RE: Architectural recommendations in the SOAP 1.2 Rec

> -----Original Message-----
> From: Mark Baker [mailto:distobj@acm.org]
> Sent: Tuesday, June 24, 2003 11:20 PM
> To: www-ws-arch@w3.org
> Subject: Architectural recommendations in the SOAP 1.2 Rec
> 


> 
> - identify resources by URI, not by other means

Right.

> - use the uniform interface constraint for retrievals

Huh? What leads you to that conclusion?  SOAP 1.2  suggests using the GET
web method feature when appropriate, but has a list of caveats.  You seem to
have missed the caveats.

> 
> I believe this requires one of two things;
> 
> - REST style Web services given preference, as they have already
> incorporated both those constraints, or
> - some SOA style extension which includes these two constraints,
> but which is also given preference

How about if you get David Fallside to explain to us that this is REALLY
what SOAP 1.2 had in mind, and then I'll be glad to take your interpretation
seriously.  <duck>  Remember that several of us sat through hours and hours
of discussion of this topic on XMLP telcons, and I for one think you are
signfificantly over-interpreting the implications of SOAP 1.2.

Or maybe I'm over-interpreting what you're asking for.  I for one have no
problem restating the consensus of XMLP in the WSA document:  The SOAP GET
binding really SHOULD be seriously considered if the following conditions
apply: HTTP is the only protocol that the service must operate over, it's
practical to identify service endpoints with URIs, retrieval of a data is
the purpose of the service, the retrieval operation has no side-effects in
the underlying software, and there are no security or privacy concerns
making result cacheing or URI-based service invocation problematic. That's
very good advice IMHO, but I'm not sure how this implies "REST style Web
services [be] given preference."

Received on Wednesday, 25 June 2003 01:08:39 UTC