RE: Issue LC50 - MEPs

>-----Original Message-----
>From: Sanjiva Weerawarana [mailto:sanjiva@watson.ibm.com] 
>Sent: Sunday, Nov 21, 2004 22:48 PM
>To: Yalcinalp, Umit; 'Amelia A Lewis'
>Cc: www-ws-desc@w3.org
>Subject: Re: Issue LC50 - MEPs
>
>
>"Yalcinalp, Umit" <umit.yalcinalp@sap.com> writes:
>> 
>> Are you suggesting that the input message and the output message 
>> in an inout pattern do not NEED to share the binding? What you 
>> seem to suggest that the pattern allows different bindings to 
>> apply to the incoming and outgoing message. We don't allow this, 
>> the binding is for the entire pattern hence governs all the 
>messages. 
>
>Yes that's what I'm suggesting. Does our binding disallow the
>use of WS-Addressing? If not one can easily switch bindings by
>using ReplyTo!

No, it does not disallow WS-Addressing, but you are making an interesting point that I never thought that we allowed before. 

There are two issues here: 

-- Do we really enable two separate bindings to apply to a single pattern (request-response) or not? The way the spec is written today, taking WSDL as a contract, the client believes that a single binding governs both the incoming and the outgoing message hence the same transport/protocol. My interpretation has been, and I am asking this question to the rest of the wg, is that there is a single binding for an MEP only. Do others have the same understanding? 

-- Lets assume that WS-Addressing allows us to override this mechanism. Where does the service enable such an override mechanism? It should be obvious to the client.
After all, WSDL is the contract that is provided by the service itself. Hence, it is the service that makes it obvious to the client that single binding per MEP is overrideable somehow. Wouldn't you agree?  

>
>Practically, its not likely that someone will switch out of SOAP ..
>or whether that will even work! However, IMO its perfectly reasonable
>to receive a request response over SOAP/HTTP and get the response
>via SOAP/SMTP.

Wouldn't you agree that this is not a complete contract on its own right however? For example, if I were to override via ReplyTo to send the response via a different binding, the service provider as the provider of the WSDL and the service NEED to support the alternative binding indicated by the ReplyTo? Where is this information supplied in WSDL? 

Having the capability to override using WS-Addressing is necessary but NOT sufficient. 

For example, if you want me to send the reply via pigeons to you, you need to know that I can send the reply to you via pigeons, right? If nowhere in my WSDL that I provide to others I made any mention of pigeons, i.e. there is no indication of a binding (transport/protocol) supported for that service, I can not support a ReplyTo header as I will not know what to do with it as I will not support a different binding, right? 


>
>Sanjiva.
>

I am trying to understand what we are really enabling here and how complete the WSDL contract is as it appears to be a bit different than what we specified. If that is the case, we need to be clear. 

Thanks. 

--umit

Received on Tuesday, 23 November 2004 01:33:25 UTC