has involved the organisation of two full reviews, in December 1988 and June 1989, plus all the routine administration, report production, and contract renewal. In addition, preparations have commenced for the Esprit conference and exhibition at Brussels in November and for a follow-on project funded by EVEREST II.
DCSD is to be involved in the pilot project and GAL was invited to act as Secretary to the organising committee. KFH is the co-ordinator of the Informatics participants, with the Department Head as co-ordinator-in-chief of the SERC Pilot Scheme.
World Computer Congress, San Francisco, August/September '89, GAL
At the start of the report period the Alvey Unit contained the following staff:
When it was created one of the objectives of the Alvey Unit was to disappear by March 1989. This it duly did. Martin Dunn transferred to Central Computing Department to work on Administrative computing. Jill Cheney also moved to Central Computing Department. Terry Mawby transferred to the Transputer Co-ordination Unit. Keith Dancey transferred into Design Division to work on the Ecstasy project. At the end of March, Sheila Davidson and John Smith joined Management Services Group of DCSD.
The sole survivor, Mike Russell, became the founder member of the IED Unit. In August he was joined by Tony Lubbock (AJKL), who continued to work for Adrian Wheldon, himself now nominally at Swindon. Both work for the Systems Engineering Director in DTI's Information Engineering Directorate.
Thus at the end of the report period the IED Unit consists of:
The objectives of the Alvey Unit were clear, if impossible. They were to maintain all the invaluable services which Informatics Department provided to the Alvey IKBS, Software Engineering and HCI Directors, and their successors in IED. In the meantime the budget was cut in half and the staff had to be reduced to zero by 31 March 1989. The hope was that by then SERC and DTI would have agreed how these or similar activities should be continued for what became IEATP. In the event the decision was taken that none of these activities should continue.
On the other hand, it was agreed that Mike Russell should put his experience to use in organising the monitoring of the Systems Engineering projects in IEATP. The objective of the IED Unit is therefore to provide a home for Informatics Department staff who spend much of their time at DTI's Kingsgate House, and whose duties do not fit naturally into the structure of Informatics Department.
It is easy to imagine the problems of trying to maintain services for as long as possible whilst the expertise and experience necessary was evaporating at an increasing rate. It is greatly to the credit of all the staff involved that they turned their hands to many new tasks knowing that they would probably only be doing them for a short time.
During this period all three mailshots continued to appear at regular intervals. Monitoring officer reports were chased and consultants paid. The IKBS Bulletin Board was maintained. Deliverables were chased and published regularly. Workshops were organised, the bills paid and reports obtained.
Increasingly effort was put into persuading IED to take over some of these activities or even show much interest in the Alvey Programme. In fact, all attention was focused on defining what the new Programme should be like in spite of the fact that a significant proportion of the Alvey projects were still running. This seems like a very sad end to what had started off with such high hopes and had, in fact, made substantial progress towards vital strategic objectives.
The withdrawal of staff to continue the monitoring of the Alvey programme contributed to the termination of publication of the abstracts of Alvey project deliverables. These abstracts constituted the prime method for dissemination of information about the results and products of the programme into the public domain. Hence, one of the prime objectives of the Alvey programme - the dissemination of the research results - was prevented by default.
The abstracts also enabled quantitative assessment of the performance of the projects to be made for programme evaluation purposes. The requests for further information about the research results prompted by these abstracts provided additional data for evaluation purposes.
The main objective for Mike Russell was to set up and operate a Monitoring Officer (MO) programme for the new SERC/DTI IT programme. By the end of this reporting period monitoring officers have been selected for >90% of the approved projects in the IEATP programme in Systems Engineering.
These MO reports will be used to judge progress of the 60-odd new projects and the several remaining Alvey projects. To assist the Director of Systems Engineering, Mike Russell is required to analyse the MO reports and prepare a summary indicating problems and achievements. On the basis that the projects have been selected for technical excellence and that adequate support has been provided, then the main uncertainties relate to the previously unspecified project staff. The MO reports will pay special attention to the nature of project staff interactions in addition to the quality of the project deliverables.
The MO reports will form a coherent record of project performance using the deliverables as a quantifiable parameter of achievement. In the event of a project failing to perform satisfactorily the MO reports will be used to justify any actions taken by the Directorate. Good communications between team members is very important and MOs should attempt to identify where in-house staff training would benefit the project.
Individual research has been conducted by Mike Russell into previously unexplained defects in natural mica crystals. Strong circumstantial evidence has been found for a new solid-state recording process of exceptional sensitivity. It is possible that this new recording process might have practical implications in the area of data storage and the feasibility of duplicating the process in the laboratory is being examined by a RAL/Shanghai (China) collaboration. Recent detailed study of the (supposed) positron tracks in mica has shown that the new recording process responds to both elastic and inelastic lattice perturbations. Recording processes based on inelastic scattering, such as caused by ionization, are well known. However, the observation that the recording process in mica can respond to elastic events is a significant discovery with wide ranging implications.
If this interpretation is correct then it is predicted that very directional, high energy, soliton-like entities may be created in nuclear interactions within a crystal. A RAL/Frankfurt/Brighton collaboration has been formed, to seek experimental evidence for Supra-Ballistic Phonons, to investigate this exciting possibility.
F M Russell, Nuc Tracks Radiat. Meas., Vol 15, 1-4, 41-44 (1988). Identification and selection criteria for charged lepton tracks in mica.