SWSA Telecon Minutes of Dec 16, 2003

Attending: Massimo Paolucci, Mike Huhns, Mark Burstien, Michal Zarembe, Tim Finin,
Chris Bussler, Carol Goble

Last week: discussed Fipa approach - effects: how you modify beliefs, but
no notion of committment. 

MP: Had discussion with Katia Sycara - diff between agents and services.
BDI gives structure of agent, FIPA ACL says how to effect state of agent -
assumes communication channel but doesnt say much about it. WSDL talks
about channel but not the agent.  Should we make assumption beyond the WSDL
interface? Are we doing this by the semantics?

Tim: Working on 2 use cases. One based on the Fujitsu demo given at the
last F2F, describing and composing services in wireless environment. The
idea is that a person wants to compose services in particular way, publish
result as service. A second use case will cover more automatic service
composition. New device offers some services - wants to do others in
conjunction with locally found services. They both have opportutites for
addressing security and privacy. - Dont want sensors announcing who (users)
is around locally - policy based. Will talk with Norm Sadeh about this.

MB to CB: Would be good to have the B2B use case address the issue of SWS
support for versioning for mediation of protocol content and object
definition changes.

CG: Currently requires agreement that everybody can upgrade. Will try to
add a use case.  E.g, for EDI, there are US and worldwide committees. A
formal process of request for change/extension. As they discuss the
changes, vendors do software updates and testing, raise issues. Implemented
standards are worked thru before release. Then release happens. EDI AS1,
AS2 - over internet. Testing occurs before standards release. If you buy an
AS2 implementation - pairwise testing has already been done, certificates
released.

CB: The issue is how to use SWS for changes between versions - maybe issue
deltas and conditions on mediation rather than just new versions.  Adding
attributes is straightforward, the real issue is what is the impact of the
change? Does it make anything inconsistent? Are there new rules required to
avoid inconsistency?

CG: Intuitively, the second one is where the SW could apply - reasoning to
notice or take account of inconsistencies. Need local model of how you are
using the standard as well - adding the attribute implies a completely
different set of parents.

Explicit meta data for use must happen. 

CG: paper on changes to Gene ontology will be forwarded to the list. 
Also grid resource usage use case.