
	*Logic Programming Paradigm Meeting* Jurgen Dix This is the short version of a report about the Logic Programming Pradigm, an international meeting in Shakertown near Lexington, Kentucky, USA from April 24 to April 27 (see \url:URL:http://www.cs.engr.uky.edu/~inna/lpp.html). The full report is available from the author and is planned to be published in the AI Communications. Introduction The aim of the whole meeeting, organized by Krzysztof Apt (president of the Association of Logic Programming), Wiktor Marek, Mirek Truszczynski and David Warren, was to get a clear picture of the current activities in Logic Programming and its role in the coming years. In particular * to present current research directions in Logic Programming worldwide, * to analyze future research directions, * to boost Logic Programming in North America, and * to discuss the relation between Logic Programming and related fields. There were 13 regular talks, each scheduled for 45 minutes and followed by a 15 minutes discussion chaired by an expert of the topic of the presentation. Sometimes very lively discussions, stimulated by the chairs who started the discussion with some comments, emerged. In addition to the talks, there were two panels, one on the past and one on the future of Logic Programming. One evening was devoted to the idea of founding a north american chapter of the Association of Logic Programming: an idea that found quite some support and will eventually lead to such a subchapter. Most of the influential people in the area attended the workshop: 23 from North America, 9 from Europe, 2 from Japan and 1 from Australia, constituing a total of 35 participants. Talks The first lecture was delivered by Bob Kowalski: From Logic Programming to Multi-agent systems. Bob started with a list of successes of LP (CLP, ILP, logic grammars, knowledge representation) and discussed the case against LP (no procedural syntax, no logic for change of state or for input-output, unclear about concurrency and about treatment of objects). He then related objects and properties of LP to those in multi-agent systems: logic programs single agent's beliefs, integrity constraints single agent's goals, no operating system observe-think-act no change of state beliefs, goals and environment. Gopal Gupta gave a talk on Horn Clause Denotations and Their Applications. Gopal claimed that the use of logic programming as the notation for specifying denotational definitions of programming langauages leads to many advantages: (1) both syntax and semantics are executable and can be specified in the same notation; (2) compiled code (sequential or parallel) for a program can be automatically derived from its denotation via partial evaluation; (3) program verification and model checking can also be done within the same framework. The last speaker in the morning was Luc de Raedt who talked on (A perspective from Inductive Logic Programming. He gave a nice overview on the rapidly evolving area of inductive logic programming and its roots in machine learning. The success story of this branch of LP may be due to the fact that there was always a nice balance between (1) theoretical work, (2)practical applications and (3) developed systems. He ended with a proposal of something like a relational data mining language. The afternoon started with Danny de Schreye's talk on Assessment of some issues in CL-theory and -program development. He argued how, in the area of development and implementation, basic research in Computational Logic has reached significant breakthroughs, but that it is still far from its objective of formal semantics supported semi-automated software development. Danny tried to show how integration on various levels can be required. He considered insufficient work on methodological and software engineering issues as a major weakness and pointed to some wishes for semantics-based programming tools. The last talk on the first day was presented by Manuel Hermenegildo: a tutorial on Programming with Global Analysis. He argued that global data-flow analysis of (constraint) logic programs is reaching a comparatively high level of maturity and asked whether it is now time for its routine incorporation in standard compilers. Such incorporation can make good sense only if: (i) the range of applications of global analysis is large enough to justify the additional complication in the compiler, and (ii) global analysis technology can deal with all the features of practical languages (e.g., the ISO-Prolog built-ins) and scales up for large programs. Several people noted in the discussions that although LP has had some impact on computer science it did not have as much as it should. LP languages are still perceived as non-serious languages and thus many programming people regard LP research as obsolete. It was argued that producing finished products would help much more than just writing papers. The second day began with Jacques Cohen's talk on Computational Molecular Biology: A Promising Application Using LP and CLP. Jacques first introduced us to the basics of molecular biology, explaining the overall process and the transformation of DNA into pre-m-RNA, m-RNA, aminoacids and finally into proteins. In his talk, he considered in detail the step from pre-m-RNA into m-RNA. He considers problems in molecular biology in this area to be of great interest to the LP community and tried to attract the attention of LP researchers to this exciting new area. Saumya Debray gave the second talk on Link-time Optimization of Multi-Language Programs. One of Saumya's main points was that not many real-world applications are likely to be written entirely in a single language (such as PROLOG). Thus for PROLOG to survive, the issues relating to the use of multiple languages in a single application need to be