Re: Review of Authoritative Metadata

Roy Fielding writes:

> I do believe that the finding applies equally to SOAP and WSDL.

Indeed.

> Under no circumstances is a directory service more authoritative 
> than the service itself.  A discrepancy may indicate an error in 
> either the directory or the service, but the agent must treat the 
> service as authoritative because it is far more likely to be aware 
> of its own evolution over time than a disconnected directory.

I strongly agree.   Furthermore, I believe it to be the case that the SOAP 
processing model mandates this interpretation.  It specifically dictates 
the normative interpretation of a received SOAP message, and it does so 
based purely on the content of that message interpreted according to (1) 
the  SOAP 1.2 Recommendation and (2)  to the specifications for the 
specific QName'd headers that may appear in the message.   In no case does 
SOAP provide for WSDL to override the correct interpretation of a SOAP 
message.

Of course, when there is a conflict, it's perfectly reasonable to signal 
an error at the WSDL level.   The nature of that error is:  "The message, 
which I correctly interpreted based on the SOAP Recommendation, was not 
what the WSDL led me to expect."   That's as opposed to incorrectly 
saying:  "I used the WSDL to override the normative SOAP interpretation of 
the message."

--------------------------------------
Noah Mendelsohn 
IBM Corporation
One Rogers Street
Cambridge, MA 02142
1-617-693-4036
--------------------------------------

Received on Tuesday, 4 April 2006 20:51:38 UTC