10th June 2004 SWSL Telecon minutes

6/10/04 SWSL telecon

Attending: 
  David Martin
  Michael Kifer
  Sheila McIlraith
  Rick Hull
  Steve Battle

David: the June SWSL face-to-face will occur.  

Notes on SWSL-Profile meeting
-----------------------------
Four key topic areas
1) what expressiveness?  Pretty much determinied - LP with non-monotonicity,
and possibly some bells and whistles (e.g., prioritization of rules).
  
Shiela: any notion of subsumption?
MichaelK: some notion of frames to give taxonomic structure, but
  not subsumtion in DL or OWL sense
MichaelK: we decided on an intersection of OWL-S and LP paradigm,
  emphasizing the LP side.
  i.e., OWL capabilities that can be expressed within LP
Sheila: so the basic language will be undecidable
MichaelK: right
David: so taxonomic info is important for SWSL-Profile?
Sheila: yes
   - it is natural for people to use
   - we should make sure that we've got reasonable computational
         machinery to work with this
MichaelK: conceptually this is important, but it doesn't
  add expressive power unless we get into the non-monotonic reasoning
  about it
-------
2) Surface Language: this has bubbled to the top as a key issue
David: it would be a quick win if we could find a surface language
  that is appealing and easy to work with
Sheila: will the surface syntax be "general-purpose" or "services-targeted"
MichaelK: no specific keywords for services.  Some vocabulary will
  come from OWL-S.  "service" will probably be a class in this
  
Sheila: that the surface syntax is domain-independent,
    but we will be able to express a services-specific ontology,
    which would typically be used
Dave: analogous to PSL, which uses FOL as surface syntax
  and then a domain theory which provides a number of domain-specific 
  predicates, etc.
Dave: that there are 2 proposals for surface syntax
MichaelK: Benjamin's proposal is LP with priorities (based on his work)
          a second proposal is F-logic
MichaelK: that Michael and Benjamin will work to merge the 
  2 proposals, next week
  that this might all be embedded somehow into RuleML
    RuleML continues to grow, eg, to include F-logic 
David: is the Forum effort developing an XML-based syntax for serialization?
MichaelK: there are several F-logic-based languages
   What Deiter is using
   what some company is using
   what MichaelK is using
   what Stefan Decker is using
 we are trying to come up with a unified syntax,
 and also an embedding into RuleML as the XML-ified version
-----
3) Deciding on the domain theory, the core concepts, 
   in which you say things about services, time relationships,
   policies
   agreed to adopt a layering approach, analogous to PSL
   start with a core, and then various modules you can add on
David : one thought is to try to tie with the W3C Arch group ontology
MichaelK: what about tieing to the SWSA ontology
David: do they have a clean, concise listing at present of an ontology?
Rick: we should also coordinate this with the SWSL-Process
David: XACML was mentioned, Rosetta net, UBL (an update to ebXML)
---------
4) Show some value/apps to web services world at large
-------
Steve: that Katia was making a case for exploiting OWL-S profile
Dave: yes, to add rules onto profile, and we'll have a new service
  language which will let us express things in a more appealing fashion.

Notes on SWSL process modelling
-------------------------------

MK: In the process modelling subgroup we decided to postpone the issue of
the surface language to concentrate on core concepts from PSL. We require
something more like a constraint language than a pure process composition
language. The work for the next few weeks is to decide what we need from the
process model.

MK: Reified formulae can be used to represent preconditions and
postconditions to bridge between the profile and process models.
SM: One problem we had was the issue of the closed world assumption. Are
preconditions conjunctive, or do we also allow disjunctions?
MK: Do you need to make the closed-world assumption for match-making or for
planning?
SM: For planning. The compilation step from the process model to the profile
has the disadvantage that it results in complex formulae.
MK: LP can deal with certain kinds of disjunctions. In the body, yes, but
not in the head.

DM: The OWL-S profile advertises a single process - we want to advertise a
service, which may comprise many processes.
DM: Both the profile and process models express IOPEs. The difference
between them is in the level of description.
SB: I see the difference as static verus dynamic service descriptions.
MK: Is redundancy of IOPEs inevitable?
SM: A major complaint about OWL-S is this redundancy, the lack of a mapping
between IOPEs. Originally the plan was that the IOPEs in the profile would
be generated from those in the proicess model.
DM: Do we really need IOPEs at all in the profile, or are they described at
a different level of granularity? In OWL-S the profile is really only for
advertisement, it should address what we really mean by 'service'.
SM: Even if IOPEs are only used to select services there should be some
relationship between them.
MK: How does this relate to choreography?
SM: The preconditions and effects of a composition can be derived. For a
sequential composition A;B the preconditions are those of A and B (modulo
the effects of A). The effects are those of A (modulo the effects of B) and
B.
MK: And for more than simple sequences?
SM: Choices are reflected as conditionals.
SM: What is the profile being used for? Some complex contracting scenarios
seem to require a full process model.

SM: What is now the purpose of the Thursday telecon? What commitment is
required to join the subgroups?
DM: Anybody is welcome at any time - so long as they don't impede progress.
MK: External reviewers play a useful role as well.

Received on Monday, 21 June 2004 09:18:16 UTC