This report is the outcome of a workshop on User Interface Management Systems held at Glasgow, from Saturday 19th to Monday 21st September 1987. The workshop was the first of a series instigated by the Computer Facilities Technical Advisory Group to assist it in formulating policy in a number of important areas. Representatives were invited from a wide range of disciplines and organisations, both academic and commercial, all with experience or requirements as users, designers or investigators of user interfaces to computer systems.
he report gives full details of the objectives and recommendations of the workshop and appendices list the organising committee, the programme of activities and the names of participants. The principal recommendations are as follows:
Tools and systems in EASE should not be required to operate with only an alphanumeric terminal for communication with the user.
The Workshop was set up at the instigation of the Computer Facilities Technical Advisory Panel. Membership of the Organising Committee is given in Appendix I. Administration and planning was handled by Geoff Lambert of RAL, with Alistair Kilgour assisting with local arrangements in Glasgow.
A letter from Malcolm Atkinson, Chairman of CFTAG, enjoined the organising committee to be guided by the following aims:
Appendix III gives a list of all those who actively contributed to the workshop (although not all attended for the full three days). We believe that the distribution did come close to fulfilling the first four aims above, although there were inevitably some gaps. (About 60% of those invited were able to attend).
The workshop took place at the Kelvin Conference Centre, West of Scotland Science Park, from lunch-time on Saturday 19th September till lunch-time on Monday 21st September. A detailed programme is given in Appendix II. From 16.15 onwards on Sunday afternoon parallel working groups were organised to discuss requirements, tools and research issues, and these groups reported back on Monday morning. This arrangement worked well, although several participants felt that more time for discussion, with correspondingly less on presentations, would have been helpful.
The working groups were asked to say
while bearing in mind the general constraints of around 4 man-years/year of support and £50 K/year of EMR funds available for the user interface area. Integration of the working group reports was necessary in order to arrive at the conclusions in this report, which CFTAG requested (in Malcolm Atkinson's original letter) should contain:
The following section of this report is intended to answer satisfactorily these requests.
CFTAG identified the topic of this workshop as "User Interface Management Systems", but it became clear during the proceedings that no generally agreed definition of a UIMS could be found. (It is perhaps reassuring to note this diversity of views is reflected in the wider community as well, as indicated, for example, in the proceedings of the ACM SIGGRAPH workshop on Tools for User Interface Management reported in Computer Graphics Vol 21, No.2, April 1987).
Participants interpreted the remit broadly to include any software components which could form part of a user interface to an application or set of applications running on a single-user scientific workstation with high resolution screen. Attention was particularly focused on window managers, toolkits of interaction techniques, and dialogue specification and control systems, and to a lesser extent conventional graphics packages. The need for modularity, and compatibility between the components, was borne in mind, particularly in view of the EASE policy of providing a kit of parts from which research groups can select those components of particular relevance.
A major unresolved issue emerging out of the debate on the nature of a UIMS was the possibility, and desirability, of separating the user interface part of an application from the computational components, allowing the user interface to be separately specified and controlled, perhaps by a different person from the application programmer. Several participants with experience of large engineering applications packages indicated how centralisation of the communications between the user and the application had become essential in order to keep the maintenance problem within bounds, apart from the parallel desire to allow migration to a workstation environment Nevertheless, it was recognised that the nature of the interface may place unanticipated demands on the functionality of the applications modules, especially as regards incremental updating and constraint satisfaction.
Although there was little direct experience or knowledge among the delegates of the properties of X-11 or its associated toolkits, there was unanimous agreement that X-11 and its toolkits should be components of EASE, and a firm commitment should be made to supplying and supporting these.
No direct cost was estimated for acquisition of the X-11 package, which is expected to be in the public domain.
Reasons for support of X-11 included the widespread industrial backing it enjoys both from manufacturers and software developers, making it almost inevitable that it will become a de facto standard. This view was reinforced by participants at the workshop from software houses and CAD software developers. However it may be noted that this universal support was expressed in spite of the absence of hard information about functionality or performance of the system.
An urgent need was identified for education of both software developers and users in the capabilities and properties of the X-11 system. It was unanimously recommended, therefore, that CFC should commission one or more workshops on X-11 inviting experts from the U.S. where appropriate. (For the X-11 system itself, Robert Sheifler of MIT and Jim Gettys of DEC were suggested. Tony Williams at RAL can obtain names of major contributors to the X toolkits).
An equivalent cost of 3 man months was estimated for this activity. It was noted that several books on X are in preparation, and due for publication in the next few months.