addressed. Among such issues (semantics, interface issues) he addressed in particular the performance issue and described the current direction of the work in his group: postprocessing of executable files to apply to them some optimizations not available at compile time. Philip Wadler, the only alien (being an advocate of functional programming) in this meeting, ended the morning with his lecture on How enterprises use functional languages and why they do'nt. He pointed to half a dozen serious uses of functional langauges in the real world (like HOL and Isabelle, systems for telecommunications etc.).In the second part of his talk, he analysed why functional languages are not so widely used. He claimed that a language should support interlanguage working, possess extensive libraries, be highly portable, have a stable and easy to install implementation, should come with debuggers and profilers and be accompanied by training courses. He also mentioned that performance is no more a serious reason nowadays. A lively discussion followed, where several mentioned that they would like to do applications, but that (1) it is difficult to get academic credit for such work, (2) they do not really know how to do it. Others strongly opposed that viewpoint by saying that trying to build applications in cooperation with industry is a waste of time and would take us away from the work we should really do. The afternoon started with Donald Loveland's talk on Concerning the future of Automated Deduction. Don's talk summarized the key points of a report on the status and future directions of the area of automated deduction (AD) (available from \url:http://www.cs.duke.edu/AutoDedFD). According to Don Loveland, the AD area is at the same time both stronger than ever and facing diminishing support. After discussing the reasons for that, he suggested to focus on near terms, to use commercial applications as a measure, to find applications in other domains and to identify grang challenge problems. He also mentioned the missing user-friendlyness of theorem proving systems as an important drawback. The last talk was given by Raghu Ramakrishnan: Deductive Databases: Achievements and Challenges. Raghu complained in his enthousiastic talk about the lack of applications of DDB but also noted that they influenced real DBMS (via optimization techniques and some language constructs). However, entering the commercial mainstream takes place via SQL (DB2, Oracle, SQL3) with almost no recursion. The current status of DDB is that there are many prototypes around but yet no finished commercial products. Raghu claimed that DDB capabilities taken alone will not sustain a DBMS product but that such capabilities can be added to existing relational DBMS. The discussion again turned around the importance of applications and that they should influence the theoretical system design. Others still opposed. VS introduced the importance of changes, updates, materialized views and uncertainty in DDB. The last day of the meeting, Monday, saw three lectures. Vladimir Lifschitz started with Representing causation by logic programs, Veronica Dahl talked about The Logic of Language and finally Kazunori Ueda lectured on The next 10 Years of concurrent Logic Programming. Vladimir Lifschitz considered in his talk two important knowledge representation tasks where LP-technology can and should be applied: reasoning about actions and causal dependencies (not involving actions). He pointed to important work done in the last 5 years and emphasized the recent approach to planning by Nebel/Kuhler/Dimopoulos using smodels as a stable models engine. Vladimir introduced his action language C and suggested to extend the work of Nebel et.al.to C. Veronica Dahl argued in her talk that logic is an inseparable part of language and can play an important role in natural language processing. She first examined families of formalisms from linguistics and computational linguistics, drawing their connections to logic. She also considered a linguist's wishlist in terms of programming language features that would make his/her task straightforward {scalability, debugging, avoid left recursion, easy ways to change parsing strategies}). Kazunori Ueda gave the very last talk of the meeting. He started with giving a review of the last 13 years of {concurrency} and its many achievements. He argued that the power of concurrent languages comes from {logical variables as communication channels}. Among the failures and problemes in this area he mentioned the absence of good textbooks and the difficult and tedious way to learn the foundations. Panels The two panels were scheduled for Sunday, Panel 1 before dinner, Panel 2 just after it. Panel 1, Logic Programming and Software Engineering: has anything gone wrong? was chaired by Krzysztof Apt and consisted of 10-15 minutes contributions of Howard Blair, Leora Morgenstern, Michael Gelfond and Alan Robinson. Howard recalled one of the remarks of Bowen, namely that Logic Programming has failed to carry through its logical promise based on the logical duality. He stressed that we should be aware of the fact that Logic Programming has been a result of some 50 years of research originating with Herbrand. He mentioned his current interest in using Logic Programming for a study of dynamical systems. Michael stressed that Logic Programming is a formalism not only for Software Engineering but also for Knowledge Engineering and pointed out that for the latter modularity is a non-trivial issue due to the presence of negation. Also, he pointed out the importance of declarativity by citing an ambiguity in the definition of the forall built-in. Leora argued the importance of non-toy examples and quoted her work on inheritance networks as an example of a theory that after 10 years turned out to be highly