- From: Gerhard Wickler <Gerhard.Wickler@informatik.uni-stuttgart.de>
- Date: Fri, 14 May 2004 17:29:52 +0200
- To: Gregory Huczynski <greg@dcs.gla.ac.uk>, public-sws-ig@w3.org
Gregory,
Having worked on the problem of capability description and discovery for
a while now, I believe I understand your problem. As you pointed out in
your example, a simple IO description is not sufficient to help Alice
find the right service. The two potential solutions you suggest involve
describing the service in terms of preconditions/effects or typing the
service.
The way I see it, these are not mutually exclusive but should go hand in
hand. Suppose you have a service ontology which defines a "Printing"
service type. I believe that the best way to define this action type in
an ontology would be in terms of its preconditions and effects. Whether
you do this by inheriting from a more general type of action and modify
it, or by explicitly representing them here is not important, as long as
they are represented in the ontology. Now any service that advertises a
"Printing" capability indirectly also states the effects it can achieve.
Thus, Alice should be able to send her request to the matchmaker either
as a task to be performed ("I need this document to be printed") or as a
goal to be achieved ("I need a hardcopy of this document"). The
matchmaker should be able to find the same services that could do the
job by using the definition in the ontology.
You are right, of course, that this approach does indeed rely on human
standardisation, i.e. service ontology development.
Gerhard
Received on Friday, 14 May 2004 11:33:54 UTC