The Annual Report covers the year ending September 1988. 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 and its objectives.
The Divisional Structure in September 1987 was:
The funding for the first two Divisions came largely from the Alvey Programme. Software and Knowledge Engineering had research projects in the IKBS and SE areas and provided technical support to Alvey Directors in these areas.
The Infrastructure Division supported a hardware and software infrastructure for the IKBS, SE and MMI areas. This started in 1984 providing multi-user system support via a distributed set of GEC Series 63 and SYSTIME VAX systems. More recently support has also been provided for SUN systems. This Division also provided coordination and support in the areas of IKBS, MMI and SE for the relevant Alvey Directors.
Engineering Computing Division provided support for the Engineering Board's Computing Facilities Committee. This was a range of activities from application to hardware support aimed at providing support for research funded by the Committees of the Engineering Board.
The Alvey programme will be substantially complete by April 1989 and the last year has been one of running down the Alvey work as the programme comes to an end. In consequence, the Divisional structure became rather unbalanced. For this reason and several others, a new structure was introduced in July 1988 which is oriented more towards function than funding. The new structure has four Divisions:
The major changes were to split Engineering Computing into its functional components, to move the Human Computer Interaction work of the Infrastructure Division into Design and to move the Image Processing parts into Computational Modelling.
The organograms at the start and end of the year are given as Appendices. The format of this report is largely that of the structure at the beginning of the year.
After the enforced changes in structure and management which occurred during the previous reporting period, this period was relatively stable with DAD and GAR leading the combined Software and Knowledge Engineering Group from September 1987 -May 1988 when RWW returned from his American sabbatical.
In the July 88 reorganisation of Informatics Department the synergy between SE and KE was recognised more formally when SKE Group became Systems Engineering Division.
Visitors
Sandwich Student
The main activities in the Software Engineering area are:
Development work is funded directly from Millbank to SERC/RAL. The research work is funded differently. This either follows normal Alvey rules, or goes through the peer review system of IEC/CSSC. At any instant the aim is to have more than one project underway with successors in the pipeline. Systems Engineering aims to have both Alvey and CSSC funding simultaneously. It is hoped that funding via ESPRIT can also be obtained.
The section undertakes specialist infrastructure work for the Alvey SE programme. The way of working that seems most satisfactory is for the section to work alongside a specific Alvey project to provide a piece of infrastructure needed by that project, but to do so in such a way that it will be of use to the whole community. An example of this is the project to provide a yacc-like tool for Standard ML.
The aim of this project is to produce a yacc-like parser generator written in SML which produces output code in SML. This project is being undertaken on behalf of the Alvey FORSITE project.
FORSITE is a collaborative project between the University of Oxford, Racal ITD Ltd, the University of Surrey and System Designers Plc. The FORSITE project is producing an environment to support the development of formal specifications in the Z and CSP notations. Both are products of Oxford's Programming Research Group. The main components of the environment are a multi-font WYSIWYG editor for mathematical texts and a type checker for Z. The latter is written in SML.
An SML/Yacc system is needed by Forsite to provide a Z parser that will integrate neatly with the other components of the environment.
The specification for the system was drawn up by Michael Hedland and Brian Matthews (BMM). Following Michael Hedland's return to Sweden, Stuart Robinson (SKR) has become involved in this project. By December 1987, the system was in a sufficiently good state as a parser for LR grammars to be a practical tool. However the extension to full LALR parser generation proved technically very difficult and computationally very expensive.
During the winter period, the first version of the LALR lookaheads generation was implemented, the final computational problem to be overcome in producing LALR parsers. The algorithm given in Aho Sethi and Ullmann's book on parser construction was implemented. This proved computationally very expensive in ML so some time was spent optimising the code and doing as much precomputation as possible to get the run time to a reasonable level. Another lookahead algorithm as given by Pennello and DeRemer was implemented when further improvement in the ASU algorithm seemed impossible. This was claimed to be a better algorithm, but the efficiency of the ASU version has not been achieved. The project is now winding down and a final product is scheduled for October 1988. There is some tidying up of the code to be done. Example parsers will also be produced and tested including a parser for the FORSITE Z grammar.
The new methods developed in the project are being written up for publication, and a users manual is being written.
Working with functional languages has suggested a method for the production of parsers which is based on parser composing operations. This is completely different to the traditional table-based methods used in this project to date. Some exploratory work is being done in this area before the project terminates.
In February 1987, the Laboratory for the Foundations of Computer Science at Edinburgh University asked if the Division would collaborate with them in developing graphical user interfaces for software tools being constructed.
Edinburgh are developing a number of tools for supporting formal software development methods that would benefit greatly from graphical user interfaces. One such tool, the Concurrency Workbench, seemed particularly appropriate for a pilot study. The Concurrency Workbench is a suite of tools for manipulating and analysing concurrent systems described in their CCS (Calculus for Communication Systems) notation. The first prototype, with a simple textual interface, is now completed, and the aim of the collaboration work is to develop a graphical user interface for these tools.
HCI section of the Infrastructure Group are providing the people for the project, funding comes from the Alvey Software Engineering programme. Technical progress is reported under the HCI entry.
Theorem provers are software tools to aid in the process of proving theorems. In the software engineering context we are primarily concerned with theorems about programs, for example that a particular program has a particular property.
Some 30 copies of the Cambridge LCF/Rutherford SML system (described in the 1986/87 Annual Report) have now been distributed and more are in the pipeline. Much interest was aroused by Larry Paulson's (Cambridge) book on LCF. Larry Paulson has produced a SUN version of the system which is about ready for distribution by RAL.
The group also acts as UK distribution point for the Alvey SE Infrastructure for the Boyer-Moore theorem prover from the University of Texas.
The overall software engineering research them is Quality Certification of Software Product. Within this theme, the main research interests of the group are in formal specification, the development and application of machine-assisted proof systems and the role of formal reasoning in software development.
In September 1987 David Duce (DAD) held a research grant from SERC's Computing Science Sub Committee entitled Theoretical Studies of Emerging Computer Graphics Standards.
DAD and Mike Parsons (MSP) worked on the project during the year, but progress was impeded by DAD assuming the role of SKE Group Leader and the diversion created by the Esprit II programme.
Three papers were prepared for the Eurographics GKS Review meeting in 1987. One covered proposals for improved input facilities in GKS. The second was a specification of the polyline, polymarker and fill area output primitive of GKS in the framework reported in the Arnold, Duce, Reynolds paper at Eurographics 87. The third paper recorded the lessons learnt from the GKS specification work so far and areas in which the GKS definition could be improved. The papers were well received at the Workshop and it is noticeable that, with the exception of the USA, there is now considerable sympathy in ISO for the use of formal methods in the development and presentation of standards.
The GKS Review Workshop produced many interesting ideas for directions in which GKS could be simplified and at the same time provide richer facilities. A good example of this was the idea to replace the GKS segment store with a primitive store from which primitives could be selected using a name set/filter mechanism.
Many suggestions were made concerning output primitives: the current set of primitives in GKS is unsatisfactory for several reasons.
MSP subsequently looked at more fundamental and structured ways of defining primitives. The underlying philosophy is that the set of output primitives will never be fixed, people will always want to add new primitives to do a specific job, or to exploit particular hardware. To allow for this extensibility, a proposed framework approach for graphical primitives is proposed: as long as new primitives can be shown to fit the framework model, they may be considered to be primitives. The case of poly primitives has been looked at in some detail and a working paper using a functional specification style has been produced. This specification has been partially implemented using Miranda (a functional language developed by Turner at the University of Kent) and Postscript.
The subject of Reference Models for Computer Graphics Standards reared its head again during the year. It is a subject on which everyone has an opinion (and different expectations!), yet few are prepared to put pen to paper and produce drafts for discussion. DAD produced a document for discussion by BSI in February 1988. In revised form this was transmitted to a special SC24 advisory group meeting on future policy in April 1988. This meeting came out with a recommendation that the BSI reference model should be developed further, together with a recommendation that a Components/Framework model should be explored as a basis of the development and management of future graphics standards. At a subsequent meeting of the BSI Reference Model Group, it became clear that the BSI Reference Model could be integrated with a Component/Framework model, essentially by recognising a connection between abstract data types and components. A revised paper was produced which formed part of the BSI input to the SC24/WG1 meeting in July 1988.
The Component/Framework model was well-received by European nations, but met with considerable opposition from the USA. Further work to refine this model within Europe is being planned.
MSP did some joint reference model work with Clive Ruggles at Leicester University. This started from the recognition that in a graphic system there are two dimensions in which graphics data can be reified (made more concrete). The first is a data reification, the second a graphical reification. A working paper describing their ideas went through several iterations.
DAD and MSP both participated in the BSI Computer Graphics Panel and in the GKS-3D international review. DAD attended two GKS Review meetings and the SC24 meeting in June/July 1988. DAD has recently been appointed document editor of the GKS Review.
Jeremy Dick's (AJJD) research continues in the field of Automated Equational Reasoning 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.
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 the 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.
AJJD and John Kalmus (JRK) have continued the development of ERIL.
Several more termination orderings in ERIL have now been successfully implemented. These include a version of the Knuth-Bendix Ordering that allows the user to assign weight to function symbols (userKBO); three versions of the Recursive Path Ordering, one which requires the user to assign the function precedence in advance (userRPO), another which automatically determines the appropriate precedence (RPO) and another which also determines appropriate function status (RPOS).
Initial experiments have been made with AC completion in ERIL.
A new version of ERIL is currently being designed in collaboration with the Royal Holloway and Bedford New College (RHBNC).
A joint paper of AJJD and JRK with Dr Ursula Martin (Royal Holloway and Bedford New College) is now complete, and awaits submission to an appropriate refereed journal.
AJJD's paper on structures for the representation of rewrite rules has been suspended pending the results of an MSc project that he is supervising on the subject at Imperial College.
A major task for AJJD and JRK during the period was the production of the ERIL users' manual, of which preliminary copies have been distributed for general comment. It has been registered as a laboratory report (RAL-88-055), and will be issued in September 1988.
Another RAL report on the prolog module structure used in the new version of ERIL is planned. AJJD's recently completed thesis will also be issued as an RAL report.
AJJD and JRK have held extensive collaborations with RHBNC, having held 5 one or two day meetings, which have at times included Dr Muffy Thomas from Glasgow and Stuart Matthews from Hatfield Polytechnic.
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 applied to join the project from April 1986 and, after some delay, received its grant allocation at the beginning of July 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.
During the period September-December 1987, the RAL work concentrated on the design of the theory store for the formal reasoning IPSE (FRIPSE) being designed and built by theme C. FRIPSE is split into a left hand side, providing specification language support, and the right hand side, providing theorem power support. Initially it was intended that the left hand side and right hand side would communicate through the theory store, but this was eventually felt to be too complex an interface and a simpler interface was designed, and the theory store has been accommodated in the right hand side.
The requirements analysis work for the theory store is written up in an Alvey deliverable Theory Store Analysis.
Following a reorganisation of the IPSE 2.5 consortium in the theme B area, RAL has been charged with the design and implementation of a minimal left hand side for use by theme C whilst the fuller version from theme B is awaited.
Work began in January on the specification of the minimal LHS component of Theme C - a support system for VDM (as a design methodology as well as a specification language), whose main purpose in life is to provide interesting proof obligations for the RHS (theorem prover/theory store) to work on. Originally, the keyword for this design was minimal; however, in consideration of the present status of the theme responsible for a full-blown support system, RAL have come under some pressure to make the design less minimal.
Design and presentation of a preliminary formal specification of the LHS has been the major task over the first half of 1988. A VDM data model has been developed to encapsulate the construction of completed design developments, in accord with the process laid out in Jones 86 of specifications of varying degrees of abstraction linked by reification information (how one specification can be formally viewed as an implementation of another, both through reification of data models and through redesign of operations and functions upon them). This model went through numerous iterations.
Several working documents which were published in June, were amalgamated into the preliminary specification which was reviewed and has been released as a project document.
Since then, effort has been concentrated upon one of the largest gaps in the preliminary specification, namely how operation decomposition could be handled. The main task here is the design of a method for annotating statements in a Hoare-like fashion. A working document on this should be completed soon.
BR and JCB have also contributed to the rest of the theme C work, participating in informal discussions as well as more formal document reviews.
BR co-authored a paper with Paul Taylor (Edinburgh) on the Interactive Proof Editor (IPE the subject of BR's PhD thesis). The paper was presented by Paul at a workshop on Hardware Verification in Canada in June.
Jim Woodcock and BR initiated an attempt to use the IPE to prove a Z proof obligation. It was interesting that the proper structure of the proof obligation only became apparent during the construction of the proof. To BR's relief, the IPE coped quite well with this style of development and it is intended to write up the exercise.
February saw the start of BR and JCB's involvement with BSI VDM. The intention is to make RAL's specification follow the BSI proto-standard as closely as possible, hence the involvement in the BSI committee. As work progressed through the following months, errors in the proto-standard were found and it became clear that certain choices that had previously seemed arbitrary could have complicating consequences on the theme C work. Thus contributions to these meetings have increased. BR has been persuaded to take over document control of the main proto-standard document from Cliff Jones (Manchester); so far, this has required little effort - the main changes to the BSI proto-standard so far seem to have arisen from the IPSE 2.5 work.
Jim Woodcock (JCPW) continued his researches into the area of state-based concurrency. In collaboration with He Jifeng (PRG, Oxford) JCPW is working on the combination of the Z notation with CSP, the basic idea being to use Z to describe states, and CSP to describe concurrency and interaction. Jifeng has now fixed the theoretical underpinnings and has produced a complete set of rules for data refinement. JCPW has produced a case study which uses the new techniques. Part of the work was presented at the Z users meeting in Oxford in December, and at a series of lectures at IBM UK Laboratories, Hursley Park.
Working with material from the CICS formalisation project at IBM, JCPW has developed some case studies in specification and refinement. Current work concerns proof rules for event refinement, based on Jacob's rules for local replacement. This seems promising. The second approach to the description of state-based concurrency is Cliff Jones' extension of VDM with rely and guarantee conditions. Little work has been done using this extension since Jones published his D.Phil thesis in 1981. Working with a colleague at GEC Telecommunications, JCPW has produced the first draft of a report which investigates the applicability of the technique. In his thesis, Jones concentrated on operation decomposition, and in particular parallel decomposition. The report takes a more general view of development, where a development step may involve both a data refinement and the decomposition of an operation. The notions that Jones suggests for proving refinements correct are shown to have major shortcomings. The most important of these are the inability to change the degree of atomicity in a refinement, and the need to know how the environment of an operation is being developed in order to make a development step. The first problem militates against proper abstraction whilst the second militates against hierarchical design. The work has been presented at a seminar at the Hatfield Polytechnic on 20 November, and at RAL on 27 November.
In January JCPW completed work on a paper describing this work. The paper describes previously unknown incompleteness results for his notion of refinement. A summary of this work was presented at a workshop on refinement at York in January, a paper has been accepted at the VDM 88 Symposium, and a full report submitted as a monograph at the PRG, Oxford. In collaboration with Jifeng at the PRG, JCPW has a draft of a new set of rules for refinement, which is believed to be complete. The proof of this has yet to be discovered.
Work on verification of processes in the notations of CSP has also continued. Often small processes and simple specifications lead to surprisingly long proofs. Using ideas of fully abstract states, a normal form for such a specification has been developed so that it denotes the same behaviour as a corresponding normal form process. A proof rule for determining the correspondence has been verified. This reduced the burden of proof in the examples that have been studied. Proofs of satisfaction are replaced by chains of equivalences in the model. The normal form specifications are written in the readiness model for CSP, and are closely related to operations in concurrent Z.
JCPW has contributed to the work on a Standard for Z.
Working with Jim Davies, (JCPW's D.Phil student) a specification and implementation of the Ethernet protocol in timed CSP has been produced. This is the first application of the timed model, and it is revealing many interesting insights and theoretical problems. There seems to be much work to be done here.
JCPW has recently completed a book on formal methods.
Mario Martins (MM) from the University of Minho, was a Visiting Scientist from March 1987 to December 1987. During the period September 1987 to December 1987, MM's work was concerned with the specification of user interfaces. Two papers were produced during the period, one of which describes an extension to the production system formalism which can be used for the specification of concurrent systems. The second built on earlier work with Jose Oliverra and describes the use of constructive and algebraic specifications in formalizing some aspects of the user interface.
A Call for Proposals for Phase II of the ESPRIT (European Strategic Programme of Research in Information Technology) was formally issued in December 1987, with a response date set of 12 April 1988. From the first ESPIRIT II Proposers Day held at the end of the ESPRIT I technical week in September 1987 until 12 April 1988, DAD and many others were almost completely absorbed in endless rounds of European travel and meetings formulating research proposals for ESPIRIT II. The effort involved in ESPIRIT proposals is infinitely greater than that involved in formulating SERC RG2 proposals. AJJD and MSP gave very valuable assistance when the workload was at its peak.
The outcome of the ESPIRIT II proposals is not completely decided at the time of writing, but Software Engineering Group hopes to be involved in one project. Applications are also being formulated for the UK IED collaborative Programme.
The group now has a large number of links with other research groups, both national and international, and it is hoped that some of these will mature into formally funded projects. The following list is some of the organisations with which we have formal or informal collaborations.
BMM is taking an MSc Course at Imperial College London in the Foundations of Advanced Information Technology (FAIT). This is on a part time basis, two days a week for two years. This course is intellectually very demanding, but gives a very good foundation in-theoretical computer science. BMM has now embarked on the project which will conclude MSc. This is on Strategies for Theorem Proving in an Equational Reasoning System, and is under the joint supervision of Jeremy Dick and Chris Hankin at Imperial College.
AJJD is secretary to the Alvey SE Formal Methods Advisory Group.
DAD continues as Vice Chairman of Eurographics, participates in the Executive and Professional Board activities and is Programme Chairman for the Eurographics 88 event in Nice.