relevant for practice. Finally Alan reviewed his almost 40 years long involvement in programming and admitted of having used Lisp but not Prolog. He confessed that he does not believe in the dictat of typing that in his opinion limits the freedom of programming. In the discussion a lack of public domain standard Prolog was pointed out as a main stumbling block in using Prolog for commercial projects. Further, lack of work on modules and objects in the logic programming framework was mentioned (though reputed by some, for example by pointing out to work of Dale Miller). Others pointed out that speed considerations are nowadays hardly an issue, citing Java that is some 10 times slower than C++. After dinner, the second panel on the future of LP took place. The speakers were Jurgen Dix, Maarten van Emden, Ray Reiter and Koichi Furukawa. Jurgen was complaining about the lack of real applications and the lack of serious systems and implementations. He pointed to two particular promising applications namely in planning and diagnosis where a theorem prover for stable models ("smodels") resp.~a theorem prover for positive disjunctive programs ("Protein") compared favourably to special purpose engines. He also mentioned applications of LP for intelligent Web-querying (mediators as extended logic programs) and a recent application of the "smodels" prover in program verification, where "smodels" is used as a fixpoint engine for a model checker based on µ-calculus. Maarten recalled the expectation of logic programmers in the late seventies, namely that logic programming could unify computer science. He argued that the opposite seems to have happened with logic programming, which has branched out into too many application areas. He asked whether we still have a common approach to these diverse application areas? What sets LP-people apart is a preference for a more principled and fundamental approach. Koichi Furukawa considered Inductive Logic Programming as an important and promising research area for Logic Programming. From a computational point of view, the new school of ILP, as represented by Stephen Muggleton, developed a procedure for performing induction by deductive computation plus very much simplified inductive computation. Furukawa-san argued that future research will and should extend in broad fields of Logic Programming. For example ILP should include non-monotonic reasoning, constraint logic programming, and try to integrate abductive logic programming and the event calculus into ILP. Ray Reiter, finally made some heuristic observations about how nonmonotonic logics might contribute to improving the efficiency of commonsense reasoning. All nonmonotonic formalisms have very high complexity, so there seems to be something wrong with this argument. However, in most cases in the literature, whenever a nonmonotonic theory can be compiled into an equivalent, monotonic logic (e.g. by solving the circumscription axiom), the result is a definition. Normally, such definitions support efficient reasoning. The discussion after the panel fell short because both the panelists as well as the audience were quite tired. In addition David Pearce spent some bottles of wine ... Miscellenea The evening of the first day of the meeting was devoted to discussing the possibility of forming some kind of organization of North American logic programmers. There was feeling that some organization was important to try to build relationships, research and otherwise, among North American LP researchers. The goals included increased mutual knowledge of research and research directions of groups in North America to allow for joint funding efforts, student exchange, etc. There was also some feeling that yet another organization was not necessary and we would better spend our time reaching out to other communities. The conclusion was that David S.Warren would form and coordinate a steering committee that would try to put together a proposal for how to proceed. Also we would organize a day or two informal workshop, probably to be scheduled around ICLP'99, at which North American researchers would provide broad overviews of their research directions and goals. David Pearce, the network coordinator of Compulog Net (\url:URL:http://www.compulog.org/, mentioned that many theoreticians from Compulog Net's academic nodes, have little exposure to the vast number of powerful and significant practical contributions the technology currently makes in business, industry and science. He drew attention to some of the current success stories and olunteered to collect pointers to these systems. In particular, Compulog Net is organising a one day industral conference during JICSLP 98 in Manchester ((\url{URL:http://www.cs.man.ac.uk/~kung-kiu/jicslp98.html}) which will describe some of the best work being done in LP, CLP and ILP applications. Given the successes and the unique infrastructural role of Compulog Net, Gopal Gupta is working on and trying to put forward Compulog Americas, a Compulog Net Europe equivalent like structure. Interested people are referred to URL:http://www.cs.nmsu.edu/~complog. Conclusion Like four years ago, the workshop site was still very charming: a large outdoor museum embedded in the gentle hills of the countryside near the Kentucky and Dix (!!) rivers. The Shaker's erstwhile craftmanship is always nice to admire and there are not many possibilities to lodge in such a unique 19 century environment. Wiktor Marek, Mirek Truszczynski and in particular Deborah East, Matt Levy and Inna Pivkina, did a wonderful job in organizing the event -the only drop of bitterness being the cold and rainy wheather during the whole weekend. Jurgen Dix, University of Koblenz-Landau Computer Science Department Rheinau 1, 56075 Koblenz, Germany URL:http://www.uni-koblenz.de/~dix/

