Paul Kiel wrote:
> Hi Dave, 
> I've been through this discussion as well and in the end, there was a
> decision to use the payload schema.  In essence, the generic wsdl just
> renders the soap message as a less useful envelope that is discarded (hence
> the REST comment from Stefan).  Now most people end up discarding it anyway,
> but somewhere you need to model the expected request-responses.  If it isn't
> in your wsdl, then you need to accommodate this in the application.  What is
> the application expected to respond with?  What should it do if it fails?
> How will it know what to validate the payload with?  These questions need to
> be answered somewhere.  Why not put them in the wsdl so that it is part of
> the explicit agreement between trading partners.  If you push this into the
> application, you run the risk of people making assumptions that later become
> problems.
>
> At least those were some of the arguments we experienced.
> FWIW.
> Cheers,
> Paul
>
>
>  
Yeah, what I've been reading, and even reviewing in various WSDLs, is 
that a generic approach would require an additional out of band 
communication.   You have to provide another mechanism that would 
indicate which services or messages your application would accept to 
processs.   I believe some organizations may use the WS-MetaData 
specification to accomplish this secondary out of band 
communication.     I would assume that there would be much more 
application specific routing of information to get it to the right 
place, where with the specific the destination and service is already 
defined in the wsdl, so less routing of information would need to occur.

I'm also finding that this is particular design topic is about as bad as 
which is better RelaxNG or XML Schemas or Schematron.  Everybody likes 
their way of doing things and thinks it's the correct way.

Anyways, thanks for the comments, it helps to get outside opinions.

Dave