DAD edited a book on Knowledge Representation with Gordon Ringland, which contains contributions from many members of the Division. The book was published in April 1988.
DAD and AJJD spent a week in Barcelona at the Technical University in November 1987, teaching a course on algebraic specification. This led to an opportunity to do a little bit of graphics research through a simplified system showing how PHIGS could be defined on top of GKS-3D.
The UK Term Rewriting Group, formed by AJJD and JRK jointly with Ursula Martin, has had several informal meetings. Two excellent formal meetings were held in Glasgow on Induction and Inductive Inference, and in Manchester on Extensions to Term Rewriting.
AJJD and JRK are both involved in the organisation of the BCS FACS meeting in term rewriting to be held in September 1988 in Bristol.
Work has continued on the Para1fex project investigating architectures for know1edge-based systems. A number of extremely promising directions are being investigated, within the general framework of a unified architecture which is being developed to allow experimentation with the ideas. The possibilities for reusability of knowledge are being investigated, and it has been discovered that there is great potential here for making the process of building knowledge-based systems faster and more reliable. It has been found that there is a greater variety of approaches available than is found in reuse of conventional software. Knowledge-based systems, including the Source of Finance Adviser on which the Paralfex work is currently concentrating, may have different user roles with much knowledge in common but a certain amount varying. The reuse of the common knowledge would reduce the need for knowledge acquisition when building such complementary systems. There is also the idea of abstract generic tasks, applicable in specifying problem-solving strategies across domains, and of reusable packages of domain knowledge (for instance, the knowledge associated with the balance sheet/profit-and-loss account).
Another direction of work is concerned with the explicit representation of control knowledge by means of prototypical tasks. It is well-established that control or strategic knowledge must be made available to a knowledge-based system explicitly if the system is to be able to explain its reasoning adequately. At one level the representation of generic tasks contributes to this; another approach is to represent control knowledge as prototypes describing the typical sequence of actions associated with any entity in the domain from simple numerical quantities to the consultation as a whole and possibly including the request and generation of explanation.
Finally, a last theme is the integration of graphical explanation facilities aimed at the end-user.
The current version of the Source of Finance Adviser is Version 1.6 which works on the Symbolics Release 6.1/ART2.0. With the acquisition of ART Release3.x for both the Symbolics and the SUN which provides much improved facilities, the SOFA system is being reimplemented to make use of these enhancements. As a familiarisation exercise for CYLK, SOFA 1.6 has been ported to the SUN/ART3.l. This was to gain more knowledge of both the internals of SOFA and programming in ART. There is a great deal of difference between the Common Lisps on the two machines.
The Symbolics had its OS upgraded to Genera 7.1. This took a couple of days. First to back up the existing file-store (200MB to 40MB cartridges) and then the man from Symbolics did the actual installation of 7.1. In addition, the SUN 3/160 LIGNUM was upgraded to run ART. This required memory increase from 4MB to 16MB and the disk reconfigured to increase the swap space to 80MB. The purchase of Common Lisp had to be chased up due to a slip up of communication somewhere. We didn't get the documentation and tapes until late July. When the software arrived IDUS management had to be cajoled into providing 60MB of disk space to unload/install ART and SUN Common Lisp. There was a problem of getting LIGNUM to realise 80MB of swap. There were also problems of installing ART due to incorrect version of SUN Common Lisp. This was rectified after the visit to Ferranti International.
A sandwich student from Teesside Polytechnic, (LNW), is now working on the project. Her first assignment is to port the Source of Finance Adviser into Goldworks, an AI toolkit running on a Compaq 386 machine. This will extend the range of comparative versions of the Adviser available to us, and put us in a position to experiment with forms of input and output other than the standard questioning of the user (e.g. the use of spreadsheet files). It is known that such a facility would greatly enhance the acceptability of the system and others like it to their intended users.
The review paper presented at last years SERC Knowledge Acquisition for Knowledge Engineering Workshop by MDW has been expanded to cover the variety of task models proposed by others for design systems in greater detail and will be published as a book chapter.
The RAL report produced last year describing a Machine Readable Psycholinguistic Database has been published as a journal article (MDW).
Temporal Logic has been studied, including:
A seminar was given to SE and KE on temporal logic.
A book, Approaches to Knowledge Representation: An Introduction edited by David Duce and GAR was published in April 1988, containing contribution from the above Group members, together with colleagues from HCI and SE.
The SERC-funded Intelligent Front End (IFE) project addresses the difficulties that end users experience when attempting to use existing engineering appraisal packages. It tackles two issues, firstly the poor quality of the current user interface and secondly the high level of skill/knowledge required to drive the appraisal packages.
The IFE currently consists of a central blackboard, accessed by three knowledge bases - for dialogue handling, rudimentary user modelling and building the user conceptualization. The dialogue knowledge base uses a dynamic form-fill package, (developed in conjunction with Strathclyde University) together with other supporting packages such as a map utility (MDW), 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. It is expected to have a demonstrator system by mid October.
At the request of the Alvey Directorate GAR with Dr Roy Leach of Heriot Watt conducted an extensive evaluation of the MHM (034) project during November and December.
During the year, project monitoring duties for Alvey have been undertaken by members of the group. MDW has been monitoring officer on 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.
The secretaryship of the committee SIG AI has been undertaken by a member of the group (MDW) since July 1988. SIG AI has been an advisory committee on Artificial Intelligence to SERC and Alvey in the past, and is expected to continue in that role under the new DTI IED structure.
CYLK is a member of the Main Working Group representing the SERC. The representation also covers one of the sub-committee meetings, that of the Built-In Predicates (BIP) sub-committee. The BIPs work is concerned with the definition of what predicates and their behaviour are to be included in the eventual Standard Prolog. The main committee meetings discuss the contributions of all the sub-committees collectively.
The state of the effort at July 88 is that the definitions of modules in Prolog has to be included in the standard (as most of the current implementations have some module facility) and the discussions on the subject have re-started.
Portable Common LOOPs (PCL)
This is a publicly available Object Oriented Programming system (OOPs) obtained via the network. A successful attempt to get PCL to work on top of Kyoto Common Lisp took about two weeks in November. The difficulty in this exercise was the lack of documentation. Although that was available free as well, it had to be ftp'd from the USA and at the time, the cost of this was too high.
Kyoto Common Lisp
Four weeks were set out just before Christmas last year with the aim of porting Kyoto Common Lisp to the Pyramid RISC architecture machine. If this was completely successful, a comparison of the performances of KCL on SUN and Pyramid would be done. The port was relatively straight forward as the documentation clearly laid out what source files needed to be re-written or altered and was completed right on target.
Performance Comparison
The main part of January was taken up benchmarking KCL to measure and compare the performances on the SUN and Pyramid. The bible of LISP benchmarking by Gabriel was heavily consulted during this process. The actual programs were kindly supplied by SUN Microsystems, UK.
Work has started on identifying appropriate areas within Administration at RAL where Knowledge Based support would be of value. It is intended to implement a demonstration system by April 1989.
From October to April GAR, SCL and MDW were heavily involved in the formulation of proposals for ESPRIT II. Papers arising from these proposals have subsequently been submitted for publication. Proposal 2474 - MMI2, concerned with Multi Modal Interface for Man Machine Interaction with Knowledge based systems has been accepted and we expect to be involved with this project over the next five years.
From July to mid-August GAR, SCL and MDW were concerned with formulating a number of IED1 proposals. The result of this activity is still awaited.
Initial contacts have been made and an outline proposal developed for submission to the CEC under the Advanced Informatics in Medicine (AIM) initiative. This proposal will be for a project to demonstrate that knowledge can be acquired from both medical and technical experts in different European countries and combined together with advanced sensor technology into a diagnostic KBS which can be used within the health systems of different European countries.
CYLK is a Member BSI Standards Committee on Prolog.
MDW has accepted to be European Representative on the Editorial Board of the journal Interacting with Computers and undertaken the relevant editorial duties. He has also accepted membership of the programme committee for the ACM conference CHI' 89 to be held in Austin, Texas in April 1989.
Mario Martins returned to his University in Portugal at the end of December 1987 after a very valuable 9 month visit to the Laboratory. Mike Parsons left the Laboratory in June 1988.
Mike Parsons, Brian Ritchie, and Jeremy Dick were awarded PhD' s during 1987/88.
Rob Witty was on sabbatical at Xerox PARC until May 1988 after which he resumed his role as Head of Software and Knowledge Engineering Group (aka Systems Engineering Division). At the invitation of Dr Bill Spencer, Vice President of Xerox's Corporate Research, RWW spent April 87-April 88 in the Intelligent Systems Lab of Xerox PARC, California. This allowed RWW to pursue his training goals in mathematical logic and IKBS through a combination of personal study and external courses run by Stanford, ACM, and the IEEE.
Residence in Palo Alto enabled RWW to appreciate the strategic factors contributing to the successful phenomenon (or phenomenal success) known as Silicon Valley.
The senior management of PARC were kind enough to educate RWW about the challenges and rewards which arise in the pursuit of research from an industrial perspective.
RWW would like to record his heartfelt thanks to the Alvey Directorate, RAL and to Bill Spencer, John Seely Brown, Frank Squires and Ron Rider and their colleagues in Xerox for enabling RWW to partake of such a wonderful opportunity to improve and expand his understanding.
Brian Bainbridge left in December 1987.
Damian Mac Randal was transferred to Design Division in the July 88 reorganisation.
Lana Woodhead joined on 11 July 1988 for an industrial sandwich year as part of her Teesside Polytechnic BSc course.
The work of the Division has been split up into sections on Computational Modelling, Design and the Transputer Initiative which reflect the new structure in the Department.
On 18 July, 1988, the reorganisation of Informatics brought together a number of activities into a new Computational Modelling Division. Most of these were previously part of the Engineering Computing Group.
RAL is involved in two closely related projects in two-dimensional modelling:
These are collaborative projects involving partners from industry and universities. The overall aim of the projects is to design and implement a flexible software system (called TAPDANCE) for modelling of semi-conductor processes.
Industrial and academic partners in these projects are:
The system consists of a shell provided by RAL into which a set of modules or kernels fit. RAL's involvement is the production of the shell, a number of the infrastructure modules and, as part of VLSI034, the Device Modelling kernel, DEVMOD. Other kernels are being produced by Kent (Implantation), Reading (Diffusion) and Swansea (Oxidation).
The people involved with their main responsibilities are:
C Greenough (CG) has been involved in both Alvey projects from the start and has been responsible for the overall concept and design of TAPDANCE. CG is the overall project manager and attends the Project Management Steering Committee meetings as director of software development. As the Alvey team has grown, some of the day-to-day project management has been delegated to JVA.
J V Ashby (JVA)'s role has been partly managerial, planning and overseeing the day to day work of the 066 project as well as liaising with the university and industrial partners on the software aspects of the projects.
JVA has participated in the implementation of TAPDANCE shell with CG and the integration of the various kernels and directing RAL effort in those 066 kernels for which RAL has responsibility. In addition, he has worked on various elements of the shell.
The TAPDANCE shell comprises a database through which all data transfer is effected, a manager to handle it and to allow limited interaction with the operating system, a display module for post-processing of results and an overall controller. During the year several improvements have been made to the database, notably to extend its capability to handle such things as I-V graphs. At present the database is a single direct access file made up of several domains, each of which contains a 'snapshot' of the semiconductor device. This has the disadvantage that any changes to a domain are immediate and irretrievable. JVA worked on a version which uses a temporary copy of the current domain and works on that, thus giving greater flexibility.
During the year JVA has supervised a Sandwich Student, Karen Whitaker, in work on the simple command parser used within T APDANCE. They have changed the command tables to use an efficient sparse storage scheme and developed a rudimentary full-screen form filling mode of interaction which uses GKS to ensure portability.
Much time has been spent revising the DISPLAY module. JVA has introduced the capability to produce isometric plots based on a general triangular finite element mesh based on hidden line and hidden surface algorithms developed at University College Swansea. With the advent of RALGKS at level 2b zooming has been revised. An additional command has been put into the module to enable the user to draw results of simulations such as the I-V curves referred to above. Documentation of the DISPLAY module has been started.
TAPDANCE has now been released to four external sites, the universities of Kent and Swansea, STL Harlow and Plessey Caswell, where it is running on Micro-VAXes. The need for these releases has involved JVA in preparing a VMS version of TAPDANCE and feeding back knowledge gained on portability into the master version which is held on Prime-H.
Through the year JVA and CG have attended several Alvey 066 project meetings as well as the Alvey Layer Processing Club meeting in Edinburgh. In addition they organised a special software meeting at RAL for the 066 project. CG has attended the project management steering committee meetings.
R J Fawcett has been implementing RAL's 066 kernels within TAPDANCE. Early in the year, RJF finished writing and testing the SIMPLANT kernel. SIMPLANT is a kernel which simulates ion beam implantation into any target made of semiconductor device fabrication materials. Such implantation is an important step in the manufacturing sequence of semiconductor devices. A typical target consists of irregular abutting regions of silicon, silicon oxide and silicon nitride. SIMPLANT incorporates some analytic results in order to speed its operation. It is a faster, more robust but less rigorous alternative to the IMPLANTATION kernel. SIMPLANT has proved its worth in the months which have followed its incorporation into the TAPDANCE package.
More recently RJF has undertaken major development work on the GEOMETRY modelling kernel. This kernel provides the TAPDANCE user with the means of specifying and modifying the geometric descriptions of semiconductor devices.
In its original form the GEOMETRY kernel (developed by C J Hunt and R F Fowler) accepted geometric data from the screen with minimal checking. More complete checks were made before geometric descriptions were written to the TAPDANCE database but an option was available to store invalid descriptions. This option existed to ensure that the time spent in data entry should not be wasted because of an abrupt end to a session. The first piece of development work on the GEOMETRY kernel was to improve data checking and to place it at the moment of data entry. This proved to be rather complicated because of the need to check the dependents of any modified item. For example, movement of a point would demand that the point be checked along with any lines and regions incorporating it. Dependency tables therefore had to be generated and maintained. The changes made the GEOMETRY kernel much easier to use, with errors producing instant diagnostic messages, and removed the possibility of storing invalid data in the database.
RJF has continued the development work on the GEOEMTRY kernel by adding five new commands: to delete geometric objects, to clear a geometric description, to define or modify an electrical contact, to apply a mask to part of a device and to etch part of a device. The first three of these commands were fairly straightforward while the last two were particularly complex. Both the masking and etching commands allow isotropic or anisotropic operations to be simulated. They incorporate the automatic generation and deletion of points, lines and regions with the consequent modification of dependency tables.
RJF has produced a laboratory report on the SIMPLANT kernel and is currently writing user manuals and system documentation for the SIMPLANT and GEOMETRY kernels. The SIMPLANT user manual is close to completion and the GEOMETRY user manual is well on its way.
The integration of the university produced kernels and DEVMOD, RAL's device modelling kernel, has been a continual task through the year. The University of Kent delivered their full implantation code and JV A successfully installed it. Collaboration with Reading University and Swansea has made progress in the coupling of the Diffusion and Oxidation kernels, and under RAL's guidance Swansea have been developing their Oxidation kernel to cope with general geometries and to use a triangular mesh. DEVMOD is now fully integrated into TAPDANCE and simple examples passing from process steps through to device modelling have been run.
The contribution of the Rutherford Appleton Laboratory to the 034 project is the production of a general purpose, two dimensional semiconductor simulation code, currently known as DEVMOD. DEVMOD solves the fundamental semiconductor device equations with the current transport given by the drift-diffusion model. The material model allows for varying semiconductor composition, and is a basis for future modelling of heterostructure semiconductor devices. The emphasis at present is in modelling silicon.
The device equations; Poisson's equation and the current continuity equations are solved by a simultaneous solution method, with a decoupled approach used for the initial solution guess. The program makes use of a number of state-of-the-art features: the device domain is discretised by a triangular finite-element mesh that possesses the DeLaunay property. This allows for arbitrary geometrical problems to be catered for. The linear system of equations make use of the ICCG and CGS solvers. (ICCG is based on one developed by CG and CGS was written by Dr P Mole of GEC.)
Through TAPDANCE, the user can define the geometry and mesh via the appropriate modules and access results, which are stored in the database, via the display module. The various process modelling modules can provide impurity distributions to DEVMOD, otherwise simple uniform and Gaussian doping profiles may be specified within DEVMOD.
Progress has been made during the past year and DEVMOD is now able to solve a range of basic problems. The most important development to the code in this time has been the inclusion of the ability to solve the on-state problem, ie when current is flowing in the devices.
EMA has been involved in the design of the module and in the derivation of the discrete algebraic system. EMA and RFF have been implementing these discrete equations.
RFF has carried out tests on a set of benchmark problems, ranging from finding the current flowing in a block of semiconductor material with uniform doping to the steady state operation of a MOSFET (a metal-oxide-silicon field effect transistor, one of the basic building blocks of modem large scale integrated circuits). Comparison with data from other sources has confirmed that the simulation code is giving satisfactory results.
RFF and EMA are currently improving the performance of DEVMOD, both in terms of speed and robustness, since it is intended to be a general purpose simulator that will be used by a wide range of users, rather than a research code.
DEVMOD has been written to run within the TAPDANCE shell that has also been produced at RAL. This system gives an integrated environment for running DEVMOD, along with other programs, or kernels, that calculate the structure of the semiconductor devices from the fabrication processes. The device modeller is able to use the information on the doping from the process simulators and predict how such a device will perform. RFF has written an interpolation routine to allow DEVMOD to use the dopant information within the data base of TAPDANCE.
During this period RFF has also spent some time porting the TAPDANCE system to run on the SUN/Pyramid UNIX systems.
Non-Parabolicitv Effects in Semiconductor Transport Models
The effects of nonparabolic energy-wave vector dispersion have to date been ignored in the hydrodynamic modelling of charge carrier transport in semiconductors. In the case of some semiconductors this can be justified on the basis that the effect introduces only a small correction. However in the case of III-V semiconductors, non-parabolicity can introduce significant changes, for example the current density in InAs under degenerate conditions is reduced by 40% compared with a parabolic model. With the increasing technological importance of III-V semiconductors a revised transport model is required.
