Minutes of SWSA Telecon of 10 June 03

Participating: Mark Burstein, Mike Dean, Stuart Williams, Enrico
  Motta, Massimo Paolucci, Tim Tim
Regrets: Chris Bussler, Carole Goble
Scribe: Tim Finin

The meeting was begun around 12:10 UTC-4

Mark noted that Chris Bussler is changing jobs, leaving Oracle and
joining Deiter Fensel's DERI project in Galway Ireland.  This means,
of course, that our European co-chair is now working in Europe.

Mark suggests that today's goal should be to come up with a proposed
set of tasks and milestones for the group.  Developing a timeline for our
use cases is one set of possible milestones.

Enrico asks what the relationship will be between the functionality
offered by our architecture and the ontology assumed by the SWSL
subgroup.  each might presuppose the other.  A discussion ensued.
Mark pointed out that the SWSL committee thinks we are behind and
should be guided by what they have done.  Tim suggested that we will
need to bounce back and forth.

Mark suggests we look at the semantic and lifecycle functions
identified in recent meetings.  Here was the list from the minutes of 6/3/03
(email sent 2:15 on that date by Mark)
-----------------------------------
The list was organized it into two categories, those that are related to
agent/service lifecycle and those that are related to semantic
discovery/composition/interoperability.

semantic functions
----------
service invokation (request/query/negotiation/agreement)
message mediation (message translation & elaboration)
process mediation
candidate service identification (matchmaking)
candidate service selection (from candidates)
reputation service
process composition
task management / task status
ontology management (for communities sharing ontologies)
community management (partnering services)

lifecycle functions
----------
registration & nameservices
service (operational) status
authentication
instantiation (e.g. service factories)
service migration (between hosts)

---------------------------------------

Tim suggests two approaches: (1) start with a blank slate and develop
a set of boxes and messages/APIs that result in a SWS framework, then
try to map parts onto regular web services. (2) start with some of the
current web services components and look for ways to augment them.
Tim prefers the first approach.

Mark talks about a planning example. user wants to plan a trip with
airline, car, hotel.  may require ontology translation.

 - client starts with a complex goal.
 - client sends same to matchmaker.
 - does the matchmaker have it's own local DB of services, or does it
   use a separate service registry?
 - how does this work with ad hoc p2p networks

Mark describe a possible reputation service similar to ones used on
comparative shopping web sites as a potential function of SWS network.

Tim: How about authentication?  A matchmaker might reveal certain
services onto to some clients.  Similarly, only some clients will be
authorized to use certain services.  Thus, some form of authentication
like service may be an important part of an architecture.

Mark: A matchmaker might engage in a clarification or information seeking
dialog in order to make its recommendation. It might also ask the services
for quality of match to requests (moving down slope toward distributed
matchmaking).

Massimo discusses other ways to carve up the problem -- matchmaker,
broker, facilitator, ...  we could look at how to distribute the
fundamental capabilities, resulting in different architectures.

Mark: certain architectures are appropriate only if you distribute
them in a certain way.

Mark: we should be careful not to just assume an agent architecture
and not think about other, more web like, architectures.  On the web,
all one needs to do to be discovered is to publish a web page that
someone points to.

Mark suggests the following homework for next week.  Each committee member
should pick an example of a scenario such as booking a trip or buying a
book.  Using this scenario, each of us will try to identify the atomic SWS
architectural elements (semantics-related functions) involved in completing
the processes involved.  We should remember to include elements like
dealing with a mismatch of Ontologies.  We should all send our results to
the group via email no later than Monday. Discussion via email encouraged,
to be followed by telecon next Tuesday.


The meeting was adjourned around 13:10 UTC-4