- From: Sanjiva Weerawarana <sanjiva@watson.ibm.com>
- Date: Fri, 1 Nov 2002 05:46:22 +0600
- To: "WS-Desc WG \(Public\)" <www-ws-desc@w3.org>
Don Mullen [mailto:donmullen@tibco.com] writes: > > It strikes me that we are discussing two use-cases: > > 1) Event mechanism for point to point (strong use case in BPEL). > output and output-input not necessarily required. Better > event mechanism for WSDL would be. > > 2) Pub/Sub where publisher might not know/care about specific > endpoints (netnews example, other messaging systems). > 'Subscriber's might dynamically tap into published information > without the publisher's knowledge, so from the server's > [publisher's] point of view, there *is* no 'subscription'. > output and output-input required (with enhancements/clarifications). BTW, I don't grok what the last sentence in (1) is trying to do: is it leading on to (2) or trying to state an assertion about (1) being better?? I susepect its the former, but please clarify. Its a good idea to agree on the scope first! IMO we need to solve problem (1). WSDL is a language for describing a service so that a user of the service has all info necessary to use the service. Even if (2) is the mechanism by which events are delivered from the source to the sync, the client (the sync) *does* need a mechanism to indicate to something that it wants to subscribe to the event. That subscription request may not find its way all the way back to the event source (as in case (2)). If not subscription, one can imagine another model such as polling where the client (the event sync) goes to some place and asks for any new events. Clearly the polling approach is a different message exchange pattern from the subscription approach. IMHO what we are trying to do is come up with a syntax for another message exchange pattern. So its important to agree which MEP(s) we are talking about! My feeling is that we must solve the case (1) MEP and we could solve more if we are so inclined. Bye, Sanjiva.
Received on Thursday, 31 October 2002 18:48:38 UTC