In order to make the problem tractable, a first order non-parabolicity approach was taken and a set of high-field transport equations derived by taking moments of the Boltzmann transport equation. The resulting model, which also accounts for heterostructure and degenerate semiconductors clearly reveals the influence of non-parabolicity. Furthermore, by taking the low field limit, the commonly used drift-diffusion current density can be corrected by a number of non-parabolicity coefficients.
This work is due to appear in J of Applied Physics, 18 August 1988.
Modelling High-Field Effects in Submicron Semiconductor Devices
The program HETEM has been written to solve the hydrodynamic transport equations, including the energy and momentum conservation equations. The energy transport model allows the possibility of investigating the hot electron effects in submicron semiconductor devices, including velocity overshoot and energy and momentum relaxation. The program has been applied to modelling the GaAlAs/GaAs heterojunction bipolar transistor.
RAL is involved in an important ESPRIT device modelling project. A primary aim of the project is to develop efficient algorithms to solve the drift-diffusion form of the device equations in three-dimensions. As semiconductor devices become smaller their three-dimensional nature begins to show itself. For European industry to predict the behaviour of sub-micro devices if of paramount importance. The project has ten industrial and academic partners:
The project is divided into five workpackages:
RAL is the prime contractor and leads the whole project as well as directly the most important Workpackage, Workpackage 5 the Project Research Code.
RAL is also technically involved in Workpackages 2, 3 and 4.
Over the past year C Greenough, C J Hunt, T K Patel and D Gunasekera, have been working on the development of a comprehensive three-dimensional modelling system.
CG as well as being the co-ordinator of all RAL effort, is the director of the software programme within the project and co-ordinator of Workpackage 5, and is responsible for utilising the eight man-years of effort from four partner sites (RAL, TCD, UCS, STL). CG is responsible for the overall design of the system which consists of four major program modules coupled through a simple neutral file system. The modules are: geometric modeller and mesh generator (TCD), impurity profile generator (UCS), the analysis code (RAL) and the post-processor (UCS). STL have been involved in the evaluation of the system. RAL is responsible for the overall planning and control of the Workpackage, the development of the analysis module and the provision of the integration tools.
CJH and DG have been involved in the design and implementation of the solver module. The system can now solve the on-state semiconductor problem in three-dimensions. The complete integrate system was demonstrated at a successful Commission review.
TKP has been involved in the development of the neutral file system. This was based on the one design and implemented by C R I Emson et al. This work involved the conversion of the system from sequential ASCII files to direct access binary. This gave a speed improvement of a factor of two.
The NAG/SERC Finite Element Library is now in the third release. Release 3 of the software has been with NAG Ltd since the autumn.
There are now over 100 licence holders of the Library world-wide and a large user population. Although the software runs to some 150 Fortran Sub-routines and 14 example programs, only two significant bugs have been reported since the release of Release 2.0 in 1984.
CG is in control of the whole project and has planned the developments and content of the next three releases of the Library.
CG and CJH have been involved in developing both new routines for the basic library and new example programs. Some work has also continued on developing the graphical output capabilities of the Library and the pre-processing has been enhanced with a simple mesh generator and bandwidth minimiser.
Release 4 is now in preparation for delivery to NAG Ltd. An important addition to the Library will be two example programs on the solution of the Navier-Stoker equations.
PARFEL - A Parallel Finite Element Library.
DAP-FELIB (C Greenough)
Fitted Finite Elements (C Greenough)
The staff involved in these activities have been
NASTRAN has continued to be the main structural analysis finite element package supported, although use has been largely in-house.
LUSAS has been mounted on the PYRAMID, both to assess the capability of the software and to gain experience of operating applications packages in a UNIX environment.
The Pre- and post-processors FAMBUILD and FAMRESULT (formerly FEMGEN and FEMVIEW) are now installed both on PRIMEs and SUNs.
A neutral file system RALBIC now facilitates the interchange of finite element data between these pre- and post-processors and users' programmes. It is anticipated that this will be superseded eventually by the CAD*I neutral file development.
The latest versions of the RAL electromagnetics software packages, PE2D and TOSCA, upgraded by Vector Fields Limited, have been mounted on the PRIMEs, the SUNs and also on CMS. The solver module for the 3D TOSCA package has been modified to run on the CRAY-XMP/48. This executes at about 9 times the speed of the IBM in the solution of non-linear problems. C J Collie helped A R Mayhook (CCD) in this activity.
A study is being made of the solution of dense matrices using the AMT DAP with a view to its application to packages such as BIM2D, which is a Boundary Integral computer program for analysis of magnetic field problems.
A course was held in April 1988 for users of the PE2D package. This was attended by 9 people, all from UK universities.
C S Harrold has developed Hextet - a program which performs the automatic subdivision of a mesh of irregular hexahedral elements (8-node bricks with planar places) into tetrahedra without introducing extra nodes. The triangular faces of these tetrahedra need to be coherently matched. Surprisingly, there are 74 such subdivisions (2 using 5 tetrahedra and the rest being sextuplets) for a brick element. Such subdivisions only generate 64 triangulations of a brick surface. This allows us to place constraints on the choice of subdivision. In particular, the aim is to avoid tetrahedra with negligible volume. However, only random access to brick elements can be assumed, matching brick faces is a challenging problem but a working solution can be obtained by using group theory. Some more coding will be required before this program can be used in conjunction with a neutral file system.
The main activity in the past year has been the completion of an SERC funded project involving RAL, Bath University and Imperial College on 3D Eddy Current analysis. The project has resulted in the production of a research computer program written at RAL capable of analysing 3D linear transient eddy current problems. Also as part of the project, the implementation of a Neutral File can now be used to interface the eddy current package with commercial pre- and post-processors namely FEMGEN and FEMVIEW, as well as the Post-viewer written by Imperial College as part of the project.
As part of the activities in electromagnetics research at RAL, CRIE has also been involved in a number of other tasks. These include the assessment of a 3D eddy current code developed by the group at Electricite de France, the implementation of the RAL eddy current software on the SUN computer, and the organisation of another in a series of eddy current seminars held at Cosener's House (see Publication list below). During the seminar, the first in (hopefully) a series of meetings was held to enable collaboration between a very active group in Italy and ourselves in UK on 3D magnetic field studies. The meeting was attended by 4 professors from Italy, who were also attending the seminar, as well as 3 professors from UK. C J Collie produced one of the test programs (eddy currents in a hollow sphere) used at the workshop.
CRIE has also been involved in the submission of grant applications to continue funding for the electromagnetic work at RAL. This has included 2 submissions for a project on high frequency electromagnetic device analysis, neither of which were approved. He was also involved in the preparation of a large grant proposal for ESPRIT II on electrical device design using knowledge based systems. This too was not funded.
It is envisaged that SERC will make use of CHEST (The Combined Higher Education Software Team), who are based at Bath university, for supply of commercially available software for engineering applications. This will, of course, only be necessary in respect of those grant holders who cannot obtain the required facilities directly via their local computing centre.
The SERC normally has a relatively passive role in the selection of the research areas which it supports. However, recently the SERC has tried to concentrate some of the available research funds in areas perceived to be of national importance. This means that the passive role of SERC has been partly replaced by more selective mechanisms for enc0uraging research in these selected areas such as the setting up of Specially Promoted Programmes (SPPs). It has long been recognised that the success of any UK energy conservation programme is dependent upon reducing energy consumption in buildings and then associated services, since this accounts for approximately half of the total amount of prime energy consumed nationally for all purposes. Consequently during the 1970 's there was considerable growth in the number of Research and Demonstration projects in this area, and it was against this background that the Energy in Buildings Specially Promoted Programme was set up by the Building Sub-Committee during 1979.
Although personnel at RAL have supported Energy in Buildings related activities for 8 years now this is the first year that Informatics Department has become involved. Two broad areas of research are being supported at present and it is anticipated that these areas will continue to develop.
Firstly a stated objective of the Energy in Buildings SPP was to increase the fundamental understanding of, and develop the methodology for, the prediction of the dynamic thermal behaviour of buildings. There was a need for further development of models for the prediction of the thermal dynamic behaviour of buildings and systems. In particular there is an awareness of the need to improve the confidence in, and the credibility of, thermal models if they are to gain widespread acceptance. Consequently from the inception of the SPP the establishment of a Model Validation exercise has been a central concern and RAL has been one of the four institutions participating in this validation exercise. The emphasis of RAL's contribution has been towards the identification, implementation and preliminary application of techniques which will identify the accuracy, sensitivity and range of appropriate use of simulation models and the algorithms contained therein. These objectives have been achieved by using existing time series analysis and statistical quality control techniques and where necessary refining them or developing new techniques. For example a stochastic sensitivity analysis technique has been developed which relates the non-linear system sensitivity functions to the response functions. Also the statistical vector T2 test (telling test) originally applied to comparing the experimental against simulation model predicted time series output has been extended to perform vector tests on the correlation functions on power spectral densities. In addition a novel technique has been developed which should statistically test the eigenvalues of the impulse and frequency response matrices. It is recognised that these techniques will be sensitive to outliers in the data and a robust covariance estimation technique is being developed which examines the derivatives of the covariance. RAL and Cranfield are currently preparing a joint research proposal for submission which will not only continue development of the statistical validation techniques but also include them in the Energy Kernel System and apply them to a limited range of algorithms in the EKS.
Secondly RAL's initial involvement with the SPP was in the area of performance assessment. RAL is and will continue to support this area of research. For example at RAL in 1988 a non-linear theory of thermal transmission through the building fabric has been developed9 which is based upon a technique developed for the extraction of Volterra kernels from the data and the transmission theory is being tested in collaboration with the university of Newcastle. Performance assessment tools, based upon Volterra kernel extraction using central statistical moments, will be developed in collaboration with Bristol University and by using existing experimental data and simulation models the convergence, accuracy and range of appropriate use of such tools will be assessed before discrimination.
Finally the rest of this financial year is to be devoted to clearing the backlog of papers arising from the above work.
C J Collie has started providing A D Irving with support in these activities.
This is a new section which is to take responsibility for visualisation of complex data, which could be complex multi-dimensional results or complex models or one superimposed on the other.
Some preliminary work has been done in the old SUS/Applications section during the year by JH, who has been assessing workstations with added 3D graphics power. An assessment of functional capabilities has been carried out and a benchmark has been written. The benchmark is adjustable in the quality of rendering that it demands and in the complexity of the scene and is expected to be suitable for testing the new superworkstations becoming available.
Other work during the year has been done by RP on ray-casting on transputer systems and is described under Transputer Developments (3.2.9.3).
The new section will consist of:
Transputer Developments during the year have been in two main sections. Chris Wadsworth has run a small team concerned with technical developments involving the transputer. For the first half of the year these people were also involved in the co-ordination activities. The main goal now is to build up the Department's technical involvement in projects involving the transputer and parallel processing in general.
A separate activity, but now under the same management, has been assessing the ability to enhance the power of the IBM 6150 workstation by adding transputers to it. IBM have funded the hardware for this activity in return for access to the papers produced by the project.
The staff involved in the two activities are:
CPW was Deputy Co-ordinator for the Initiative until this was handed over to Cyril Balderson in January 1988. This entailed involvement in all aspects of getting the Initiative started - the academic loan pool, establishment of the National and Regional Centres, priorities for initial development contracts, etc - and continuing publicity, presentations and demonstrations. Presentations about the Initiative were given to three conferences (Liverpool, Bangor, Bath), to several supplier companies, and to an assortment of meetings, including demonstrations to Council and to the Engineering Board in support of the case for the future funding of the Initiative.
CPW continues to liaise closely with the RAL Co-ordination Team, particularly in technical matters, and is a member of the Transputer Application Management Group (TAMG), the steering group established by Computing Facilities Committee to oversee and advise on the running of the Initiative. He also chaired and led the Workshop on Transputer Development Environments held at Cosener's House in November 1987. The Workshop produced a detailed set of recommendations on suggested developments (not all of them short-term!), which have now been absorbed by T AMG into the Initiative's plans for development contracts for 1988/89.
An interim Centre, with CPW as Head, was established at RAL as soon as the Initiative started in April 1987, in order to provide an initial focus for the Initiative until the National and Regional Centres became fully operational. During this period the interim Centre ran two courses given by INMOS staff and hosted a variety of users, mainly loan pool applicants whose needs are best met by access to Centre facilities. Following the confirmation of RAL as the permanent Regional Centre for London and the South East, Richard Parkes has taken over as Manager of the Centre.
Technical backup and support to the Centre is now provided mainly by the Parallel Processing Group in the new Department structure, with assistance from others as appropriate from time to time. Much of this work to date has consisted of initial discussions with, and technical advice to, prospective industrial customers of the Centre, some of whom are now booked to use the Centre facilities. It is also expected that some of these contacts may well lead to consultancy or joint development work with particular companies.
One project has been worked on to date, and one probable future project is currently being discussed.
This general area covers anything concerning tools, advice, or other aids to assist users in porting existing software to transputer systems. An initial six-month development contract was placed with Professor Hey's group at the University of Southampton. The goals for the six-month contract, which ended in March 1988, were:
CPW collaborated with Southampton on part (b). A summary of the work given by Professor Hey at the Transputer Initiative Seminar on reports from the first round of development contracts, held at RAL in April 1988, was well received and it is likely that this work will be extended.
This was an EMR (in conjunction with Dr Stuart Robinson of Brunel University) to develop a parallelising Fortran compiler. An initial feasibility study was carried out and a report was produced. The long term aim is to take existing sequential Fortran code, automatically extract the parallelism and run it on a network of Transputers.
In particular, DJJ was involved with the development of a static Fortran analyser, that would produce statistics from Fortran source to reveal the nature of any parallelism present.
DJJ has produced a number of application independent harnesses that should simplify the task of moving code on to a multi-processor system. These harnesses were given built-in monitoring facilities so their design could be improved on an iterative basis.
HAP involves research into parallel algorithms which can be used to accelerate and support an existing sequential package STABLE-H for H+¥ design of finite dimensional, multivariable, time invariant feedback control problems. The first step has been a feasibility study to identify the main areas of the package where parallelism can be applied and a set of core algorithms defined. The result of the study appears in [4]. The remaining time has been spent on developing test algorithms using transputer hardware for the core algorithms identified above, this has included Ricatti (Lyapunov) equation solution, Singular Value Decompositions (SVD's), Model Matching problems, and Multivariable Frequency Response (see [6][7][8] and [9] for details). At this time the practicality of the accelerator has been demonstrated and the type of parallel algorithms required identified construction of the accelerator has been delayed until appropriate software and tools for developing a PC/package interface have satisfactorily evolved and become available (eg SUN interface and TDS Toolset). The Research for better algorithms and techniques is on going.
To complement the designs above, attempts have been made to formalise a general mapping technique for converting Feedback Control System models into Neural Networks. The idea is to map control problems into neural networks and train the weights of the network for a given control problem. An inverse transformation then yields the coefficients of the transfer functions for implementing the control problem. Other benefits such as fault tolerant designs, distributed control, and expert networks for CAD packages will result from a working model. So far a generalised framework has been identified but more analysis is required before the implications of the design in control and AI contexts can be fully analysed. Hence no significant results are available.
RP and GMM have been developing a ray casting machine in OCCAM based on the SCRIP architecture. Most of the programming has been done by RP and a primitive system is almost working.
The IBM 6150/Transputer project was a project of two tales. It began with a joint study with IBM and the objective was to increase the processing power of the 6150 by interfacing it with the INMOS transputer. This involved implementing a device driver for the transputer board and a server for the host/transputer communication. Software development on the 6150/transputer architecture could be done either in the Transputer Development System (with tools such as the folding editor) or the Unix System environment (with tools such as vi, make, sccs etc). Utilities were provided to facilitate transferring files between the two environments. Recent advances made it possible to dispense with the Transputer Development System completely which marked a significant change in the way the transputer could be used. The approach was, to a certain extent, novel at the time it was conceived, but was gaining wider recognition and acceptance as people started to appreciate the transputer was not the ends but only the means of computing. It was believed that a transputer acting as a master with a host acting as a slave peripheral (as in the early days when the host was dominated by the PCs) was extremely restrictive and in the long term more expensive. On the other hand, a suitable host with a suitable operating system environment would open up all other opportunities, such as networking, high quality graphics, windowing, program development and support environment, document preparation, database management etc to the users. It was clear that a good workstation with a Unix operating system would be ideal for the transputer. Conversely, the justification for having the transputer on the workstation is based on applicability, performance and cost effectiveness. Unless one was prepared to pay a hefty sum for a super-workstation, presently a workstation could deliver typically not more than 10 MIPS which would easily be dwarfed by a transputer system. In short, the performance of a workstation can be improved many times at an affordable price by equipping it with a transputer system which would allow it to be used in applications that are not possible today because of insufficient processing power.
With limited resources, it had always been our firm belief that research and development work should be driven by real applications. No matter how convincing our workstation/transputer architecture argument was, it would have to be proved by genuine demands for such a system. Therefore, in the second part of the project, we were very pleased to collaborate with Vector Fields in implementing BIM2D, an electromagnetic design package, on the 6150/transputer system. The implementation was designed to explore parallel processing as offered by the transputer system and high quality graphics, which would be essential for any CAD packages, as supported on the 6150 workstation. Apart from having a deliverable product, the main result of this part of the project was to assist in sharpening the requirements in the RAL/IBM joint study.
