service / agent terminology

I'm somewhat concerned about the direction of today's discussion of terminology.

I agree with David B that there's a lot of baggage around "service" as a provider.  However, that baggage is not imposed by abstract WSDL... rather it accrues by the common usage of request-reply as the prevalent operation signature.  I'd like to lose the baggage.

WSDL got many things "right", in my view, and the inclusion of "out" and "out-in" operation signatures is one of them.  This lets me define an entity that has a clear boundary of identifiable, typed communication endpoints... I'd like to call that entity a service, even if its only interaction with the outside world is periodic publishing of time and temp (on an appropriate transport).

Providing a service need not imply reactive communication... services can be proactive as well.  This means that "provider" need not imply "responder".

HTTP tunnel-vision has made a similar discussion on the Description group somewhat long and animated, apparently.  I think we have a great opportunity to define an architecture that extends beyond existing Web transport protocols to communation protocols in general.  (Will I now be skewered by REST proponents?)  I hope our charter does not preclude that.

Scott
-- 
Scott Vorthmann                    mailto:scottv@tibco.com
Senior Architect                     mailto:scottv1@imcingular.com
                                              office: 919 969 6513
TIBCO Extensibility                  mobile: 919 593 2349
TIBCO Software, Inc.               http://www.tibco.com

Received on Thursday, 20 February 2003 17:23:07 UTC