Although there was a minority view that a major effort should be made by SERC to establish a leading place in X toolkit development, a majority felt that developments in the US (particularly by Hewlett-Packard and DEC) should be monitored as they are released, and adopted or extended where necessary to adapt them to the U.K. research environment. An independent U.K. initiative without detailed information about what was being done elsewhere could lead to unnecessary duplication of effort. There was, however, some support for the view that SERC should immediately place EMR contracts with suitable groups for developments specifically using X-11, so as to speed up the process of gaining X-11 experience, and to underline the commitment to X-11 as the supported approach.
It is recommended therefore that an X-11 Toolkits Working Party be set up, based at RAL, but involving other groups with relevant expertise and enthusiasm, to develop detailed knowledge of X toolkits and evaluate them for a range of applications. This group should also arrange to disseminate its knowledge periodically to the wider community by means of reports and seminars. Gaps or shortcomings identified in the received toolkits could be rectified by development work at RAL. The extent to which this will be needed is difficult to quantify at preset. Suggestions for tools which might be locally developed included a font manager, a general-purpose schematic drawing tool, and a report generator program.
There is a clear commitment to provide continued support to groups who have built applications using the RAL-developed WW portable window support library. Therefore it is recommended that a version of WW running on top of X-11 should be developed and supported.
The estimated effort for this is 3 man months.
This system has been adopted by the ECSTASY control engineering software project, and assessment of its performance and limitations, in the context of a large integrated applications suite, will emerge as that project progresses. Initial reactions from Prof. Neil Munro at the workshop were generally favourable.
Developments are in progress at PA to integrate their product with the GRAPE system, the rights to which they have purchased from the now-defunct Cambridge Graphics. It is recommended that a licence for the integrated PASet product be obtained by RAL, and that benchmark tests be carried out on one or more small applications systems. Several candidate applications within RAL were suggested.
This system, developed by Kasik and others at Boeing over several years, is now commercially available from Team Engineering Inc., 735-A Chestnut St., Santa Cruz, CA 95060. It provides less built-in functionality than PASet, but supports a wider range of dialogue specification techniques.
It is recommended that a licence for TIGER be purchased, and that it be evaluated in the context of an existing application, and a new one. Ideally it was felt that it would be valuable to test TIGER with ECSTASY or a project of similar scale, but because of resource constraints it was accepted that the applications used for TIGER evaluation", were likely to be small-scale. Nevertheless, some insight into the likely effect of scaling up may be gained.
The estimated resources for purchase and evaluation of both PASet and TIGER are 1 man year plus £25K capital expenditure.
Martin Prime of RAL presented a valuable review paper at the workshop. (Copies of his report, entitled "User Interface Management Systems - a Current Product Review", are available from him on request). However, his analysis of the products was based solely on descriptive material provided by the vendors, and not on hands-on experience. It was to go some way towards remedying this situation that the purchases of PASet and TIGER referred to above were recommended.
It is recommended that the review of commercially available products by RAL staff should continue, with revised reports being issued from time to time. Where possible opinion from users of the products should be sought, and review material from other sources should be collected, ego from SERC-funded investigators who have recently visited sites outside UK. The Alvey HCI centres in the UK (at QMC, Manchester and Heriot-Watt/Strathclyde) may also have valuable experience to provide.
A product of particular interest is the Apollo Dialogue System. Currently this runs only on Apollo workstations, but a portable version implemented in C++ and utilising X-windows is due to be released soon. This may provide a useful alternative to PASet and TIGER.
The desirability was discussed of providing either a demonstration centre, where interested parties could come in order to evaluate a UIMS product with their application, or a "lending library" of systems with an "itinerant licence" which could be assigned to the borrower for the period of the loan. On balance the latter idea received greater favour, and it was recommended that enquiries be made as to whether suppliers would provide a suitable form of licence.
Another alternative, suggested by Tony Williams and already adopted by several suppliers, is to obtain cut-down versions of the target system, which are often provided free of charge with permission, and indeed encouragement, to copy and distribute without restriction. These demonstration versions have some vital component missing, or have a built in limitation on the size of application they can handle, which prevents them from being commercially exploited, but still allows users to get the flavour of the full product. It is recommended that SERC should encourage all suppliers to follow the example of the few and provide evaluation versions of their systems free of charge.
It was recommended that this system should be made available as part of EASE for use by those carrying out research into interactive styles and techniques, but that researchers should not be encouraged to use as the basis for building interfaces to large applications.
It was noted that a version of this built on X-11 was being produced. This, together with Sun emulation of the PC, could make a large body of low-cost software available for use on Sun's. It is recommended that this situation be monitored.
Commercial software developers at the workshop who had used Windows were of the opinion that it was not yet a mature product (compared, for example, with the Macintosh system).