Vector Fields originally intended to move the Fortran code for the solver part of BIM2D onto the transputers, and rewrite it as a parallel program. However, because of delays in the release of the Inmos Fortran compiler, and subsequent delays in acquiring the Occam toolset (which is needed to run Fortran processes in parallel), MHR wrote a parallel Gaussian elimination algorithm in Occam, which has been interfaced to BIM2D. BWH installed a GINO graphics simulator to support BIM2D, this simulator provided graphics support on both the IBM 6150 and the SUN machines. BWH spent some time tracing an apparent fault in the graphics system on BIM2D through the BIM2D FORTRAN code. BWH finally isolated the error and proved that it was an error in the IBM FORTRAN compiler. As an aid to the graphics support BWH also installed BIM2D on the SUN, this allowed the debugging of the GINO simulator on the IBM 6150 to be checked against the GINO simulator on SUN and the real GINO package on the PRIME machines.
To demonstrate output from the transputer board producing graphics on the 6150 display, HKFY and BWH wrote a Mandelbrot program in Occam, linked via a Unix pipe to a C program running the display. Following the installation of an additional board with four transputers, a multi-transputer version was developed by BWH with assistance from MHR.
A more flexible interface, using System V Inter Process Communications routines, has been developed by MHR to allow client programs running on the 6150 to communicate with server programs on the transputer board.
To learn more about multi-transputer programs MHR rewrote the mandelbrot demonstration program to do dynamic load balancing using the farm approach. This produced a significant improvement in performance.
The Fortran compiler and stand-alone Occam toolset have since arrived and been installed. To support Fortran and the stand-alone Occam (which run without the Transputer Development System), the Inmos 'alien language file server' was ported to the 6150, and MHR added the RAL inter-process communication enhancements which allow a server process on the transputer to talk to a client application running under Unix on the 6150.
To enable the Occam toolset to be used on the 6150 MHR has written some interface programs which emulate the MS-DOS interface programs provided by Inmos. Source to a makefile generator is provided by Inmos; this determines the dependencies within the components of an Occam program and creates a makefile suitable for building the program. MHR made some changes to get this running under Unix, (the original is designed for use with Microsoft make under MS-DOS).
The stand-alone toolset integrates better with standard Unix development tools, (such as make and SCCS), than the Transputer Development System.
MHR has used a simple test program to show that multi-transputer can be built and run on Fortran programs the 6150 transputer system. The way is now open for Vector Fields to integrate some of their Fortran routines from BIM2D into the existing Occam Gaussian elimination program.
MHR made use of the Transputer Centre Inmos ITEM rack to connect an array of 32 T800 transputers to the 6150. The Gaussian elimination program was reconfigured and ran successfully on the larger system. Further tuning of the algorithm, to make more efficient use of the T800s needs to be done, but the initial test showed a speed up of around seventy times when compared with a single T4 transputer.
BWH worked during the year on the rtserver program. This program was very restrictive as it only allowed access to the Transputer Development System (TDS) editor or an IBM 6155 terminal. BWH initially provided a new help screen mechanism within TDS. He then converted rtserver to run under the curses terminal interface package, thus allowing rtserver to display on any ASCII terminal with a terminfo file. This still left rtserver with a user interface which relied on special keys found only on the IBM 6155 terminal. He then provided a standard user interface to allow users to access the TDS editor in a common fashion. This interface uses standard Alpha-numeric keys and the Escape key, the interface provides pop-up menus to aid the user to access the 37 functions available within the TDS editor. This interface allows users over the Ethernet to access the transputers on the IBM.
BWH has also been working on providing other utilities for the IBM 6150. These consist of a Screen-blank program and a suite of programs to help with the TDS system. The screen-blank program monitors the use of the console display and blanks the screen if the console is not used for more than 15 minutes. The TDS utility suite consists of four programs:
This suite provides a TDS/UNIX programming workbench allowing the user to move easily between the TDS filesystem and the UNIX filesystem.
BWH has been learning to program in Occam during the year. This was initially required to understand the TDS editor. BWH worked on providing a simple mandelbrot test to be used for benchmarks. This program helped to introduce Occam and parallel processing. A version which could run on one transputer was provided. When a 4-board transputer card was added to the IBM 6150 BWH extended his program to run under 5 transputers. Graphics using GSL {the IBM graphics support package, were used to provide the graphics needed to display the Mandelbrot images.
BWH has supported the work on the IBM 6150 during the year by installing various packages. These packages have included PHIGS, X-windows, BWH assisted in installing TCP/IP and an Ethernet card. BWH has also helped install two new versions of the Operating System AIX 2.1.1 and AIX 2.1.2. Simple testing of the PHIGS and X-windows systems was carried out by BWH, these resulted in producing small demonstration programs to show the systems in operation. BWH has also taken part in the installation on three disk drives and two transputer cards on the IBM 6150.
BWH spent part of the year on the TRANSA program. This program could solve transient dynamic problems using an explicit method for the integration of equations. It was expected that since this package was written in Occam that it would be a relatively simple task to install it on our IBM 6150. This program for the transputer arrived in an early version of Occam 2 on DOS disks. But BWH transferred this to AIX, converted the Occam code into a format compatible with the current Occam compiler and placed it in the TDS system. BWH converted it from multi-transputer to single transputer format to simplify testing and debugging. Although the program was written for DOS operating system, BWH managed to provide read support for DOS in the rtserver. This support enabled the program to be started and a set of test data read into the transputer. Unfortunately the program would not progress any further and no calculations or results could be generated. It was found to be difficult to extract debugging output from the various parallel processes during execution of the program. Not enough time was available to insert a suitable debugging framework, so no further work has been carried out.
As an introduction to Curses on the Unix system BWH investigated the terminfo entries for various terminals. BWH produced a terminfo entry for the Cifer 2634 on the pyramid and IBM 6150.
BWH has provided a Device Driver to handle the Transputer card on the IBM 6150. The Unix structure on the IBM is quite different due to the inclusion of a Virtual Resource Manager (VRM) to isolate the Operating System from the Hardware. BWH created a character device which provides the transputer with a portable interface. The interface also provides security against multi-user access of the transputer. BWH has provided a version of rtserver using the interface to test the Device Driver.
BWH is currently writing reports on both the TRANSA and Transputer Device driver projects.
A D Bryden (ADB) was a member of the Computational Fluid Dynamics (CFD) Advisory Group which held six meetings during the year. The Group submitted an interim report to the Electro Mechanical Engineering Committee (EMEC) in February 1988 followed by a final report which was approved by EMEC in June 1988. ADB provided the computing background to ensure that the recommendations on computing are compatible with EASE. He played a major part in drafting the final report.
As a result of EMEC's acceptance of the report, ADB is planning some of the details for implementation of the report. A visit is being made to NASA Ames Laboratory to discuss acquisition of their computational data. A Workshop will be held in the Autumn to make more detailed proposals relating to the policy agreed by EMEC.
Before the reorganisation on 18 July 1988, the Section had the following staff:
DT was successful in obtaining promotion during the year.
Work has continued on ESPRIT Project CAD Interfaces (CAD*I), which was described in last year's report. This has occupied MM, DT and JVM for most of their time, and BFC for about 20% of his time. Except for a small involvement in standardisation activities, RAL's participation in this project is due to end on 31 October 1988.
During the past year the work on CAD*I consisted of finalising the reference model and the software, and also the writing of the final report. The reference model has been completed and is able to describe finite element modelling, analysis and results. Work is progressing well on the software. The programs now finished or nearing completion enable the transfer of data from a geometry neutral file to FAMbuild, from FAMbuild to neutral file, from neutral file to Nastran, and from neutral file to FAMview. Non-exclusive licences have been granted to a partner in CAD*I (GfS, a company based at Aachen in West Germany) to incorporate some of this software into commercial products.
A successful International Workshop was organised by the section on behalf of CAD*I in Oxford, 21-22 March 1988. Out of 75 participants, 50 were from commercial and academic organisation not involved in CAD*I; most of the member states of the EEC were represented.
DT and JVM have continued to participate in International Standards Organisation committee meetings. This work enables us to influence the development of the STEP standard (STandard for Exchange of Product Information).
Progress has been made by MM on the development of a compiler for the data modelling language EXPRESS, which has been adopted by STEP.
BFC has spent about 40% of his time on work associated with finite element processing, interactive graphics and user interfaces for single user workstations. He wrote a GINO graphics emulator package for the SUN and, together with B W Henderson (Computational Modelling Division) produced a version for the IBM 6150. The SUN version relies on the ww interactive graphics library which was produced within the Division.
BFC has continued to collaborate with Vector Fields Ltd of Oxford. The GINO emulator has been used to implement the finite element electromagnetic programs PE2D, SCARPIA, TOSCA and OPERA (originally developed at RAL) on SUN workstations. Work is continuing on a new two dimensional potential modelling and analysis program, though progress has been slow due to the commercial success of the programs just mentioned and consequent demand for improvements and support from customers.
| 006-87 | 9 Nov 1987 | Report on ISO meeting in St Louis, USA, 12-16 Oct 1987 | JVM |
| 007-87 | 20 Nov 1987 | Report on ISO meeting in St Louis, USA, 12-16 Oct 1987 | DT |
| 008-87 | 20 Nov 1987 | Current Status of CAD*I and PDES Specifications for FE data exchange | DT |
| 009-87 | 1 Dec 1987 | Interfacing geometry to FEM using CAD*I; examples of practical Problems | MM |
| 010-87 | 18 Dec 1987 | Minutes of CAD*I WG6 meeting at GfS Aachen, 3-4 Dec 1987 | JVM |
| 011-87 | 18 Dec 1987 | Specification of Exchange of Product Analysis Data. Version 2 (draft) | JVM, MM, DT |
| 0001-88 | 26 Feb 1988 | Specification of Exchange of Product Analysis Data. Version 2 | JVM, MM, DT |
| 0002-88 | 25 Mar 1988 | FE Analysis Control - IDEF1X Reference Model | DT |
| 0003-88 | 15 Apr 1988 | Report on ISO meeting in Washington DC, USA, 28-31 Mar 1988 | JVM |
| 0004-88 | 15 Apr 1988 | ISO TC184/SC4/WG1 meeting in Washington. FEM committee report | DT |
| 0005-88 | 17 Jun 1988 | Substructuring in CAD*I and FAMbuild | MM |
| 0006-88 | 17 Jun 1988 | Substructuring in CAD*I and Nastran | DT |
The following meetings were attended:
| 7-8 Sep 1987 | Madrid | CAD*I WG6 | JVM, DT |
| 16 Sep 1987 | London | BSI AMT/4 | BFC |
| 8-9 Oct 1987 | Madrid | CAD*I Project Board | BFC |
| 12-16 Oct 1987 | St Louis | ISO TC184/SC4/WG1 | JVM, DT |
| 27 Oct 1987 | Nottingham | NAFEMS CAD/FE exchange WG | DT |
| 5-6 Nov 1987 | Aachen | ESPRIT Phase II Proposal | BFC |
| 26 Nov 1987 | London | BSI AMT/4 | DT |
| 3-4 Dec 1987 | Aachen | CAD*I WG6 | JVM, MM |
| 3-5 Dec 1987 | Abingdon | CFTAG Workshop | DT, BFC |
| 14-15 Jan 1988 | Aachen | ESPRIT Phase II Proposal | BFC |
| 25-29 Jan 1988 | Rotterdam | ISO TC184/SC4/WG1 | JVM |
| 8-9 Feb 1988 | RAL | CAD*I WG6 | JVM, MM, DT |
| 15 Feb 1988 | Birmingham | NAFEMS CAD/FE Exchange WG | DT |
| 22-23 Feb 1988 | Karlsruhe | CAD*I Project Board | JVM |
| 3-4 Mar 1988 | Aachen | ESPRIT Phase II Proposal | BFC |
| 21-22 Mar 1988 | Oxford | CAD*I International Workshop | JVM, DT, MM, BFC |
| 27-31 Mar 1988 | Washington | ISO TC184/SC4/WG1 | JVM, DT |
| 20-22 Apr 1988 | Leeds | CFTAG Workshop | BFC |
| 21-22 Apr 1988 | Madrid | CAD*I WG6 | MM, DT |
| 17 May 1988 | Brussels | CEN/CENELEC Conference | JVM |
| 18 May 1988 | RAL | NAFEMS CAD/FE Exchange WG | DT |
| 24 May 1988 | London | BSI AMT/4 | DT |
| 20-21 Jun 1988 | Copenhagen | CAD*I WG6 | JVM, MM |
| 27 Jun 1988 | Frankfurt | CAD*I Project Board | JVM |
| 20-21 Jun 1988 | Denver | ISO TC184/SC4/WG1 | JVM, DT |
| 6-11 Feb 1988 | Santa Cruz | Course on TIGER UIMS | BFC |
| 22-23 Feb 1988 | Uxbridge | METIS Tornado DBMS | DT |
| 28 Jun - 1 Jul 1988 | London | Relational Databases | DT, MM |
The policy and budget for much of the work of the Engineering Computing Group is determined by the Engineering Board's Computing Facilities Committee (CFC). The latter is advised on technical matters by the Computing Facilities Technical Advisory Group (CFTAG). The ECFE is responsible for providing input to these groups and for executing their decisions.
The ECFE does not correspond to a particular branch of the group and the following staff participated:
From ECFE MRJ attends CFC and MRJ, ADB, JRG, GAL (Secretary) attend CFTAG.
The major concerns of CFTAG have remained
CFC Meetings: 21/12/87, 29/6/88, 20/9/88.
CFTAG Meetings: 2-3/11/87, 3/3/88, 4-5/5/88, 5/9/88 (MRJ, ADB, JRG, GAL) .
Prime and the UK Academic Engineering Community 1976-1988 (MRJ) - a paper presented at the First World Congress of Prime Users held in Australia on 9-15 October 1988.
Engineering Computing Facilities (ECF)
Staff:
GAL acts as Secretary to the advisory group and in addition to the routine paper writing tasks associated with the role produced a report on the Applications Software Questionnaire which was published as RL 87-074. As an essential part of the Applications Software Review a series of four workshops were organised by GAL:
The reports from the workshops formed the basis for the software aspects of the EASE policy which was accepted, in principle, by the Engineering Board in July 1988.
GAL now has responsibility for the management of the project. RAL is prime contractor. Immediately after the project review meeting, held in November 1987, GEC indicated their intention of withdrawing totally from the project. This caused a period of feverish activity to find a substitute partner, STC Technology Ltd, and make all the necessary changes to the contract and forward details, before the end of the contract in March. In spite of all efforts final renewal of the contract was not completed until September.
GAL remains responsible for the initial scan of all computer related grant applications referred to RAL for comment. It is noticeable that an ever increasing number are requesting SUN based workstations and that those requesting resources on the ECF multi-user-mini computers are reduced to a trickle.
Fran Childs (FMC) continues to act as the editor of the Newsletter, with editions being produced at quarterly intervals. Production, against an exacting timetable remains a demanding task with well over 2,500 copies being distributed in each edition.
The decision was taken to bring forward the closure of the ECF MUMs, at a meeting of the community held at RAL in December 1987. The following timetable was accepted.
The responsibility and use of the RAL GEC community has been passed on to CCD.
Fran Childs (FMC) remains responsible for the loan pool of the Initiative. Activity has remained high with large injections of new equipment requiring check-out and distribution. FMC has also been involved in a series of displays etc publicising the activities of the Transputer Initiative.
One meeting of the Engineering Computer User Group (ECUG) was held (organised by GAL and FMC), however a second meeting was cancelled, due to lack of attendees.
GAL now has responsibility for all the contractual activity of the ECF, including all the hardware, software and support contracts. In addition there are an increasing number of EMR agreements being offered as a result of the EASE activity.
The purpose of the Operations Section is to provide operational services, on whatever hardware, to the Engineering Computing Community, to the Alvey community, and to the Department as a whole. In much the same way, the User Support Section provides assistance and advice to users of these facilities.
The staff in post at September 1987, the beginning of this period of report, was as follows:
Operations
Support
The major occurrence in this period, affecting the work of the Section, was the decision of the CFC, in December, to run down the use of the GEC and Prime multi-user-mini-computers, terminating the service by 31 March 1990. No new full 3-year allocations for such resources would be taken on and no new allocations beyond 31 December 1989 awarded. Closely following this decision, many machines were either closed or handed over to the University sites. It had already been decided to cease funding some uneconomic machines so that, during this period, GECs closed at Cambridge, in December and Cardiff in May; with Primes the dates were, East Anglia in December, City, Surrey and UCL in March, and UMIST and Warwick in May. These measures did not exclude RAL where both the Department's GECs went, one scrapped in December and the other transferred to CCD in May; the development Prime was transferred to UMIST in February.
The operating system on the GEC had already been frozen but now the same fate befell the Prime.
With the rapid growth of SUN workstations within ID and the imminent closure of the Alvey program, it was decided to close down the two DEC and two Systime VAX computers by March 1989. Already one of them, rl.vm, has been loaned, indefinitely, to NAG following the loss of their machine in a fire.
In September, MJL was the solitary person working on GECs. As System Manager, he had overall responsibility for the day to day running of the system, providing limited updates, hardware and software problem identification, and system security monitoring.
He provided user support by electronic mail, telephone and personal contact, and provided and maintained documentation.
He fixed all known faults in the Kermit file transfer protocol software.
He resigned in January, bringing to an end all GEC expertise and support within the Department.
In a similar way, PCP provided support for the Prime computers and their users.
He also assisted in resource management by designing, programming and administering two databases providing information for finance and management.
When he resigned in December, the software updating was transferred to ACD and the general user support to UMIST. Local support was taken on by MEC and PDA.
The third member of the Support Office, BAA, also resigned in December. Although all three members had their own specialist area of support, each was able to cover the other areas to some degree. With the UNIX service gaining momentum, in contrast to the GEC and Prime service, there was a need to replace BAA immediately and, to this end, NJK was thrown in at the deep end.
She took over the UNIX support on SUNs, VAXes and Pyramid and, in doing so, mounted Mk11 of the NAG Fortran Library on the Pyramid.
Taking over the post of Secretary of the monthly held UNIX Service Meeting, she has been responsible for the production and distribution of the Service Notes. In this series, she has written one as a quick reference guide to the Unipress emacs V2.l0.
