Meeting Notes for the WSML phone conference on 24 Nov 2004
Notes taken by Uwe
Participants: Michael Kifer, Christina, Uwe, Reto, Ruben, Holger, Livia, Titi, Ioan, Jacek.

General:
 - Add Jacek to the member list
 - Benjamin Grosof might join the phone conferences from time to time 
 
Discussion d5.1:

- Introduction part still too long and do not really make clear the point
- Distinction between Service and Web Service notions:
    * Still there has not been an overall discussion on the terminology
    * It's not completely clear to all people what Dieter understands by this notions.
    * Some opinions and thought of the attendees:
        - Use a different terminology, for instance Service Instance and Service
        - Relation to Chris Preist's conceptual architecture (CA):
            + Service in WSMO = Concrete Service in CA (contract has been established, 
              all details clear for executing/implementing the contract)
            + Web Service in WSMO = Abstract Service in CA
        - Basic distinction between both wrt. Discovery:
            + Web Service Discovery is about posing a query (disovery request) against 
              a set of services. It uses basically the semantic annotation and does
              not involve *interaction* with the single services (or their providers)
              in some way (e.g. a discovery interface or partial execution of the service)
            + Service Discovery is about checking whether a Web Service can give access
              to some specific concrete service (a service instance). This goes beyond
              looking at the mere semantic annotations of services and involves some
              interaction (of what form ever) with the service or service provider.
            + A similar distinction has been made in the paper on Rule-based Discovery by Michael K. et. al:
            Discovery & Contracting
            + One could have as well different layers of capability descriptions: One for discovery (the one we have
              right now) and one more detailed (based on predefined predicates by the service that realize the interaction
              on an abstract level)
            + Very similar mechanisms can be used for both:
                * WS-Discovery: Pose a query against a set of services
                * S-Discovery: Pose one or several queries against a specific service (using some interaction)
       - Initial experiments with set-based modelling in SWF based on a theorem prover show that it is possible
         to implement the framework using ATP systems, but depending on the concrete ATP system and their configuration
         for proof search the reponse times for a single matching check can shorter or longer.
         Using a non-trivial ontology and Vampire the response times have been 1-2 seconds. Vampire terminated for
         all examples (positve and negative examples for matches) in around a similar amount of time.
         Increasing the ontology certainly has a significant effect on the response time.
       - Some more experiments on scalability will be done by Holger and Ruben using  DL reasoners.
       - Michael estimates that the Rule-based approach could cope with several hundered or a couple of thousand
         web services in a KB.
       - Clearly, for scalability discovery has to use additional techniques (in all approaches) to prune the search space
         as much as possible. (Some sort of indexing of service)
         Ruben wants to use set-based modelling, pre-classification and DL-reasoners for this task.

- How to proceed with the deliverable:

        - Work on the Introduction 
        - Wait for the ESWC paper on discovery and rework the deliverable afterwards
        - Future work: Introduce a notion of contracting and investigate how to connect
          contracting to choreography
          
- New Version on Dec. 24th 2004