Logic Programming Paradigm Meeting

Jurgen Dix

This is the short version of a report about the Logic Programming Pradigm, an international meeting in Shakertown near Lexington, Kentucky, USA from April 24 to April 27 (see \url:URL:http://www.cs.engr.uky.edu/~inna/lpp.html). The full report is available from the author and is planned to be published in the AI Communications.

Introduction

The aim of the whole meeeting, organized by Krzysztof Apt (president of the Association of Logic Programming), Wiktor Marek, Mirek Truszczynski and David Warren, was to get a clear picture of the current activities in Logic Programming and its role in the coming years. In particular

* to present current research directions in Logic Programming worldwide,

* to analyze future research directions,

* to boost Logic Programming in North America, and

* to discuss the relation between Logic Programming and related fields.

There were 13 regular talks, each scheduled for 45 minutes and followed by a 15 minutes discussion chaired by an expert of the topic of the presentation. Sometimes very lively discussions, stimulated by the chairs who started the discussion with some comments, emerged.

In addition to the talks, there were two panels, one on the past and one on the future of Logic Programming. One evening was devoted to the idea of founding a north american chapter of the Association of Logic Programming: an idea that found quite some support and will eventually lead to such a subchapter.

Most of the influential people in the area attended the workshop: 23 from North America, 9 from Europe, 2 from Japan and 1 from Australia, constituing a total of 35 participants.

Talks

The first lecture was delivered by Bob Kowalski: From Logic Programming to Multi-agent systems. Bob started with a list of successes of LP (CLP, ILP, logic grammars, knowledge representation) and discussed the case against LP (no procedural syntax, no logic for change of state or for input-output, unclear about concurrency and about treatment of objects). He then related objects and properties of LP to those in multi-agent systems: logic programs single agent's beliefs, integrity constraints single agent's goals, no operating system observe-think-act no change of state beliefs, goals and environment.