During November and December, RP passed the responsibility and experience of Prime Managership of RLPA to PDA. Before moving into the Transputer Initiative, RP worked together with PDA in two particular tasks, namely the upgrade of RL. PH from a 2655 to a 2755, and the reformatting of all the discs to conform with the new and SERC final major release of Primos, Rev20.
In her role as both Manager and Resource Manager, PDA has since been involved in the move of grant-holders from the remote sites to RL.PA, especially from UCL where a complete disc had to be transferred. A replacement 2250 on a grant at UCL required her assistance as there is no longer a Manager at that site.
One of Prime's newest and largest discs was installed on RL.PG and this required necessary reformatting.
An outstanding BRMS (dumping procedure) problem has been, and still is, giving cause for concern.
A pilot program to look at forecasting for the Engineering Board has been developed and PDA is assisting in the process.
ACD, who was assistant to RP, transferred to PDA when she became Manager.
IV has been the System Administrator for the Pyramid, rl.pyr-a, the two production VAXes, rl.vc and rl.vd, and the Pyramid workcentre, nfs4. Apart from the overall responsibility for these machines with regard to performance, fault investigation, liaison with manufacturers for hardware and software faults, a major task was the installation of release 4.1 on nfs4 and rl. pyr-a. Due to the way in which Pyramid distribute their releases, involving the completion deletion of /usr, this was an undertaking with drastic repercussions for two weeks.
She has participated in the integration of rl.pyr-a and rl.vd into the Mail service.
IV has also handled the administrative maintenance of Netnews, updates of the NRS database, and Arpanet registrations and mail queries.
ASJ has assisted IV particularly in the operational support. When agreement was made to cease CCD involvement in machine dumping, ASJ took over the task to perform these with minimal disruption during prime shift. Subsequent file retrieval, on user demand, and files tore checking is her responsibility.
Analogous to IV, RT has served as System Administrator for the SUNs by maintaining user filestore, maintaining hardware and software through contact with SUN UK, installing new software, reconfiguring hardware and resolving operational problems.
Over and above the type of problems expected in an operational environment, he has been involved, generally in leadership of several related projects. These have been the major rearrangement to include additional server nfs5 and 24 clients, an upgrade to as to 3.4, installation of the ID Mail Service, the move of servers to comply with new Computer Room plan, and the addition of a further server nfs6. Preparations have started for the as release 4.0.
RT has been responsible for the day to day Lab 11 facilities.
Working on a part-time basis, LJR has been assistant Systems Administrator, primarily for the SUN workstations. This has included such tasks as helping to manage filestores, to check and maintain printers, to add discless clients and, on disc machines, to upgrade from 3.2 to 3.4, to install DWB software, and to install the DAP software. She has maintained the SUN on-line manual pages, added new users, and been responsible for issuing new user ids and group ids.
She has arranged in-house SUN user courses.
LJR has revised the IDUS Note on getting started for new users of IDUS.
MEC has been undertaking a survey of fire safes and of Postscript laser printers for attachment to the Ethernet.
Although Prime system work has been covered by a UMIST contract, the assessment of Prime UNIX (Primix) was made at RAL. Following user involvement from all sites and a summary by MEC, it was decided, on its performance, not to install the software.
The other major new software installed was JTMP.
AMJ, with assistance from ASJ and ACD on occasions, was responsible for linking the old Computer Applications Group of Technology Division into ID with regard to equipment needed individually and the merger of their partly owned terminal pool. An analogous exercise followed when the Image Processing Section was also merged, together with their Microvax computer.
He has, in effect, been the Systems Manager for the AMT DAP when this was purchased in December and eventually installed in Lab 11, linked to a SUN. He has served as liaison with AMT for hardware and software updates and queries.
Assistance has also been given with networking connections.
JRS has been responsible for the maintenance and upkeep of the many INFO databases held for use by ID. These include -
| Sept 7-11 | Prime System Principles & Tuning - Southampton | PDA |
| Sept 10 | CSMA/CD (Ethernet) Mechanism & Practice - London | AMJ |
| Sept 16 | Network File System - London | RT |
| Sept 29 | Managing Awkward People - Swindon | MJL |
| Nov 19-20 | AMT DAP Programming - London | BAA, MJL |
| Jan 25-29 | Pyramid System Administration - Farnborough | IV |
| Mar 7-10 | C Programming on Cray - Atlas | RT |
| Mar 23-25 | Introduction to UNIX - Harwell | ASJ |
| May 24-27 | Induction Course - Daresbury | RT |
| Sept 14-15 | INFO User Group Conference - Sheffield | PCP |
| Sept 21 | UMIST Prime User Meeting - UMIST | MEC |
| Oct 22 | 21 Oct 22 Systems Industry DEC Seminar - Abingdon | MEC |
| Nov 18 | Compec Exhibition - London | MEC, MJL |
| Dec 14-15 | UKUUG Workshop on Networking | IV |
| Jan 7 | SUN Seminar - Oxford | PDA |
| Mar 4 | Pyramid Upgrade Seminar - Farnborough | MEC, NJK |
| Mar 16-24 | Assistance to UCL Manager to move Users - London | PDA |
| May 5 | Pyramid User Group Meeting - London | NJK |
The section has been responsible for the support and development of Single User Systems for Engineering Board researchers and for the Alvey Programme, graphics on ECF machines, and text processing on ECF UNIX machines. Some of these activities are also carried out for the Departmental service IDUS.
During the year, Mark Phillips gained promotion to SO, and Martin Prime to HSO.
On Friday 13 May, Bill Hewitt and Mark Phillips suffered injuries in a car accident and were eventually able to return to the Lab in August and November, after recovering from their injuries. By the time they returned, Informatics reorganisation had taken place.
The assessment project for Single User Systems in EASE started in June and Tony Lucas, Subodh Chanda and David Barlow joined the section briefly until reorganisation took them into a new Assessment Team.
The team led by PK has been responsible for support of centrally purchased and maintained Single User Systems funded through Engineering Board committees and the Alvey Programme.
The team supports over 400 SUNs and a small number of remaining PERQs. PK took responsibility for advice to SUN people preparing a grant application and specifying SUN equipment for the committee secretariats. During the year some of this work was gradually handed over to KML.
All machine ranges are held online with the largest and most active set the SUN3s - being upkept by Janet Smith.
Maintenance is coordinated at RAL. The contracts procedure has been streamlined moving away from the old quarterly contracts to a long- term one, where the actual equipment lists are agreed between SUN and RAL on a frequent basis. PK and KML have been involved here.
The SUS Support Office telephone (with Atlas' Service Line screening out the hardware calls) has been staffed by WJH, MC and MDP.
Software distribution this year has mainly been concerned with providing new users with a standard software set and distributing new versions. NAG numerical library and NAG graphics has been licensed and distributed on request. New software expected to be distributed next year includes X-ll window system and the DWB text processing software suite.
With the large number of sites to distribute cartridges to, efforts have been made to set up a software and information distribution system accessible to any user connected to JANET (JAM or Janet Access Machine). Progress in setting up communication has been slow due to problems with the implementation of the X29 coloured books software.
MC and Jeremy Isserlis have spent much time testing SUN communication software and MC has set up file structures for allowing the user to find the information that is needed.
DCS led the Basic Software Team. The purpose of the team can be summed up as developing, providing and supporting widely usable basic systems providing common interfaces on which level software and applications may be built.
The work of providing a uniform and useful text processing system on the UNIX machines has been continued during the year. Documenter's Workbench 2.0 and Transcript 2.0 are now in normal use on all the IDUS gold service machines.
A survey on the usage of text processing macros was carried out by Nicola Watkins and resulted in the choice of mm as the major set of macros. Also arising from the survey, were produced some Informatics macros and profiles (mid) ; this allows Informatics papers, notes, memos and letters to be conveniently created.
As the text processing system evolved, IDUS Note 14 was kept up to date.
The result of this work (except the Informatics macros) was packaged up for ECF distribution.
The policy of using manufacturer's version of Transcript 2.0 was thrown into question by two events. One was the arrival of psfig, which allowed Postcript files to be inserted into DWB documents. The effective use of psfig needed an alteration to Transcript. The other event was the test and use of new Laserwriters. The technical policy favoured was to adopt Transcript source, alter it and compile it for IDUS Gold and ECF machines. Unfortunately staff shortages meant that this is impractical for the time being.
Within IDUS, the text processing policy has been guided by a text processing meeting, chaired by the IDUS Technical Manager, Bob Day. Chris Rust also contributed to the text processing software.
With the demise of the support for ICF Primes and GECs, conversion of graphics programs from GINO to GKS has achieved some prominence. DCS gave a talk on the subject to the special meeting of grant holders in December and wrote a letter to users, distributed with the Engineering Computing Newsletter. Little response was received from users despite their earlier worries. Some enhancements to RAL GKS have been included in the final release on the Primes to ease the transition (RMK).
Development of the master source of RAL GKS has continued this year, with several significant enhancements being completed.
PLP finished work on a driver for Tektronix emulators. This makes use of the scrolling text screen for all alphanumeric text, and also allows drawing in the background colour (ie erasing).
PLP added an extra choice input device to the Tektronix and Sigmex drivers to allow any printing character to be used for choice input. This ties in with the new ESCAPE functions designed and implemented by RMK. The purpose of these ESCAPEs is to associate GKS locator and choice devices, so that they can both be triggered by a single operator action. This allows the use of GINO CURSOR-style input, and has been implemented to ease the conversion of GINO-F software to GKS.
PLP has started work on a GKS driver which generates PostScript output. This is based on a driver written by a Starlink VAX/VMS user, which MJP converted to run on UNIX. PLP is correcting this driver and also enhancing it to make better use of postScript.
The new system interface routines for RAL GKS were designed and implemented on UNIX and Prime (RMK, TAW, PLP). These perform character and byte output to a file, and are already being used by the PostScript driver.
In August, RMK produced RAL GKS master source 1.20. This included all of the developments described above (apart from the postScript driver which has not been completed).
In addition to work on the master source of RAL GKS, libraries have been produced and released on various operating systems.
When PJWR left in January, RMK took over graphics on the Pyramid and released the Pyramid GKS library he had built.
RMK also released RAL GKS 1.20 on the Primes; this is intended to be the last GKS release on these machines.
In June, RMK released versions of the NAG Graphical Supplement for use with GKS and GINO-F on the Primes. These libraries allow a phased transition from GINO-based graphics to GKS.
Work is currently underway to provide NAG Graphics libraries for use with GKS on the SUNs and Pyramid.
Several diverse areas of work were covered during the year. The main areas of responsibility being GKS on SUNs, other graphics standard software on SUNs, emerging standard window management software and UIMS, and evaluation of high performance workstations. The team consisted of JH, MJP (half-time) and PJWR (until January 1988) and was headed by TAW.
This is an ongoing collaboration with Tony Arnold of Manchester University, Computer Graphics Unit. The Draft International Standard version of RALGKS was upgraded to the International Standard on SUNs. Several improvements were made to the SUN workstation driver. TAW added support for colour workstations and fuller functionality for line drawing. PJWR wrote many speed and functionality improvements for the output and control of the SUN workstation driver. Tony Arnold made several improvements to the input functionality. This was packaged up and given to support for distribution.
PJWR wrote a SUN raster image to GKS metafile converter and documented it. There was a need for some performance tests for GKS. These were written and tested by TAW.
An X11 workstation driver for RALGKS is being written as part of the same collaboration. A Postscript driver was ported to UNIX by MJP and modified to conform to the International Standard. Some of the Postscript code production routines were altered so that they would produce correct Postscript code. MJP developed some new test programs to augment existing tests and so fully tested the driver.
PJWR configured GKS-3D for use on SUNs at RAL. TAW then tested FIGARO and GKS - 3D and made them available for the Department. The new version now works within a SUNview window instead of taking over the whole screen. However this version still only corresponds to a very early version of the standard.
Towards the end of the year DCS organised a training course at RAL for potential recipients of the four licences purchased for external use and assessment. it was given by UIS, the suppliers of GKSGRAL-3D in the UK and was attended by five external users and two RAL staff. Just before the course, but too late to postpone, it was learnt that a new version corresponding to the DIS was imminent and would be sent to RAL when available. This version will be distributed to the trial sites.
Standard window systems are becoming important so work was needed to provide EASE users what they need and increase their awareness of what they can do. This work started in January with the trip to the X Conference at MIT. TAW wrote some XII benchmarks and ported an application to Xlib to be used by the EASE evaluation. Having gained some experience of using X11 TAW planned future work on XII needed to support EASE users. This was documented in a paper and a SUN user note informing users of our intentions. The MIT XII release 2 on SUNs was too slow to use so TAW implemented some speedups designed by CAAG and incorporated several known fixes to the code. This will be made available to EASE users.
The UIMS assessment was instigated by the CFTAG workshop on UIMS held in September 1987. This recommended that two commercial UIMS (PASET and TIGER) should be evaluated for use in an engineering environment. The aim of the evaluation was to determine the ease of converting a reasonably sized (approximately 15,000 lines of FORTRAN) existing engineering program to use each UIMS and to determine the advantages and disadvantages of using the UIMS.
It was subsequently discovered that another UIMS under consideration, BLOX, had been purchased by a US graphics' software house and would be supported and developed. This was added to the evaluation.
DCS has managed the assessment which is being carried out by UMIST (PASET and BLOX) and RAL (TIGER) staff and MJP has coordinated the daily running of this evaluation.
MJP has continued work on the survey of UIMS, which compared proprietary systems and evaluated them against a set of criteria which he produced from papers in the subject area. He presented the conclusions as a paper at the CFTAG work shop and the survey provided a useful focus to the workshop's deliberations. The paper was then updated and published as a RAL report (RAL 88 -028) User Interface Management Systems a current product review.
This work carried out by JH is reported on elsewhere.
TAW and MJP prepared part of this course, on using the window system and gave it a few times to members of the Department and UMIST.
The Single User System Assessment project became a major project in June. This followed preparatory work by Eric Thomas (RET) who prepared the questionnaire with inputs from many people and JRG, DCS, PK and WJH. The project clearly needed more effort and AJL, SKC and DSB joined the section. Since reorganisation was known to be imminent both RET and JRG were actively involved. JRG chaired project meetings before the reorganisation and RET afterwards, with one standing in for the other during holidays, thus ensuring continuity.
DCS assisted in the design of the questionnaire and benchmarks. TAW prepared X11 benchmarks. PK produced initial filter mechanisms to remove the no-hopers and also advised on statistical methods.
After reorganisation, the assessment became the first project of an enlarged assessment team under Mike Claringbold/Bob Day, who became heavily involved.
Bob Day coordinated the paper, which Eric Thomas presented to the September CFTAG meeting.
JRG participated in CFTAG and ECFE and took a major role in formulating EASE in response to inputs from many sources. He also presented the new plans to various bodies such as IUSC and CHEST thus starting what should be a useful cooperation with them.
With Bob Day and Dave Duce, JRG and DCS represented RAL in establishing the ARGOSI Esprit II proposal, which has been looked on with favour by the Commission, although they have said combine with another proposal and, by the way, there's only half the money you asked for!
The London and South East Regional Transputer Centre was set up on 1 April 1988, after existing as an interim facility since September 1987. The Centre is now located in Building R1, Lab 10. Day-to-day operation of the centre, and the overall provision of services to academics, laboratory staff and industry is the responsibility of the manager, R Parkes (RP). RP took over responsibility for the move of the facilities from the previous accommodation in R32 and the opportunity was seized to upgrade the appearance of the centre. This was accomplished at minimal expense, in particular, a fortuitous flood in the area resulted in the laying of new carpet replacing the existing rather grubby covering. The centre has now been repainted and equipped with better lighting, fittings and furniture than existed in the interim facility. Given the financial restraints the area is now considered to be suitable for its purpose. The centre is equipped with 5 IBM PC computers containing a single T400 transputer together with development software such as the TDS (Transputer Development System) which includes a novel folding editor, and also sequential versions of the scientific languages, Fortran and C. Two of the machines are used as front ends to multi-transputer systems manufactured by Meiko and INMOS. These systems also have high speed colour graphics capability. A number of demonstration programs are also provided.
Much emphasis is placed on the search for customers from industry, who are particularly sought after, in view of their ability to bring funding into the centre, which is supposed to be self-sufficient by 1990. Progress has been steady, if not exactly spectacular, and much importance is attached to the value of advertising. The momentum is being maintained nevertheless. In July of this year our first seminar was run at the laboratory, with the aim of introducing Company Management into the idea of the Transputer, and parallel processing. The seminar was well received and it is expected that at least one other will be run later in the year, targeted at a similar group. it is hoped that lucrative activity within the centre will be generated as a result. The Centre also plans to run a more technical follow-up in the Autumn.
The Initiative exhibited recently at the User Show, which has resulted in a number of contacts in industry from which some new sponsorship deals should soon be concluded.
Usage of the centre has been sporadic; which is rather disappointing in view of the widespread advertising. Perhaps it is yet still too early to draw implications. The indications that industry will take up services offered by the centre are in evidence and uptake of services, should soon begin to accelerate. In expectation of this databases are being designed to provide the management information that will be required later.
Other general issues that concern the operation of any service such as this are slowly being addressed, such as bookings, security, technical documentation, the way machines are used and what sort of operating/usage procedures should be in existence. At the moment these are being developed according to the actual demands that are being made. It has certainly be necessary to introduce a booking scheme and the issue of software security is one that will be addressed as a priority.
Other areas of activity planned for the immediate future are the development of a standard user environment on the PCs for this centre consisting of standard utilities, compilers, development environments, eg TDS and TDT, packages, libraries, utilities and possibly online tutorials. The development of a business plan to produce an income of £100,000 per year is also a priority issue. The Centre is already in a position to charge for some of its services, some chargeable hands-on use has already been invoiced for.