This was felt to be the appropriate level at which graphics support should be provided in conjunction with X, although concern was expressed about the restrictiveness of the input model embedded in CGI.
GKS is already supported within EASE, and the recommendation was that it should continue to be made available to those who required it. Similarly, as GKS-3D and PHIGS move towards final crystallisation as international standards, versions should be acquired by SERC and added to the EASE catalogue. Developments in this area were felt to be of less urgency than those concerning UIMS's and window managers.
A major recommendation was that no graphics package should assume it has complete control over the whole screen, and that it should be possible for several different graphics packages to co-exist in the same environment.
The object oriented programming approach could have a major impact on future developments in user interface design and construction. Many features taken for granted in to-day's workstation and microcomputer interfaces were pioneered in the Smalltalk-80 system. This is now available for the Sun range in a very efficient implementation by ParcPlace, and provides a powerful and versatile prototyping environment. No recommendation was made as regards provision of Smalltalk as part of EASE, however.
The alternative evolutionary approach to supporting object oriented programming is via extended languages such as Objective C and C++. Several initiatives are currently under way to provide user interface toolkits in C++, and the outcome and impact of these needs to be monitored.
A third development in this area referred to at the workshop was the release by Apple of the MacApp environment (The book "Object-Oriented Programming for the Macintosh" by Kurt Schmucker contains a detailed description and discussion of MacApp). Although not as yet relevant to Sun users, the release of MacApp could give considerable added impetus to the use of the object-oriented approach in the area of user interface design.
This issue has already been referred to in the introduction, and is likely to be the focus for research for some time to come, particularly in respect of the linkage between interface and application (the client-server interface), the differing requirements placed on the application by different interface styles, and the need for a combination of lexical, syntactic and semantic feedback within the same interactive technique.
It is possible that alternative approaches mars be helpful in thinking about the structure of user interfaces. One such alternative is the idea of a harness or framework (referred to as an "intermediate semantic representation" by Steve Draper, who described the idea at the workshop) which matches at an abstract level the structure of a wide range of applications. Details of a specific application may then be plugged in to this harness to provide an "instant interface" to the application. There was general agreement that research on this and other alternative approaches should be supported.
It was also recommended that experience by existing users of UIMS tools should be monitored, in particular current Alvey projects.
There was general agreement that no single formalism (BNF, Augmented Transition Network, push-down automaton, CSP, Petri nets, etc) was adequate for all applications. Further research was needed both into improved formalisms for specifying systems, particularly those supporting interleaved dialogues and full concurrency, and into systems which provide a range of different formalisms on top of the same underlying dialogue control system.
Object-oriented programming, as epitomised in the Smalltalk environment, is only one of several current approaches to fast prototyping. Others include use of functional programming, logic programming and persistent programming systems. The development and application of these approaches should be monitored, in order to provide access to expertise and guidance on the strengths and weaknesses of each approach.
There was recognition that many of the exciting developments and products in the user interface area were appearing in the personal computer domain, in particular for the Macintosh (eg. MacApp, Guide, Hypercard, etc). Although porting of software for the IBM PC and its compatibles to the Sun environment was now feasible, no way of transferring Macintosh applications to the Sun is yet available.
Because of the wealth of low cost software, much of it is of high quality, appearing on the small machine market, it was generally felt desirable that the possibility of incorporating products originally designed for this domain into EASE should be continuously monitored.
During the workshop experience of porting two large application suites (for control engineering and architectural energy analysis) from a batch to a workstation environment was described. It was recognised that a major component of the work of these and similar projects was in centralising and configuring the user interface, although there had sometimes been a tendency to believe that this process was straightforward and undemanding.
The need to support researchers engaged in productising existing applications software should be recognised, so that full advantage can be gained from previous research initiatives.
There was general agreement, with a few dissenters, that it was no longer necessary to constrain supported software products to be capable of being driven from an alphanumeric VDU terminal. Future systems may assume high resolution bit-mapped graphics as a minimum.
Among the issues discussed which it was felt would be likely to the subject of continuing research in the medium to long term were the following:
No suggestions for topics for future workshops were noted, but there were a number of comments as regards organisation and format:
Thanks are due to all who assisted in the establishment and organisation of the workshop, and in particular to Geoff Lambert who drove the whole enterprise forward, and whose guidance was invaluable on both the administrative and technical fronts.
This report is an attempt to condense a largy body of opinion and comment. The assistance of the chairmen of the working groups in ably summarising the work of their groups is gratefully acknowledged. Remaining inaccuracies or omissions are entirely the responsibility of the author.