W3C home > Mailing lists > Public > www-ws-desc@w3.org > February 2004

Re: 2004-02-12 Action Item: Clarification to the OperationName feature

From: Umit Yalcinalp <umit.yalcinalp@oracle.com>
Date: Mon, 23 Feb 2004 11:54:43 -0800
Message-ID: <403A5A83.2060601@oracle.com>
To: Jim Webber <Jim.Webber@newcastle.ac.uk>
Cc: Mark Baker <distobj@acm.org>, www-ws-desc@w3.org


Jim Webber wrote:

>Mark:
>
>  
>
>>Ah, right, it's that one again. 8-)
>>    
>>
>
>Indeed it is.
> 
>  
>
>>But as I think I must have said before, you seem to be trying 
>>to make WSDL be something that it isn't.  That may or may not 
>>be a good thing to do, but every use of WSDL I've seen uses 
>>it describe application interfaces, so that's where my 
>>comments are coming from.
>>    
>>
>
>I'm not trying to make WSDL anything else, it already is a message
>description langauge. Some people like to think that it describes this
>mythical application, but I see no justifcation for that. The
>"application" that receives a message described in WSDL might be a human
>reading a fax. How does that tie in with an operation?
>
>I guess it all depends on your view point. Some people see web services
>as a point-to-point means of joining my thing to your thing  (with
>application-specific semantics permeating the network layer). I see it
>as a canonical messaging platform for joining anything to everything
>(minus any baggage from the application layer). 
>
>Jim
>
>
>  
>
Jim,

This discussion is NOT about how people see web services or a 
philosophical way of architecting web services. This discussion is 
about  solving a very practical problem that exists today:

Given an interface there may be more than one exchange that uses the 
same message structure. When an endpoint receive a message, it is NOT 
obvious which message exchange it belongs to since the current WSDL 
design does not convey an identifier of the exchange in the exchange 
itself.

This problem occurs at the interface level. Even if one should use 
distinct enpoint addresses for each service (hence exchanges for an 
interface are identifiable), it is IMPOSSIBLE to distinguish the 
exchange within an interface since there can be more than one exchange 
that may use the same message structure. Further,  there is no 
restriction such that the same address should only be used by one 
service (interface). This complicates the problem even further.

RESULT: one can not distinguish the exchange just by looking at the 
message itself. This is a problem that requires a practical solution 
that requires identifying the exchange. Hence the proposal.

Whatever the underlying architecture does with the messages or how we 
name the message exchanges ("operation" vs. "message_exchange")  is not 
really irrelevant for this discussion. However, the need for a  
mechanism for always identifying the message exchange is.


-umit

-- 
Umit Yalcinalp                                  
Consulting Member of Technical Staff
ORACLE
Phone: +1 650 607 6154                          
Email: umit.yalcinalp@oracle.com
Received on Monday, 23 February 2004 14:55:29 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 5 February 2014 07:15:02 UTC