At the start of October 1987, CB moved from Alvey Infrastructure Coordination to the SERC/DTI Initiative in the Engineering Applications of Transputers. The envisaged role was that of Technical Administrator with responsibilities for contracts, loans of equipment, liaison with Transputer Centres etc. During this period, CB attended a small number of conferences and courses aimed at establishing a transputer background.
Initial activities were in supporting the Transputer Applications Management Group (TAMG) and the Support Centre Directors Meetings (SCDM). In addition a number of development contracts were established in support of the Transputer Software Base with various universities. A round of purchases and equipment upgrades added interest to the end of financial year, as did the opening of the National Transputer Support Centre at Sheffield, on which CB was a member of the organising committee.
In the new financial year, CB took on additional responsibilities as Deputy Coordinator for the Transputer Initiative and became involved with the marketing of the Initiative services. This has led to participation in and organisation of a number of shows and seminars in terms of the Initiative or the RAL Transputer Centre. Included in this has been participation in the Royal Society Soiree, involvement in the arrangements for a publicity video currently being made on behalf of the Initiative, speaking at events such as the Transputer Forum at the PC User Show, and the IEE Colloquium on Recent Advances in Parallel Processing for Control, and participation in the Enterprise 88 event in Newbury on behalf of the RAL Centre.
There have been several meetings with potential clients of the Centre based on follow-ups from these events and other publicity, with CB involvement. There have also been discussions with potential sponsors for the Initiative.
On the academic front CB is involved with universities and companies in setting up of development contracts aimed at adding to the base of transputer software. CB is also involved with the overall management of the Loan Pool, including assessment of new applications, and in production of the Initiative mailshot.
A UK Initiative in the Engineering Applications of Transputers (MRJ) - a paper presented at the Prime International Conference on Computer Graphics held in Singapore on 15-16 September 1988.
To evaluate the 3 top of the range workstations from Hewlett-Packard, SUN Microsystems and Apollo Computers JH designed some 3D graphics benchmarks aimed to stretch the capabilities of each. For each machine JH had to learn a new graphics language, and in the case of the Apollo, the language was also being extended during the evaluation. For HP and Apollo, the graphics package claimed to be based on PHIGS, whilst for SUN based on Core. In the case of the latter two machines JH suffered problems with unavailability of the necessary hardware and software. JH soon discovered that despite upwards compatibility trying to develop software on lower range architectures is not always easy or recommended.
The HP 350SRX arrived mid-September, and proved relatively easy to work with - JH only suffered one upgrade to the operating system and one period when the machine was out of action due to hardware failure. After the upgrade the graphics package came with a good set of example programs and on reading the small print from an online manual. Apart from not succeeding to convert the tests to run within a window (which was the result of incompatible software for the given hardware) JH had reasonably completed the tests by January 1988.
The DN590T (though for most of the time a DN580 or DN580T). Throughout most of the evaluation JH had an Apollo on loan. However, only for a short period of time did it have the required 24 colour plane, with floating point coprocessor, for the rest of the time an 8 colour plane model was used. The initial release of graphics software did not contain such things as shading and the machine, being one of their demonstration machines, was liable to disappear for a while and return in an unsuitable state for my evaluation. Problems included lack of example programs and lack of reasonable documentation.
SUN-4/260C (though for most of the time a SUN-3/260C). Again JH was hampered by lack of example programs, poor documentation, changes in operating system and unavailability of the desired hardware. Not being the only contact with manufacturers for this machine caused problems in knowing what state the machine is in.
Mondays, during term-time JH has attended Middlesex Polytechnic to study for an MSc in Computer Graphics. Since a term was used at the beginning of the course to bring everyone to the same level of competence in mathematics and programming the end of first year exams were part way through the second year, in December 1987. She passed in all subjects despite suffering from an abscess (as discovered in a later trip to the dentist) and obtained As for the courseworks. She has another set of exams in December 1988 then 2 terms to complete a dissertation on a chosen topic.
JH installs the latest software and assist users to the Departments Apollo DN3000 and other Apollo at RAL. She is the Laboratory contact with Apollo and keeps up to date with developments via the Apollo User Group, the Network news and the Press.
JH maintains the Department's Macintosh Plus, assists users and advises on the purchase of hardware and software. She also keeps up to date with developments via the Macintosh User Group, the Network news and the Press.
| 28 Oct 1987 | London | The Future of Graphics S/W (BCS) |
| 26 Dec 1987 | London | Realism and Visualisation (BCS) |
| 26 Mar 1988 | London | Graphics Tools for Software Engineering: Visual Programming & Program Visualisation (BCS) |
| 5 May 1988 | London | Visualisation of Fields in Three Dimensions (IEE) |
The Infrastructure Group which disappeared in the re-organisation of 18 July was led by K F Hartley (KFH) during the report period. Most of the work was funded by the Alvey programme under the headings of Infrastructure and Coordination and Support. Technical support of IKBS and Software Engineering was provided by SKE Group. Some research in HCI and Image Processing was funded by Alvey and Engineering Board. The other major responsibility was for Systems and Communications support of Engineering Board systems.
The following sections describe the work done during the report period, but 4.3 is somewhat different. It contains the full text submitted for the 1988 Alvey Annual Report. As it is understood that it is likely to be reduced substantially before publication the opportunity is taken to include it here for the record. It contains a review of the Department's support of the Alvey Programme throughout its lifetime. As the Alvey Directorate no longer exists and Informatics Department support will end in March 1989, though some Alvey research projects will continue for a further two years, it was felt that this was the right time for such an overview.
KFH's main activities during the year have centred on the new joint SERC/DTI collaborative programme (still without a name, though lED are running a competition to find one). Comments were prepared on the draft strategy documents. Attempts have been made to persuade those responsible that the new programme must have an infrastructure, to define what it should contain and to try to ensure that the Department obtains a reasonable share of the funding. So far, all of these have proved to be frustrating and inconclusive. The decisions are expected to be made by December 1988.
Substantial effort has been needed to wind down the Alvey activities by re-deploying staff and redistributing or ending activities. This is never easy but reasonable progress has been made without too many casualties amongst the staff involved. For once resignations have actually helped, even though the skills could not be spared. Continued uncertainty over future funding probably contributed in some cases.
KFH has also contributed to management's response to the Management Services Review of Informatics Department. This started before, and will be completed after, the present report period. Hopefully some errors have been corrected; certainly some of the conclusions have been strenuously challenged. We await the outcome with interest and some patience.
KFH and the whole Group have been ably and loyally supported by Miss H V Jenkins, between bouts of injury and illness, and Mrs J Perkins nee Smith, between bouts of house purchase and marriage.
KFH attended the IED conference at Swansea in July and a Bow Group conference on the Exploitation Gap.
SERC's Rutherford Appleton Laboratory (RAL) has been providing support for the Alvey programme since its inception. As this may be the last Alvey Annual Report it is appropriate to review the origins, scope, achievements and lessons learned from this involvement.
The Alvey Report was published in September 1982 and the Programme started in June 1983. During that period RAL was involved in several SERC initiatives which were directly relevant to Alvey. The most important of these were:-
These activities meant that RAL had experienced staff in post who were able to give immediate assistance to the Directorate right at the start of the programme.
As the various Directors turned to SERC and RAL for extra effort different functions emerged and evolved over the years in the areas of IKBS, SE and MMI. RAL's work for the VLSI programme is described elsewhere. The scope of these activities eventually included:-
In many respects the achievements of this kind of work are invisible meetings happen, minutes appear, invoices are paid, mailshots arrive, contracts run and so on. They only become visible when something fails. However, the development work undertaken by RAL for delivery to the Alvey community has led to visible results; some of the highlights are discussed below.
These activities have taught us many things which should be incorporated into the joint collaborative programme.
The lessons learned tend to be negative - things which could be done better next time. However the infrastructure as a whole made many contributions to the programme and to the concepts which are now accepted as part of the normal culture. Some of these are:-
At the start of this period, the section membership was:
Arrivals
Promotions
Departures
At the end of this period, the section membership was:
The overall objectives of the section are:
The following shows the responsibilities of individuals within the Section:
This year saw a considerable change in the role of the Systems team. Classical Systems activities grew less and less. Work on the Primes came to a stop, and very little effort was required on VAXes and Pyramid (apart from Communications work, see later). UTS was run down and closed. The Alvey Infrastructure requirements were few. Instead, effort went into the IBM 6l50/Transputer link and the connection of the Linotronics 100 to the Departmental system. Effort was also expended in defining the EASE Questionnaire and putting together the benchmark suite. At one stage, it looked as if some of the ECSTASY work would also be undertaken, but unfortunately MLJ left and the work had to be moved elsewhere.
During the autumn of 1987, a pilot Primix service was set up by MHR on Prime-a to allow members of the user community to determine how useful Primix would be in their work. Following this, the Prime General Meeting agreed not to accept Primix for a full user service, and all work on Primix was wound up.
RET continued to chair SDM until the decision was taken to run down most of the machines in 1988. This effectively put a stop to any more development work. Since UMIST provide Systems support for the Primes, there was no further involvement required.
UTS had never had a large user population, and it was proving expensive to continue running it. A plan put forward by HKFY on UTS closure was accepted by both CCD and ID, and the service was terminated (with extreme prejudice) at the end of February 1988. The two main proposals in the closure plan were that (a) the computing requirements of the Cambridge users would be met by a SUN workstation, and (b) the 4250 service would be replaced by the Linotron service.
DGH and CDR were involved in a project to connect the Linotronics 100 phototypesetter, purchased by Reprographics, to the Departmental Service, thus providing on-line access. The work included the RISCS Interface (used by Repro to control the altered spooling system); alterations to the spool system on a SUN; connecting the Macintosh to a SUN, transferring files between the two; and investigating ways of creating PostScript files on a Mac that could print on laserwriters. A set of user commands based largely on the existing lpr, lpq and lprm commands, has been provided. The one major purpose of this implementation has been to prevent files from being automatically printed. A system has been devised where the control files can be manipulated so that the operator may move files between directories, change expiry and due-to-be-printed dates, and selectively print pages from files. In addition, a service was to be provided for IBM-hosted users; this has been delayed until IBM provide their own suitable software. CCD have postponed their development of this and are waiting for the IBM code. Otherwise, the project is now complete (though it seems little used).
HKFY, MHR and BWH have been heavily involved with the IBM 6150 system. The machine already at RAL underwent a number of changes of Operating system, together with a series of upgrades. It is now running AIX 2.1.2 (and will shortly go to AIX 2.2), and was changed from a model 25 (2.1 mips) to a model 125 (4.5 mips) with hardware floating point and an extra disc. It also has a 3278 emulation board (allowing it to be used as a PROFS terminal) and TCP/IP ethernet. This last item has caused some problems to the Service, and to date these have not been fully resolved. The current version of AIX includes BSD sockets, required for IBM's implementation of X-Windows.
In November, outline agreement was reached with IBM on a proposed Joint Study, which would involve connecting Transputers to the 6150 and reporting on the effect of using this system in an Engineering environment. RET organised the paperwork, and tried to chase this through the various systems. This has taken a very long time, including the involvement of senior people in Swindon. The agreement was finally signed by SERC in June. IBM's part of the deal is to provide three machines, with a variety of peripherals. These have now begun to arrive.
In anticipation of this work, investigations into Transputers on the 6150 started much earlier. The Inmos TDS system was ported, and integrated into the Unix environment. A single transputer board was successfully connected, and a version of the Mandelbrot program run (not very impressive visually on a black and white screen though!).
Modifications to the algorithm and using 5 transputers, it was possible to obtain a five-fold increase in speed. The Gaussian elimination algorithm from BIM2D was converted to run over a network of transputers, and BIM2D itself was run using this algorithm. There was a long delay before Inmos' FORTRAN compiler was available, which allowed engineering applications to be developed in a high level language (rather than Occam). In addition, the Occam toolkit had to be mounted, and the MSDOS interfaces replaced. A version of BIM2D was successfully run, and one trial involving a 32-transputer board has produced a system seventy times faster than that obtained on a single T4.
Following the definition of EASE, work was started in planning the first of the assessments which would be required by the new policy. Systems had a full share of this, providing benchmark programs for Languages (Fortran, C and Pascal) and Unix Systems (AIM2, Byte, Spiral and sysv_interface). In addition, HKFY was responsible for producing the first draft of the questionnaire, which would be used to elicit relevant information from suppliers in as painless a manner as possible. RET spent some time interacting with EC Group to refine this questionnaire as the EASE requirements changed.
Once the initial set of test routines had been provided (and other members of the Department bullied into producing applications benchmarks), the set was documented and handed over to EC Group for running. Further tests were provided by them to cover X-Windows, and by NBMC to evaluate NFS. This last was considered very necessary, following experiences with different manufacturers.
The actual Questionnaire was sent out in June, and most of the replies have been received. Benchmark machines have also been obtained, and much testing undertaken. It is intended to provide a paper for CFTAG in September, giving a general view of the ability of the contenders to perform in an Engineering environment.
The Communications Section suffered the loss of three members during the year. Ian Gunn left to take up a programming job with Reuters, and Tony Lowe transferred to A&G Division to join a project studying the behaviour of ocean waves by radar. Mike Woods left in February to become an IBM PC expert. In addition, IJJ has had considerably mixed fortunes, having had a severe bout of flu during the winter, and then being involved in a serious car accident. All this has made it very difficult to carry out the work required. On the credit side Andrew McDermott transferred from Graphics Section; he has taken over the management of the LAN from Tony. DFS has proved to be a great asset also.
Since ABL's departure, most of this work has been undertaken by RAD and APM. Much of the requirement has come from the office moves. The main work has been the installation of thin ethernet in R1, and the R25 connections. The interface to the RAL Site LAN required a complete rewiring of the ID Comms rack. In order to help keep track of all this, APM has set up a database giving details of the connections of machines to the ether.
RAD and APM have investigated a LAN modelling package called PLEXUS from Insight International. This allows a model to be built up of a LAN and the hosts attached to it, and gives details of LAN traffic and host loads. The advantage of using such a package is that it might be possible to plan for file server upgrades rationally, rather than by the ' finger in the air' approach used at present. For the evaluation a model was set up of nfs5 and all its discless clients. The results were encouraging enough to suggest a continuation of the work. The HP Monitor has proved very powerful in taking measurements.
RAD and APM have been involved in the setting up of the Site LAN. The installation moved on fairly slowly; meanwhile everyone tried to work out how to manage it once it was commissioned. RAD's contribution was to investigate the running of IP protocols over the LAN (of increasing interest as other Departments get SUN workstations). He produced some traffic analyses, and suggestions as to how things might be managed. These Management Guidelines based on our own experiences of running the ID LAN found favour with the Management Committee (a slight case of grasping anything that fills the vacuum!), and are now being tidied up to become the first stab at solving at least the human side of managing a potentially unmanageable system.
RAD negotiated for Informatics to provide advice and help to other Departments when installing IP equipment and software, in return for manpower from CCD. Help has also been given to those working on PC/NFS for the TECHELEC and Neutron Division Villages.
APM is now collecting statistics for the number of packets sent over the Bridge IB/2 onto the Backbone. Other Villages use DEC LAN Bridge 100s which are not capable of all the functions required by the ID Village but can be accessed centrally by CCD. It is therefore easy for CCD to gather statistics from other villages but not from ID, so we have to send them to CCD manually.
The JNT announced that they were looking for an organisation to evaluate MAC bridges with a view to providing advice to the UK academic community on which bridges to purchase. As this is an area in which Communications Section has an interest as well as some expertise (gained when we undertook the evaluation of bridges for the RAL Site LAN last year), RAD prepared a bid to do this work, which was accepted. This means that for the next two years the Section will perform a continuous assessment of Ethernet bridges, with a view to providing the list of approved bridges that JNT will recommend for UGC purchases. York officially starts on this on the 1st August; once again APM will be involved, along with RAD.
During April and May APM helped RAD with the evaluation of the BICC ISOLAN Primary Bridge which was subsequently bought from FOSS and is now in place as the bridge between Red and Green Ethernets.
Both IJJ and PJI were involved in trying to get SUNlink X.25 version 5 and the Coloured Books package working. Eventually, the Yellow Book Transport Service software was acknowledged by SUN to be bad and replaced. The Coloured Books Service provides users of SUNs with PAD and File Transfer facilities over JANET. The implementation is such that the access to these facilities is transparent to the user : previously, users had to login in to a machine such as Pyramid-a to use these facilities. IJJ achieved this by using the REX protocol which SUN and Pyramid supply. This RPC protocol allows the remote execution of commands on a server, and ensures that the remote command inherits the same context as the client process.
IJJ commissioned X.25 on the Pyramid WorkCenter nfs4 after it was upgraded to run OSx 4.1. This was not straightforward, as the X.25 distribution tape from Pyramid contained new versions of some programs, now with added bugs! IJJ had to back-off the new programs, and re-install earlier versions from Pyramid-a.
IJJ attended a meeting of the Joint Network Team LAN Advisory Group on January 7. The aim of this group is to provide input to section B of the JNT on technical and strategic matters concerning LANs in the UK Academic Community. The work that members of the LAG undertake includes evaluating products, providing input to standards committees etc. In representing RAL ID, IJJ can try to make sure that LAG decisions are relevant to the networking environment in Informatics. This is important because most of the LAG members come from the large mainframe, connection-oriented mould.
PJI has continued his work on the departmental mail system to handle electronic mail to the unix based workstations and, eventually, the unix system multi-user minis. So far this is based around the notorious sendmail package and SUN's Networked File System, NFS. Diversions along the way were concerned mainly with acting as an unofficial alpha test site for SUN's chargeable (and paid for) networking products, by now based on SUNlink X.25 version 5 and the York Coloured Book software as resold by SUN. The system now handles mail for all the IDUS machines, dealing with an average of 700+ messages per day.
Further developments included: improving the mail headers, providing client user commands to inspect the progress of X.25 mail leaving the mail server and providing mail access (with authorisation) to sites through the PSS gateway from sendmail via standard user mail interface programs. Some faults in mail address handling were corrected and the Departmental Mail System now knows about the common ISO international domain names.
