SWSA Minutes of Oct 28, 2003

Mihal, Massimo, MarkB, Chris, Stuart



Chris do we want to build on someone else's stack?

Do we want multiple layers of our own?

M: What did WSMF do?
C: Not a stack model, we used a conceptual model, 
so mediation means new language, how do these langs relate?
So we argued to have one language that included all. Add new constructs as req'd.

Chris will organize the distilling of reqs from WSMF.

Mark, Massimo, Tim can do reqs from the DAML-S Model. 

Then we can see how those things sit on the WSA Stack and others like it. 

Chris will revise use case. 

{We looked at WSA Architecture draft} No clear stack diagram. 

{We looked at Retsina MAS Infrastructure paper

http://www-2.cs.cmu.edu/~softagents/papers/AAMAS.pdf

Which has a stack at page 4. }

Massimo: We should look at this as a set of requirements, 
as the order of the stack has been argued to be faulty (e.g. position of security).

MB: looking at the distinction between agent-internal component requirements and 
infrastructure requirements is useful.

MP: we found that there is a reflection of infrastructure in internal agent
supports.  MB Should we have requirements for the representation of goals
within agents? This is necessary in order to support dynamic service
identification and invocation. For example, to have the goal to produce a
PO to some vendor before you know if they used SAP or PeopleSoft.

CB: Yes, you can justify on the basis of re-use. Reusable independent
representations of objectives is clearly important.

SW: ISWC talk by folks at Maryland (the SHOP2 Talk) that distinguish Task and
Method, and Enrico Motta's talk also. 

MP: I talked to Enrico - disctinction between the task description and
service description. He said that DAMLS is missing the task description. 

MB: The distinction is: what is the client trying to accomplish vs what
does the service provide?

CB: ex, goal to accomplish signing of document. then you bind
to potential implementations. 

SW: Can be related to WSMF materials and DAML-S. 

In the UMaryland paper, abstract process node is task description, the
decompostitions are methods.

MB: Let's consider having the telecons earlier starting next week. I will poll. Perhaps 10 or 10:30 EST.