Gopal Gupta gave a talk on Horn Clause Denotations and Their Applications. Gopal claimed that the use of logic programming as the notation for specifying denotational definitions of programming langauages leads to many advantages:

(1) both syntax and semantics are executable and can be specified in the same notation;

(2) compiled code (sequential or parallel) for a program can be automatically derived from its denotation via partial evaluation;

(3) program verification and model checking can also be done within the same framework.

The last speaker in the morning was Luc de Raedt who talked on (A perspective from Inductive Logic Programming. He gave a nice overview on the rapidly evolving area of inductive logic programming and its roots in machine learning. The success story of this branch of LP may be due to the fact that there was always a nice balance between

(1) theoretical work,

(2)practical applications and

(3) developed systems.

He ended with a proposal of something like a relational data mining language.

The afternoon started with Danny de Schreye's talk on Assessment of some issues in CL-theory and -program development. He argued how, in the area of development and implementation, basic research in Computational Logic has reached significant breakthroughs, but that it is still far from its objective of formal semantics supported semi-automated software development. Danny tried to show how integration on various levels can be required. He considered insufficient work on methodological and software engineering issues as a major weakness and pointed to some wishes for semantics-based programming tools.

The last talk on the first day was presented by Manuel Hermenegildo: a tutorial on Programming with Global Analysis. He argued that global data-flow analysis of (constraint) logic programs is reaching a comparatively high level of maturity and asked whether it is now time for its routine incorporation in standard compilers. Such incorporation can make good sense only if: (i) the range of applications of global analysis is large enough to justify the additional complication in the compiler, and (ii) global analysis technology can deal with all the features of practical languages (e.g., the ISO-Prolog built-ins) and scales up for large programs.

Several people noted in the discussions that although LP has had some impact on computer science it did not have as much as it should. LP languages are still perceived as non-serious languages and thus many programming people regard LP research as obsolete. It was argued that producing finished products would help much more than just writing papers.

The second day began with Jacques Cohen's talk on Computational Molecular Biology: A Promising Application Using LP and CLP. Jacques first introduced us to the basics of molecular biology, explaining the overall process and the transformation of DNA into pre-m-RNA, m-RNA, aminoacids and finally into proteins. In his talk, he considered in detail the step from pre-m-RNA into m-RNA. He considers problems in molecular biology in this area to be of great interest to the LP community and tried to attract the attention of LP researchers to this exciting new area.

Saumya Debray gave the second talk on Link-time Optimization of Multi-Language Programs. One of Saumya's main points was that not many real-world applications are likely to be written entirely in a single language (such as PROLOG). Thus for PROLOG to survive, the issues relating to the use of multiple languages in a single application need to be addressed. Among such issues (semantics, interface issues) he addressed in particular the performance issue and described the current direction of the work in his group: postprocessing of executable files to apply to them some optimizations not available at compile time.

Philip Wadler, the only alien (being an advocate of functional programming) in this meeting, ended the morning with his lecture on How enterprises use functional languages and why they do'nt. He pointed to half a dozen serious uses of functional langauges in the real world (like HOL and Isabelle, systems for telecommunications etc.).In the second part of his talk, he analysed why functional languages are not so widely used. He claimed that a language should support interlanguage working, possess extensive libraries, be highly portable, have a stable and easy to install implementation, should come with debuggers and profilers and be accompanied by training courses. He also mentioned that performance is no more a serious reason nowadays.

A lively discussion followed, where several mentioned that they would like to do applications, but that

(1) it is difficult to get academic credit for such work,

(2) they do not really know how to do it.

Others strongly opposed that viewpoint by saying that trying to build applications in cooperation with industry is a waste of time and would take us away from the work we should really do.

The afternoon started with Donald Loveland's talk on Concerning the future of Automated Deduction. Don's talk summarized the key points of a report on the status and future directions of the area of automated deduction (AD) (available from \url:http://www.cs.duke.edu/AutoDedFD). According to Don Loveland, the AD area is at the same time both stronger than ever and facing diminishing support. After discussing the reasons for that, he suggested to focus on near terms, to use commercial applications as a measure, to find applications in other domains and to identify grang challenge problems. He also mentioned the missing user-friendlyness of theorem proving systems as an important drawback.

