Rome, June 20-21, 1997

David Pearce

Participants

Luigina Aiello, Krzysztof Apt, Philippe Codognet, Tony Cohn,

Danny De Schreye, Alan Frisch, Manuel Hermenegildo, Tony Kakas,

Bob Kowalski, Nada Lavrac, David Pearce, Luis Moniz Pereira, Norman Sadeh, Domenico Sacca, Joerg Siekmann, Thomas Sjoland, Clive Spenser

Programme

Friday, June 20

14.30-16.15 Session 1: Analysis and diagnosis:

The failures and successes of logic programming (Chair: L. Aiello)

What can computational logic learn from the failures of logic programming? (Bob Kowalski)

What the logic programming community should have been aware of in the past (Krzysztof Apt)

Discussion

16.30-17.30 Session 2: Self-assessment (Chair: N.Sadeh))

Models of R&TD assessment (David Pearce)

Thoughts on supporting a market analysis of computationa logi c (Jorg Siekmann)

Discussion

Saturday, June 21

9.30-11.15 Session 3: The future of computational logic (Chair: D.Pearce)

Central themes for the future of computational logic (Tony Kakas)

Performance engineering in computational logic (Danny De Schreye)

Grand challenges for "networking architectures for massively

interconnected environments" (Manuel Hermenegildo)

Developments and applications at SICS (Thomas Sjoland)

Can LP exploit the successes of Java? Brief remarks (Clive Spenser)

Discussion

11.30-13.30 Session 4: Technological Roadmap (Chair: R. Kowalski)

Roadmaps and Industrial assessment techniques (Jorg Siekmann)

Discussion

Background, Aims and Methods

In 1996 an EU-commisioned study evaluating the concept of Networks of Excellence reported positively on NoEs, concluding that they provide sound and cost-effective frameworks for coordinating R&D and related activities between key industrial and academic players in a technological domain. As a result of the study, however, NoEs are now being required to deliver technological roadmaps for their discipline, which include the strategic long-term goals of the network and criteria and methods of self-assessment. New networks now have to provide such roadmaps as a precondition for their official launch. Older NoEs have to deliver roadmaps during the course of their current contract. The aim of this workshop was to lay the foundations for designing a technological roadmap for computational logic, which would identify the long-term goals of Compulog Net and select suitable criteria and methods of self-evaluation. Participants included members of the Executive Council of Compulog Net and ALP President, Krzyszttof Apt.

In preparing the workshop, the assumption was made that at least three key themes would have to be addressed in preparation for designing a roadmap:

(i) an analysis of the past, in particular with a view to learning from past mistakes;

(ii) the topic of technological assessment, in particular, how to characterise the technology, assess its degree of progressiveness, and evaluate its future market potential;

(iii) an analysis of technological trends and grand challenges for the future.

Lastly, some time was also to be devoted to the concept of technological roadmap and ways in which this noton might be understood in the present context. One session of the workshop was accordingly devoted to each of these three topics; the final one leading into a discussion of the computational logic roadmap.

This report summarises the main points on which substantial agreement and cohesion was achieved.

Learning from the past

Session 1, devoted to analysing the past, featured presentations by Bob Kowalski and Krzysztof Apt. In addition, several speakers in Session 3 also considered past achievements of the field.There was a consensus that logic programming (LP) has enjoyed considerable success, but now needs to learn from its mistakes. Speakers agreed that computational logic (CL) is more than just Prolog and logic programming, but thought that LP-based technologies should continue to play a strong role in the future development of CL. The main plea was that, rather than claim (as in the past) to be developing a uniform language for all of computing, one should try to indentify application areas and programming styles and environments where CL may be competitive and to focus R&D on these. Particularly important are areas with low or relatively low entry barriers.

It was also urged that today even basic, or what is now called "long-term", research needs to be applications-driven, otherwise there will be few prospects for funding in the future. There was general agreement that an important task for the present is toanalyse and promote successful applications of the technology; in particular, to determine what are the appropriate criteria by which one may judge a given application to be successful and to promote the technology by publicising its best and most striking success stories.

Standards and assessment

This needs naturally into the general subject of standards and evaluation, treated in Session 2, and by several speakers and discussants in Session 3, including Danny de Schreye and Manuel Hermenegildo. It was agreed that standards are important in programming and in software engineering. These may include eg. portability, maintainability, reliability and verifiability. Likewise, the existence of good environments, modules, objects, types, etc, as well as good interfaces to other languages, the internet and so forth are important. Good implementations are also needed, eg with small executable size, scripts, concurrency, specialisation. However, it was argued that logic programming has often been slow or deficient in some of these areas. Moreover it was observed that there is a complete lack of studies in logic programming styles.

Besides the general recommendation to improve software standards where needed, several specific actions were suggested by different speakers. These included:

- Develop appropriate indicators and measures (eg software metrics) for quality management.

- Characterise and evaluate (constraint) logic programming as a technological research programme, in particular evaluate its (past and present) progressiveness.

- Analyse the present market and future market trends. What is the nature of the market? What is it worth? What percentage of it do you want to capture?

Any of these issues might be tackled by commissioning studies on the topic in question (though it was agreed that a professional market analysis might be prohibitively expensive).

Technological roadmaps

Session 4 was devoted to the topic of technological roadmaps in general and that of computational logic in particular. Once the general concept of roadmap had been covered, the discussion returned to the earlier issue of areas of opportunity for computational logic, ie suitable growth areas with high potential and relatively low entry barriers. It was decided to use one such area as a focus for building an initial roadmap.

Participants singled-out for further study not an application area or vertical market in the usual sense, but rather a particular environment: internet/intranet. The main assumptions here were that CLP provides a good paradigm for internet and WWW programming, that this environment is not yet technologically saturated, and that there are high potential rewards. The question how LP might exploit the market successes of Java remains to be tackled.

Given the suggested focus on internet/intranet programming, the next issues that arise concern (1) the nature of the enabling technologies, (2) the specific fields of application. With respect to (1) it was agreed that all the main subareas of computational logic, as currently highlighted within the network's Research Area structure, could provide key technological inputs. One could therefore consider the contributions of Logic and Databases, Logic and Learning, Constraint-based approaches, Logic and Natural Language Processing, Parallelism, Verification Methods, and so forth. For each of these one would lay down the main technical challenges and milestones, viewed as enabling sub-technologies for the internet/intranet environment. With respect to (2), it was argued that all the industrial and commercial areas where CLP is currently successfully applied in a conventional programming setting could be considered good potential areas for internet/intranet programming. They include manufacturing, logistics, transportation, finance and banking, medicine, telecommunications, design and fashion, and so on.

Summary and actions

Participants agreed that the network should focus in the short to medium term on the following four aims:

- Try to give increase exposure to successful (C)LP industrial applications - through the website, books, articles, events, contests, etc.

- Look closer at the feasibility of commissioning evaluation and technology assessment studies which might include also market analyses

- Start to fill-in different paths on the (Internet/intranet-oriented) technological roadmap.

- Review the above and continue this exercise at a second strategic planning workshop to take place in the coming year.