NBMC has continued with the distribution of NFS to the Alvey sites, and NFS was installed at Edinburgh (the final site) in December. Bug reports have continued to come in and these have been fixed where possible. MIW also assisted NBC in some of the VAX work.
Some time was spent by NBMC aligning uids and gids on Vax-f for Yellow Pages, and helping MLJ with NFS on the ULTRIX machine (Vax-m). In addition NBC has produced a comprehensive guide to NFS and Yellow Pages (CTN/P19). A precis of this was issued as an IDUS paper. NBMC has prepared and given a talk on NFS at a number of New Users Courses.
NBMC went to Sussex to help them reinstall NFS on their VAX. NBMC also prepared for and installed NFS on Vax-c at RAL. He has recompiled and installed a number of utilities on the ID VAXs which had been affected by some missing bug fixes in the libc library. These remade utilities were sent out to the Alvey sites with VAXs. In addition, NBC prepared and distributed the binary files for NFS so that sites could attempt to build their own kernel.
NBMC prepared for running the NFS evaluation test suite on the Apollo, helped run the tests, and investigated some of the problems that arose. A paper on the results and remaining problems with Apollo's NFS was produced.
APM and NBMC have begun work on implementing Yellow Pages on the Pyramid. A client only service has been established on nfs4 and login altered to permit YP access. The YP utilities have been tested with the new release of OSX4.1. Problems with the passwd program were fixed. With help from APM the system utilities have been recompiled to work with YP.
RAD and NBMC evaluated a new 3rd party controller for SUN fileservers. Called the Rimfire, it was claimed that the controller would outperform the standard SUN offering by a factor of two. This might be true for test cases; unfortunately when it was tried with the NFS benchmarks it was found that its performance at high loads was actually worse than the SUN controller.
DFS has continued his work with the OSI standards. This has involved reading and summarising the protocol specifications, attending meetings, and delivering seminars. The first half of the X.400 recommendation summaries have now been completed. He has joined the IGOSIS IG5 and IG8 groups (Message Handling and Directory Services respectively), and is attending meetings of both groups. He has attended conferences and seminars on X.400, and attended courses on ASN.1 and X.500.
During December, negotiations commenced with UCL to get ourselves involved with a project on directory services. The project is called THORN (THe Obviously Required Name-server). As a prerequisite, the ISODE (ISO Development Environment) from Northrop needed to be installed.
In February, DFS took delivery of the THORN code from UCL, and proceeded to install it. He was requested to document any problems that were encountered, and report back. DFS modified the installation documents and returned these to UCL for perusal. Next, he was asked to help with making comparisons on the time delays between using X.2S and TCP/IP over Ethernet for accessing the directory. It is now possible for us to bulk load data across the network into the UCL directory.
RAD has been involved in organising a small project with KMT and DFS to attempt to use X.500 in a real application - as a way of allowing IDUS users to control their choice of mailbox within the IDUS Mail Service. This involves a fairly loose interface of the UCL THORN code to sendmail; KMT carried out the actual work on this project.
Various new versions have appeared at irregular intervals, and DFS has been the guinea-pig used to see if each version can be installed by an Australian-standard idiot (thus certifying the installation procedure as proof against the British-standard version), and checking that the stuff actually works once installed. He has been able to feed back a lot of "helpful" comments to UCL in both respects.
DFS has been progressing with the task of integrating the X.500 directory with the ELM mailer. This is a test bed to see how a directory should (or shouldn't) be used within a mail interface. This is getting to be quite interesting, the essentially boring bits of just getting the two to interwork having now been overcome.
NBMC spent quite a lot of time becoming familiar with the Usenet News system and uucp, following IJG's departure. As a result of this several changes were made to the newsclean procedure on Vax-d to make expiration of news more efficient and reliable. An investigation was made to see whether expire could be made to run on nfs3 but this proved to be not feasible. However, the exercise did provide further useful knowledge about the News system. As a result of this, changes have been made to the News-posting program inews, and the News-reading program rn so that articles reflect the correct mail address (for rl.inf, the new mail server) in the headers. Work is now proceeding to migrate News to Pyramid-a, in preparation for the withdrawal of Vax-d from service next year.
Along with RAD and DFS, IJJ attended Networkshop 16 in rain-swept Reading from March 22-24. The whole attitude of this meeting was different from last year, as the number of SUNs and other UNIX systems in the community has greatly increased. For IJJ, the most noteworthy point was Roger Evans, head of the Supercomputer Unit at RAL, getting very enthusiastic about running RPC-based applications between UNIX workstations and the Cray. For RAD, the highlight was to hear the JNT finally accept that distributed file systems are important, and that they will have to make recommendations concerning them without recourse to OSI protocols.
DFS was immediately involved with setting up a demonstration of the THORN project at Networkshop, following his brief visit to Oz. This involved building a system, installing the code, and setting up the completed product at the site. The demonstration was a great success.
RAD (with DFS) attended European Networkshop for the first time. These are organised by RARE, and this year's took place in Les Diablerets, a small village in the Swiss Alps. The main topic of the workshop was the proposed specification for COSINE (the intended European OSI network), along with plenty of opinions on how to implement it (or, more accurately, on how it *shouldn't* be implemented!). There were some interesting semi-technical papers too. The visit was complicated by the fact that (a) the JNT lost the applications; (b) DFS was being paid for by the University of Queensland, not RAL; (c) the University of Queensland basically believes travel should be free (hitch-hiking?), necessitating frequent referral to the ox-cart timetables etc. RAD & DFS took the opportunity whilst "almost there" to visit IBM's European Networking Centre in Heidelberg & find out about IBM's forthcoming products in X.400 mail.
RAD became involved in an ESPRIT II project proposal with DAD, JRG and DCS. RAL was originally invited to join a project with the unlikely aim of transferring Computer Graphics Metafiles across X.400 networks, mainly because the latter were there. RAD attended a meeting in Berlin where the real aim of the project was properly identified. This is to study how one uses OSI networks to achieve distribution of graphics applications, a subject not at all well understood at present. This proposal is now going ahead, and RAD is participating with the aim of eventually bringing to the section the task of handling the networking issues involved in the project, called ARGOSI (ARGOSI stands for Applications Related Graphics and Open Systems Integration; every ESPRIT proposal must have a suitably meaningless acronym!). Preparing the proposal meant a considerable number of meetings scattered around Europe Paris (January), Rome (February), Abingdon (March) and finally Paris again at the end of March. RAD also organised the Abingdon meeting, held over a weekend in the Abingdon Lodge Hotel. The ARGOSI proposal was finally kicked into shape by the end of March, allowing it to be submitted to the Commission 5 days before the April 12th deadline - something of a record. The current state of play is that the Commission probably will fund it (albeit with only 50% of the budget requested). RAD and DAD spent some eventful days in a variety of European locations rewriting the budget and persuading everyone to agree to it.
RAD spent a small amount of time setting up a SUN system to access the Atlas Cray for the purpose of job submission/retrieval. The rationale for this was twofold: a desire by CFTAG to assess the feasibility of providing "fast batch" access to the Cray, and a desire by CCD to gain experience of Unix access to the Cray, in anticipation of UNICOS developments in particular. The work involved connecting a SUN to the Hyperchannel that currently connects the Cray and Vax/VMS workstation and installing the appropriate drivers and Craystation software. This proved successful and the system has had some limited but regular use. It is planned to expand the usability of the system in the light of a grant award to the Dept of Mechanical Engineering at Sheffield University, who wish to access the Cray in this way.
This year saw the beginning of the end of the Alvey Infrastructure. In December, the Alvey Board decided to close the GECs, and to cut the funding by half in the next financial year. Hints of this had already been received earlier, so the October Site Managers Meeting was very low key. Ironically, the January meeting (in Brighton) was quite lively.
Disposal of the GECs proved easier than anticipated, as Daresbury had a use for them, and transfer to them was within the rules. Only Essex have decided to continue to run the machines. A greater problem was the EUCS Support contract. It proved necessary to invoke the termination clause, which is not something one wants to do at the best of times. Again, it was ironic that EVCS just managed to complete the NFS work before the machines were closed. If only everything had been two year's earlier.
The year also recorded two firsts: we managed to complete all the reconciliations and get all the bills in before the end of the financial year, and PAJ managed to produce a paper containing details of the statistics from all the machines. PJH and PAJ organised the Site meetings, and PJH and RET carried out the reconciliation visits. Agreement on next year's budget was also reached, although it was only possible to save £40K. In one case, the amount agreed for 1988/89 was more than previous years (since those without machines did not have to pay the 10% usage fee).
In April, RET and PJH were forced into a considerable amount of digging, caused by a request for data by the Public Accounts Committee, which was investigating the National Audit Office report. The greatest problem was ensuring that all the figures presented from the various different areas (both subject and geographical) were consistent. Eventually, since nothing further was heard, it was assumed that the figures were accepted.
RAD was involved with JRG in the ECF's Operational Requirement and subsequent Tender for Central Servers. These are defined to be machines capable of supplying NFS and paging services for single-user workstations, along with providing good communications to JANET and to University campus networks. CFC wants to place Central Servers in University Departments where a significant number of workstations are either already present or will appear as the result of new grants. Using a Central Server means that economies can be made by avoiding expensive purchases of discs for individual workstations, as well as giving the workstation users access to a better range of services.
As a result of this work SUN and Pyramid were selected as potential suppliers, and a Pyramid WorkCenter purchased for further evaluation. CFC has now made funds available for the first batch of Servers, and bids nave been invited for their placement. RAD is assisting in the evaluation of these bids.
The running of a recognised Departmental service must be considered one of the success stories of the year. A small management team chaired by KFH, with RET as secretary, handled the project. MEC reported on the service aspects, and RAD was in charge of the Technical projects. Further management input came from USM (service), ULM (strategy) and IDUUF (user forum). It proved possible to agree budgets and future work plans.
RAD has run the text processing project (which saw the establishment of its basic DWB service), and the mail service project (very much a Comms. Section show, with most of the work done by PJI and described elsewhere). RAD also purchased a new fileserver for IDUS and produced a plan for the rationalisation of the VAX systems in the Department in the light of the rundown of Alvey finance. It was decided that we should scrap everything except Vax-d. Plans are also in hand for the removal of the last SUN2s from the Gold service.
A security breach was noted, and this has caused work to start on a proper Security Policy document.
Negotiations have begun with CCD to see whether we can use their services for performing dumps. The main problem is the lack of high-speed connections to the IBM systems. At the last meeting, CCD were investigating mounting TCP/IP on the 3090. Various printers have been taken on trial to see which machines should be purchased in future. Work is still in progress.
ECSTASY is the Control Engineering environment being written at UMIST to interface various existing packages to make interworking easier. It is based on PASET tools, and Pro-MATLAB, and interfaces to the simulation package ACSL. The intention was to bring this into use during the year, both for academic and commercial engineers. RET acted as consultant to this effort (mainly helping them produce an understandable project description). This was sent round to a great many interested parties in the autumn of 1987, and the replies were sufficiently encouraging to allow work to continue. However, there have been many delays, and beta testing will now not take place till the autumn of 1988. The intention is to hand the system to RAL to organise support and maintenance (possibly with a Commercial organisation such as Cambridge Control).
In order to provide the necessary help, it was agreed to start training RAL staff. Initially, this involved HKFY and MLJ, but, having just got started, MLJ resigned. Given the shortage of staff in Systems, it was decided to move the project to a new home, and PJH and PAJ began to be involved. They were joined by KMT in February. Together, it proved possible to mount a version of the software at RAL (although there were frequent changes to the code at UMIST) and to interface a new simulation package, TSIM.
In addition, it was decided to test the portability of the software by trying to mount it on a VAX running VMS. The work is being done at Oxford, with the contract managed by RET. So far, progress has been slow but steady. The main problem has been simulating the Unix pipe mechanism.
During November, PJH became responsible for the Alvey Mail project. He carried out an analysis of the replies to the Alvey Mail questionnaire concerning the quality and usefulness of the service. Little seems to have been done with the information, however. It was decided (by the Alvey Directorate) to close the service at the end of September 1988, but then a late request was received to extend the service to the end of the year. Current thinking within DTI suggests that any future service will be based on Commercial systems, and that no further RAL involvement will be required.
PJH developed the Divisional Finance Project, aimed at automating the production of regular financial statements for the Department by the DAO. This had begun by studying the methods used by the previous DAO. A feasibility study was carried out to see what information could be extracted from FDS, the on-line system available on the central IBM. The study showed that while there was a lot of relevant information accessible, it was geared to casual use from terminals only. There were also some limitations on textual output. PJH had consultations with Central Computing Division which resulted in a variation of the FDS program suitable for this project, and wrote a set of programs to produce the tables required by IGLM finance papers. PJH produced the papers himself initially, and then organised a suitable interface so that the job could be performed by the DAO.
Some changes have been required since the original system was installed. One set was necessary to reflect changes that were made in FDS at the change of the financial year. Others were added to handle departmental data, including project definitions and allocations, pertaining to different financial years, and a method of including the DAO's descriptions of requisitions.
TIGER is a UIMS (User Interface Management System). A UIMS is used by application programmers to write their own customized interfaces. A visit to Team Engineering in Santa Cruz, California, was carried out in February. DGH helped Brian Colyer to use TIGER to re-program some of the user interface components of the apple finite element application. Although only a small subset of the apple code was actually interfaced, enough was done to convince Brian of the suitability of TIGER. Since then, DGH has been evaluating the TIGER package. Essentially, this has so far meant taking two Fortran-77 applications, removing the "ww specifics, and replacing them with TIGER function calls. The aim is to fully replicate the functionality of the original interactive graphics. In parallel with this task, UMIST are evaluating PASet, using the same two Fortran-77 programs. A comparison of the two UIMS will then be possible.
TIGER employs high-level graphics drawing routines and so certain functions that are performed with the ww routines can not be performed with TIGER. At the moment it looks as if PASet might be able to provide more functionality than TIGER as far as graphics primitives are concerned but TIGER appears easier to program.
Once again, ID played host to 7 trainee scientists, to give them some short exposure to Unix. CDR supervised them, and RET provided the customer feedback. The extra time, and the previous training in Unix and C, meant that the results were better than in previous years. However, it was the last time that UTS could be used, so new plans will be required when the next batch arrive.
PAJ automated the production of Departmental Business cards (originally designed by NBMC) so that the DAO can produce them on request with the minimum of input. Individual cards have been provided to group leaders and other selected people. General cards are available from the DAO for anyone.
PJH produced a revised set of Informatics handouts from material supplied by group leaders. These have to be generated in a variety of ways and PJH designed and implemented an interface for generating them automatically.
The section's interests lie in the general area of user interface design on advanced workstations, undertaken both by doing R & D work in the area and by supporting relevant research work elsewhere. The major funding areas are:
Staff currently in post are:
CMC and CAAG both got through the promotion system this year. Both Helen Jenkins (HVJ) and Jacqui Smith (now Perkins) have provided effective secretarial support during the year.
The Alvey MMI Directorate approved work to proceed at RAL on a Client Server Interface. The project's objectives are two-fold: (a) define a low-level software interface (known as the Client-Server Interface or CSI) to window management functions implementable on a range of workstations, and (b) implement the CSI on selected workstations as an existence proof. With vendor support for the CSI, application toolkits such as WW, which at present require extensive work to port to new environments would become much more widely available, easing greatly the work of applications programmers. Phase (a) was completed last year. The implementation on the PERQ (with limited functionality) is complete. Further work has been stopped due to (a) the disappearance of any significant UK vendor and (b) the wide availability and vendor support for X.
Some interesting prototyping work was done on a layout manager (user interface) based on the CSI (CAAG). Some experience has been gained with X version 11 including performance improvements which make the SUN version usable (CMC, CAAG) as well as the application of the MIT bug fixes. Some assessment of associated X software has also been made.
Robin Milner's group at Edinburgh have a reputation for good formal methods work, and have realised that a good user interface is necessary for effective take-up of their work by ordinary software engineers. Following some discussions the Alvey Software Engineering Directorate agreed to fund work at RAL to provide a user interface for the CCS concurrent programming workbench. MMM has done most of the work in providing a prototype user interface based on the object-oriented toolkit (see below), with help/comments/criticism from the rest of the section.
MMM has continued to work on this graphical toolkit, aimed at applications programmers needing tools to write highly interactive applications. This year's improvements include: extensions to the conv utility to enable pictures in troff or PostScript documents; pulldown menus; greying; bezier curves; flood fill; arbitrarily large cursors; more font metrics; and extensively reworked input. Some work was also done to isolate more effectively machine dependencies. WW is now available on SUN X11 and X11 release 2, and on the Whitechapel HiTech. The usual support activities have continued.
Experience with the Spy screen editor has led to MMM writing TEN, which gives more screen real-estate back to the user, as well as better functionality based on WW improvements such as filestore browsing, menus, cursors, small text areas and cut and paste with the window manager. A trial version of slow scrolling, and many changes to the user interface, were undertaken. The rewritten man page now has an annotated picture.
RALpage is the new name for the PostScript previewer. During November 1987 the software was posted to the Usenet news network and is in use at many hundreds of sites world-wide. The software has since been improved to read screen fonts in preference to the previous algorithmically generated ones. Some productising has been done to make it part of the ID Unix Service. (Mostly CAAG, with some ASW.)
There has been considerable commercial interest, and RAL Admin have (eventually) drawn up a contract, although this took some considerable time.
The Alvey Demo material has been tidied up and is now in one place (MMM).
This CSSC-funded project is investigating the ways in which extensibility can operate in a graphically-driven environment. It is now well into the first phase with various experimental tools built by CAAG - a graphical spreadsheet with a spy-like user interface, and a graphical unix terminal emulation which has programmable menus. Discussion papers have also been produced (CAAG, ASW).
A prototype basic level OO Toolkit has been developed, mainly by CMC, but with much advice (you can't be serious!) from ASW, CAAG, and MMM. This is written in C++. A set of primitive classes, useful for building more complex widgets, has been provided and will be distributed shortly. The GNU C++ compilation system has been obtained with a view to comparing it with the AT&T system.