The last talk was given by Raghu Ramakrishnan: Deductive Databases: Achievements and Challenges. Raghu complained in his enthousiastic talk about the lack of applications of DDB but also noted that they influenced real DBMS (via optimization techniques and some language constructs). However, entering the commercial mainstream takes place via SQL (DB2, Oracle, SQL3) with almost no recursion. The current status of DDB is that there are many prototypes around but yet no finished commercial products. Raghu claimed that DDB capabilities taken alone will not sustain a DBMS product but that such capabilities can be added to existing relational DBMS.

The discussion again turned around the importance of applications and that they should influence the theoretical system design. Others still opposed. VS introduced the importance of changes, updates, materialized views and uncertainty in DDB.

The last day of the meeting, Monday, saw three lectures. Vladimir Lifschitz started with Representing causation by logic programs, Veronica Dahl talked about The Logic of Language and finally Kazunori Ueda lectured on The next 10 Years of concurrent Logic Programming.

Vladimir Lifschitz considered in his talk two important knowledge representation tasks where LP-technology can and should be applied: reasoning about actions and causal dependencies (not involving actions). He pointed to important work done in the last 5 years and emphasized the recent approach to planning by Nebel/Kuhler/Dimopoulos using smodels as a stable models engine. Vladimir introduced his action language C and suggested to extend the work of Nebel et.al.to C.

Veronica Dahl argued in her talk that logic is an inseparable part of language and can play an important role in natural language processing. She first examined families of formalisms from linguistics and computational linguistics, drawing their connections to logic. She also considered a linguist's wishlist in terms of programming language features that would make his/her task straightforward {scalability, debugging, avoid left recursion, easy ways to change parsing strategies}).

Kazunori Ueda gave the very last talk of the meeting. He started with giving a review of the last 13 years of {concurrency} and its many achievements. He argued that the power of concurrent languages comes from {logical variables as communication channels}. Among the failures and problemes in this area he mentioned the absence of good textbooks and the difficult and tedious way to learn the foundations.

Panels

The two panels were scheduled for Sunday, Panel 1 before dinner, Panel 2 just after it. Panel 1, Logic Programming and Software Engineering: has anything gone wrong? was chaired by Krzysztof Apt and consisted of 10-15 minutes contributions of Howard Blair, Leora Morgenstern, Michael Gelfond and Alan Robinson.

Howard recalled one of the remarks of Bowen, namely that Logic Programming has failed to carry through its logical promise based on the logical duality. He stressed that we should be aware of the fact that Logic Programming has been a result of some 50 years of research originating with Herbrand. He mentioned his current interest in using Logic Programming for a study of dynamical systems.

Michael stressed that Logic Programming is a formalism not only for Software Engineering but also for Knowledge Engineering and pointed out that for the latter modularity is a non-trivial issue due to the presence of negation. Also, he pointed out the importance of declarativity by citing an ambiguity in the definition of the forall built-in.

Leora argued the importance of non-toy examples and quoted her work on inheritance networks as an example of a theory that after 10 years turned out to be highly relevant for practice.

Finally Alan reviewed his almost 40 years long involvement in programming and admitted of having used Lisp but not Prolog. He confessed that he does not believe in the dictat of typing that in his opinion limits the freedom of programming.

In the discussion a lack of public domain standard Prolog was pointed out as a main stumbling block in using Prolog for commercial projects. Further, lack of work on modules and objects in the logic programming framework was mentioned (though reputed by some, for example by pointing out to work of Dale Miller). Others pointed out that speed considerations are nowadays hardly an issue, citing Java that is some 10 times slower than C++.

After dinner, the second panel on the future of LP took place. The speakers were Jurgen Dix, Maarten van Emden, Ray Reiter and Koichi Furukawa.

