- From: Sanjiva Weerawarana <sanjiva@watson.ibm.com>
- Date: Fri, 8 Aug 2003 09:10:51 +0600
- To: <www-ws-arch@w3.org>
"Paul Denning" <pauld@mitre.org> writes: > > Martin has an action item from the 2003-08-7 telecon, but I thought I would > prime the pump with some thoughts I had during the telecon. > > Given that WSDWG seems to be taking the notion of RPC (or the ability to > describe RPC) out of WSDL, we need to figure out a few things. I disagree totally. WSDL's abstract part never had the notion of RPC. What it had always was the notion of a one way message or a request-response pattern of messages. The SOAP *binding* is the only thing that used the word "rpc" and that was used to mean "apply the rules of SOAP RPC as indicated in section 7 of SOAP 1.1 to generate a wrapper element for the <part>s of the message." I don't believe the direction of the WS-Desc WG has changed these fundamental concepts. What we have indeed decided is not to have the binding rule for auto generating a wrapper, but the message exchange patterns are still there and still the same (and much more powerful for extensibility). > 1. What should WSAWG say about RPC? > 2. Where in the WSA stack diagram would RPC fit? > 3. If I wanted to describe an RPC web service, how do I describe it? > 4. Is RPC a higher layer thing that Choreography should describe? > 5. Is RPC a lower layer thing that currenty has no formal description > language? > 6. Is RPC a "feature" [1]? > 7. Should RPC be included in the Message Oriented Model [2]? > 8. Should description of RPC be a WSDL extension? IMO WS-A should talk about message exchange patterns as described in WSDL and not about RPC directly. RPC's can indeed be modeled by WSDL.next just fine as a request-response (or input-output, whatever the thing is called today) MEP and thus that's all WS-A needs to consider. YMMV. ;-) Sanjiva.
Received on Thursday, 7 August 2003 23:37:21 UTC