CAAG continues to act as Mailshot Coordinator, with KR trying to sort out minor problems as necessary. CMC looks after the videotape library.
This project aims to provide access to information on MMI research interests and services in the UK. KGD took over the project from Arthur Seaton and has done most of the direct database and form design work using various versions of INGRES.
After many problems with the public domain INGRES, the SUN INGRES version was purchased. This took three months to arrive as Harwell Contracts were determined to sort out the whole software licence situation properly. After many calls of increasing intensity they relaxed their stance and shortly after Xmas it arrived (KGD, KR). The work of structuring the database and correcting the data was also undertaken. The front-end (user interface) for the database has been defined but so far has resulted in bug reports rather than software. A paper version has been produced for DTI.
TC looks after the Alvey Human Interface Club Steering Committee, under Vic Maller's chairmanship. As well as the usual secretarial duties, this has also required organising a HI Club meeting in November (with KGD), and some support work for the HI Special Interest Groups. The HI Professional Bodies Liaison Group is also supported.
PUMs are Programmable User Models - these are things that the interface designer might have to program in future to see how good or bad his interface design is. (Think about an architecture which has memory which forgets if you give it too much to remember.) Richard Young at the MRC Applied Psychology Unit at Cambridge is undertaking this work with Alvey HI funding. KR is the Monitoring Officer for the project.
COMETT is another Euro-initiative, this time for courseware for CBT. DAD, KR, ASW and TC have prepared an outline project for a course to be given on HCI in December at Darmstadt.
The Chemical Engineering Plant Design proposal finally did not get funded, with comments betraying the fact that it was not really understood. The Twinning proposal with the Technische Hochschule, Darmstadt was also not funded.
Now the good news. ASW, TC & KR spent many months preparing ESPRIT II proposals and it is now very likely that three of these will be funded EWS (a high power technical workstation), ELO (The ELusive Office - mobile teleworking), and MMI2 (multi-modal interface to knowledge based systems, mainly written by Mike Wilson). A number of IED proposals are also under preparation; and a number of discussions have been held with DEC - a large award under their European External Research Programme is likely.
The usual collection of things that do not fit neatly above are reported on here.
Exploitation of RAL Software. There has been much discussion and some actual progress with BTG on the exploitation of WW and associated software, with BTG finally relinquishing their rights. KR has probably suffered the most. ASW had managed to persuade ICL to allow spy to be distributed to commercial companies (for internal use only), and later in the year KR finally got ICL to let us sell up to 500 copies per year!
UIMS Survey. ASW cooperated with MJP on the gathering of information for the ECF Programme, generating assessment criteria and a detailed assessment of the products. TGF, a visitor from Australia's CSIRO, has been working on the assessment of Apollo's Open Dialogue.
ECSTASY. This is a project at UMIST on which ASW provides consultancy on user interface issues.
The following projects are envisaged for the coming year, given funding:
The staff involved are:
KMC has continued to organise the Alvey Vision Club, particularly its very successful three day conferences. She coped with approximately 180 delegates and all the organisation at Cambridge in September 1987. This year she is being assisted by local organisers in Manchester in arranging for 170 delegates to attend AVC88 at the end of August.
Plans are also well advanced for a 2-day workshop to be held at Cosener's House in November. The title - Exploitation of Vision - was so well chosen that it is reported that Timothy Walker, head of lED, wishes to attend.
KMC with the help of DCP has continued to provide systems programming and management for the VMS MicroVax II system which provides image capture, manipulation and display facilities. These are used by IPAL and 2D Object Identification projects. Progress on these two projects, which were within Infrastructure Group throughout the report period, is reported under Computational Modelling, in sections 2.5 and 2.6 respectively.
This system is also used by J Burren's Group in CCD.
KMC attended:
IPAL is a library of portable image processing algorithms being developed by an Alvey MMI grant. The method is to obtain contributions and provide a coordinated library, which will be taken over by NAG.
This is largely a matter of regular contacts with NAG. In addition to this RM has been convening regular meetings of all the members of the project. He acts as both secretary and chairman to these meetings. The usual attendees have been:
All these meetings are written up as PROFS documents with key-words IPA, technical, meeting.
As part of this function RM attends the main project management meetings at NAG in Oxford. These are held quarterly and are chaired by B Ford. The monitoring officer, D Rutovitz, attends and three people from outside are also invited.
The final part of this role is the yearly advisor's meeting. This is held at RAL and RM was heavily involved with the organisation of this years meeting on 23 June. He prepared a paper on the specification of the C Library and presented it to this meeting. The total attendance was thirty with five of these from RAL and three from NAG.
The emphasis in 1987 was on the creation and documentation of the core routines on the library. RM placed a lot of importance on getting agreement on common standards and conventions for these routines. The prevention of the erosion of the conventions has required frequent intervention on his part. The responsibility for final documentation lies with NAG and this has been a serious cause of delay in the project. Part of the purpose of getting together a set of documented routines was to act as a catalyst to the contributions on which this project is, in theory, based. So far the Fortran Library is based on core routines written by RAL members of the project and routines from Starlink modified by E Golton and M Carter. RM added a set of mathematical morphology routines for simple operations based on grey-levels. These and core routines which RM has written and documented are the main direct contribution to the Fortran Library.
As well as contributing routines for IPAL, KMC has maintained a public version of the entire source and documentation for the Fortran library. Two example programs have also been written. One is a simple fixed demonstration (DEMOIPAL). The other (IPALIMS) provides menu-based access to all the IP AL routines. However, its use as a test vehicle for establishing the interworking of routines is more important than its role as a demonstration. Both were shown to members of the Technical Advisory Meeting.
During the Summer a student (TCP) from Warwick University demonstrated the portability of the code by implementing it on two different transputer systems using the Parallel Fortran compiler. Both demonstration programs can now be run on T400 and T800 based systems. The latter, using a single transputer, performs as quickly as on the MicroVax II.
RM started work on the C Library from a position of total ignorance! This covered his knowledge of both C and UNIX so a first step was a visit to MRC at Edinburgh. The group visited is headed by the monitoring officer for the project and he has several years experience of producing software in C. One result of this visit was to obtain a tape of their image analysis software. After NAG colleagues had tried and failed to read this tape RM took over and UNIX support read it in. Since March RM has been experimenting with this software and has run it on the PYRAMID, SUNs and the image processing section MicroVax which runs VMS.
RM produced a paper on his experience running this software and, since the MRC software is likely to be the biggest single contribution to the C Library, he also produced a draft specification of the C Library based on this software.
RM has started work on implementing some basic routines in C as a first step towards standardising the behaviour of the MRC software. The error handling is one area which RM has identified as needing work in order to bring the software up to the standard required in a library.
KMC has successfully called C routines from a Fortran main program on the MicroVax. This enables users of the IP AL Fortran library to call the C library routines without having to re-code them in Fortran.
Since completing his High Energy Physics thesis at Durham, MC has been working on image processing activities initially as part of the 2D object identification, but more recently has been concerned with implementing algorithms for the IP AL Library. This has consisted of some chasing up of people in the Alvey vision community for Fortran code for the library; collaboration with Ted Golton on his filter and transform routines and writing Fortran code for various algorithms recommended for inclusion in the library by the project image processing algorithms survey. The latter has involved research using the original paper describing the algorithm; generating the code and checking that it conforms to the original algorithm; extensive testing of the code performance and its handling of error conditions and the production of documentation for the subroutine.
In November last year MC attended a one week course run by the Transputer Initiative on Occam and the transputer. He has written software for image display and manipulation on the transputer graphics system bought for the Image Processing Group and implemented various algorithms on the transputer based digital signal processing system bought for evaluation of its applicability to image processing problems. Some interest has been generated in using a transputer system in space applications and this has led to a small grant to assess the use of transputers in image compression of satellite imagery for the CDS experiment on the SOHO mission. This involves choice of the appropriate compression technique and design of a transputer based system for its implementation. This has also brought out the possibility of using transputers in the photon counting detector system for space/ground applications where it will allow the hardware for event recognition and location to be replaced by a more sophisticated and versatile software approach. This has involved collaborative meetings with RGO and UCL and research into centroiding algorithms and system design over the last month or so.
Papers by R Maybury:
The Alvey research project Identification of Object from 2D Images of Natural Scenes was started at RAL by Josef Kittler. With his move to Surrey, RAL has contributed to the project with effort at the rate of 2MY per year. Josef Kittler continues to manage the activities. The main people at RAL are:
Regular progress meetings with Surrey ensure that the co-operation is successful.
Research into the uses of contextual information in object labelling has continued by ERR. In collaboration with Josef Kittler, he has written a paper summarising work on evidence combining strategies for use in probabilistic relaxation. This paper was submitted to the International Journal of Pattern Recognition and Artificial Intelligence and was accepted for publication without alteration.
As an application of some of the theoretical ideas described in this paper ERR has developed a highly robust edge-labelling algorithm. This represents the first really successful application of relaxation labelling to edge-labelling. The success has results from the consistent formulation of the labelling problem adopted in the theoretical work. A paper on the edge-labelling strategy has been submitted to IEEE Transactions on Pattern Analysis and Machine Intelligence. The code for the algorithm has been contributed to IPAL.
At present ERR is concluding some work concerned with extending a technique known as discrete relaxation. The results look fairly promising and he Is preparing them for publication in the journal Pattern Recognition.
As far as future research is concerned, the plan is to extend the work on context to include studies of Markov fields and the modelling of label processes using Gibbs samplers. The work will probably be concluded with a comparative study of different contextual decision strategies.
Most of the year MP has spent working on optimum filters for edge detection.
In February and March, MP and ERR were involved in attempts to obtain ESPRIT funding. This activity involved a great deal of effort in preparing the consortium proposal document on Modular Imaging Understanding System and attendance at technical meetings with partners from France, Germany, Spain and Portugal. The proposal was in the event unsuccessful. However, the experience gained and the contacts made were both valuable. Hopefully, both will be useful in bidding against round two of ESPRIT-II.
During the year ERR continued to provide support for the Alvey Speech Club. The Club held workshops in Cambridge during September and in Oxford during March. In both cases ERR was responsible for local arrangements, finance and the preparation of a report describing the workshop proceedings. The Speech Club Committee held three meetings at which ERR acted as Secretary. Throughout the year he edited the speech component of the MMI mailshot.
The Club is at present making plans to extend its life beyond the end of Alvey Programme with a view to providing a similar role in fostering the needs of speech technologists in the IT92 and ESPRIT programmes.
Being an Atlas Research Fellow, Maria Petrou spent some time teaching Mathematics at St Hilda's College in Oxford to students of Engineering.
Papers
Internal Reports
Conferences
The main changes over the year have been losses.
To promote and support SERC's research programme in IKBS and systems architecture . This is, at present, principally funded through the Alvey programme and most of the section's work is concerned with technical support and management of the Alvey IKBS and Architecture programmes. There is also a small amount of work for the Engineering Board's Environment Committee and research and development work.
The year has been one of some uncertainty. With the Alvey programme reaching maturity, the main question has been what would follow it and what the Laboratory's role in such a programme would be. Many of the doubts remain and the section has been obliged to work within the general constraint that RAL's infrastructure support of the Alvey programme would be reduced by half in financial year 1988/89 and fall away completely in the following year. Thus the pattern has been one of continuing staff redeployment throughout the year. Nonetheless much useful work has been done in supporting the active IKBS and architecture communities built up during the Alvey programme.
A two day meeting of the KBS Club (Secretary JWTS) was organised in January at St Anne's College Oxford. The format consisted mainly of (excellent) tutorials. There was lively and appreciative participation by the Club members. A second, one day, meeting of the Club was held in June when the main items were presentations on completed projects and discussion of the future needs of IKBS research and the possible role of the Club in this.
The Architecture Club (Secretary MBD) has held three meetings during the period. The first (in October 1987) was a presentation by the ANSA project, which helped explain why this seemingly elusive project was of considerable importance to a wide range of projects. In November 1987 a meeting was held to discuss future funding options for the UK architecture community (mainly Esprit and the then unannounced UK programme). John Elmore from the EEC provided valuable comments and there was a lively discussion. Finally, in March, Alan Bagshaw's swansong as Alvey Architecture Director was a presentation on the new UK programme as he perceives it; the Club also received reports from the three SIGs it had established (Knowledge Manipulation Engines, Parallel Declarative Systems and Systems Architecture on Silicon).
The Logic Programming Club has held one major meeting over the period. JWTS is now the Secretary and is working closely with Steve Torrance, Coordinator for the Logic Programming Club. The meeting focussed mainly on projects in progress but, like all other Clubs, was understandably anxious to consider its own future in the light of funding opportunities After Alvey.
SGD has provided invaluable support in ensuring the smooth operation of all three Clubs.
For all three Clubs, the future is hazy. It is not yet clear whether they will still have a role in the new UK programme or, perhaps, more widely in Esprit. In the meantime they are serving a valuable function in keeping their Alvey communities together.
The workshops and SIGs supported by Alvey in IKBS and Architecture have been a highly successful part of the whole programme, bringing together academics and industrialists to discuss specific topics in some detail. Sadly the workshops budget has been halved in 1988/89. This has meant more work for MBD/SGD in ensuring cost-consciousness in the assessment and support of workshop proposals from the community. Despite the constraints, a reasonable programme has been maintained.
MBD has been responsible for appointing and overseeing the work of Monitoring Officers (and uncles) to the projects, with support from SGD. Apart from the purely administrative task of ensuring Monitoring Officers are in place and paid for their services, the main emphasis has been on ensuring that reports of value to the Directorate are obtained from the Monitoring Officers.
AJL has personally monitored projects in the Vision, Logic Programming and Infrastructure areas.
SGD has been obtaining and publishing in Alvey News, abstracts of work flowing from Architecture and IKBS projects . The response has been good and the publication has provided a valuable shop window for many of the projects.
The Special Interest Group in AI is a long-standing Committee which has continued to play a valuable role in advising SERC and the Alvey Directorate on the software needs of AI researchers. AJL has been its Secretary. Topics of particular concern to the Committee have been the report of the UK AI Toolkit Committee (see below) and the Information Engineering Directorate's strategy document for systems engineering (produced as part of the planning process for the new UK programme).
AJL and KMT have continued distributing new releases of the New Implementation of Edinburgh Prolog (NIP) to SERC and Alvey supported groups. The continuing development of NIP is supported by a contract administered by the section with advice from SIGAI.
AJL and KMT have also continued to oversee contracts to Sussex University for development of Poplog and to Systems Designers for maintenance of the system. Poplog is an increasingly popular environment for AI research, which includes Prolog, Common Lisp, POP 11, Standard ML, object oriented programming and other valuable features.
Following the interim recommendations of the UK Toolkit Committee, contracts were placed to incorporate part of KEATS (the Knowledge Engineers Assistant - the product of an Alvey project) into Poplog and to develop the STRATA toolkit (initially using Poplog Common Lisp). AJL has overseen these developments and RAL has been chosen as a beta test site for STRATA.
AJL and KMT have continued their work of benchmarking different implementations of Prolog and writing up the results.
SIGAI placed a contract to initiate a Knowledge Representation Systems Trials Laboratory at Edinburgh AI Applications Institute; a related (DTI) contract has established a Parallel Architectures Laboratory at Edinburgh. These, and other contracts to Edinburgh for Lisp and Prolog standards work have been monitored by KFH and AJL.
The IKBS Mailshot (run by JWTS and SGD) has continued to thrive over the year. A survey conducted during the year has shown that recipients value the service highly and, interestingly, distribute their own copies widely within their organisations. Sadly this is a service which will soon have to cease unless other arrangements are made to continue it.
The IKBS (electronic) bulletin board (JWTS) has also continued to be a valued service to the community. Occasionally technical problems have limited the extent of its distribution over one network but it is still being produced at the rate of approximately one per week. Again this is a service that will have to end soon unless other arrangements are made.
The UK AI Toolkit Committee (Secretary JWTS) finally produced a report during the year recommending a substantial investment in UK/European toolkits; the report has been considered by SIGAI and some of the results of its early work (STRATA and KEATS on Poplog) are now becoming available.
AJL and JWTS have continued to support the SERC Environment Committee's Artificial Intelligence Facility for Engineers. This consists of two IBM PCs and a SUN 3/160, with appropriate AI software, which are loaned to engineers in universities for short periods for evaluation.
KMT has written a database in Prolog to maintain records of university users of NIP and Poplog.
MBD has undertaken a fairly extensive evaluation of a number of expert systems shells available on PCs. These include Crystal (including version 3), Xi Plus, Savoir and KES. As part of the evaluation a number of expert systems have been developed, including one to play noughts and crosses! Full reports of the evaluations have been published.
MBD has been developing interests in weak machine learning techniques which might come close to imitating the behaviour of the human brain. These include some understanding of neural networks and developing learning techniques based on the classifier/genetic algorithm systems of Holland et al.
JWTS is Reviews Editor of the AISB Quarterly.
SE Section is much more intimately involved in the management of the Alvey SE Programme than IKBS and MMI support. Together with the specific topics discussed below the section is involved in financial planning of the programme and assisting in providing the IED/public interface.
The staff involved in the work are:
Involvement in the IED (nee Alvey) programme has continued throughout the year, by assisting in co-ordination of the existing Alvey Programme and planning for the new IT92 Programme. The success of the monitoring officer programme, developed by the Alvey SE Division and later extended to other Divisions, led to the adoption of a standardised monitoring programme for the whole IED Programme. The management and operation of the monitoring officer programmes has been a major part of the support function provided by RAL for the IED.
Awareness is an important fact of the IED Programme and is supported in part by the monitoring programme. This is achieved by the publication of Abstracts of project Deliverables. Analysis of the public response and interest in the Deliverables has shown that the Abstracts play an important part in the dissemination of technical information about the Alvey IT projects.