Jurgen was complaining about the lack of real applications and the lack of serious systems and implementations. He pointed to two particular promising applications namely in planning and diagnosis where a theorem prover for stable models ("smodels") resp.~a theorem prover for positive disjunctive programs ("Protein") compared favourably to special purpose engines. He also mentioned applications of LP for intelligent Web-querying (mediators as extended logic programs) and a recent application of the "smodels" prover in program verification, where "smodels" is used as a fixpoint engine for a model checker based on µ-calculus.

Maarten recalled the expectation of logic programmers in the late seventies, namely that logic programming could unify computer science. He argued that the opposite seems to have happened with logic programming, which has branched out into too many application areas. He asked whether we still have a common approach to these diverse application areas? What sets LP-people apart is a preference for a more principled and fundamental approach.

Koichi Furukawa considered Inductive Logic Programming as an important and promising research area for Logic Programming. From a computational point of view, the new school of ILP, as represented by Stephen Muggleton, developed a procedure for performing induction by deductive computation plus very much simplified inductive computation. Furukawa-san argued that future research will and should extend in broad fields of Logic Programming. For example ILP should include non-monotonic reasoning, constraint logic programming, and try to integrate abductive logic programming and the event calculus into ILP.

Ray Reiter, finally made some heuristic observations about how nonmonotonic logics might contribute to improving the efficiency of commonsense reasoning. All nonmonotonic formalisms have very high complexity, so there seems to be something wrong with this argument. However, in most cases in the literature, whenever a nonmonotonic theory can be compiled into an equivalent, monotonic logic (e.g. by solving the circumscription axiom), the result is a definition. Normally, such definitions support efficient reasoning.

The discussion after the panel fell short because both the panelists as well as the audience were quite tired. In addition David Pearce spent some bottles of wine ...

Miscellenea

The evening of the first day of the meeting was devoted to discussing the possibility of forming some kind of organization of North American logic programmers. There was feeling that some organization was important to try to build relationships, research and otherwise, among North American LP researchers. The goals included increased mutual knowledge of research and research directions of groups in North America to allow for joint funding efforts, student exchange, etc. There was also some feeling that yet another organization was not necessary and we would better spend our time reaching out to other communities. The conclusion was that David S.Warren would form and coordinate a steering committee that would try to put together a proposal for how to proceed. Also we would organize a day or two informal workshop, probably to be scheduled around ICLP'99, at which North American researchers would provide broad overviews of their research directions and goals.

David Pearce, the network coordinator of Compulog Net (\url:URL:http://www.compulog.org/, mentioned that many theoreticians from Compulog Net's academic nodes, have little exposure to the vast number of powerful and significant practical contributions the technology currently makes in business, industry and science. He drew attention to some of the current success stories and olunteered to collect pointers to these systems. In particular, Compulog Net is organising a one day industral conference during JICSLP 98 in Manchester ((\url{URL:http://www.cs.man.ac.uk/~kung-kiu/jicslp98.html}) which will describe some of the best work being done in LP, CLP and ILP applications.

Given the successes and the unique infrastructural role of Compulog Net, Gopal Gupta is working on and trying to put forward Compulog Americas, a Compulog Net Europe equivalent like structure. Interested people are referred to URL:http://www.cs.nmsu.edu/~complog.

Conclusion

Like four years ago, the workshop site was still very charming: a large outdoor museum embedded in the gentle hills of the countryside near the Kentucky and Dix (!!) rivers. The Shaker's erstwhile craftmanship is always nice to admire and there are not many possibilities to lodge in such a unique 19 century environment.

Wiktor Marek, Mirek Truszczynski and in particular Deborah East, Matt Levy and Inna Pivkina, did a wonderful job in organizing the event -the only drop of bitterness being the cold and rainy wheather during the whole weekend.

Jurgen Dix, University of Koblenz-Landau
Computer Science Department
Rheinau 1, 56075 Koblenz, Germany
URL:http://www.uni-koblenz.de/~dix/