RE: Minority objection to requiring unique GEDs or required feature to distinguish operations

Mark:

> Hmm, well do you realize that every interface definition 
> language I'm familiar with, including OMG/ISO IDL, MIDL, RMI 
> remote interfaces, DCE IDL, and every application protocol 
> ever created, has been similarly "fouled"?  They each define 
> a dispatch mechanism.  So I think we're on pretty safe ground 
> requiring it be unambiguous.

Perhaps this is why they failed to decouple the architectural model from
implementation. In CORBA for example you have the architectural model of
distributed objects, and low and behold folks implemented their IDL
interfaces with concrete objects.

(Yes, I am aware that it doesn't _have_ to be like this, but because of
the constraints imposed by the model and glued together by IDL it
generally happens this way).

My preferred model is not to couple my contract to my implementation via
the dispatch mechanism - it doesn't afford me the "room to manoeuvre"
that I ideally like for the evolution of my back-end systems.
 
> FWIW though, I was just thinking that if it were specified 
> that an agent processing the WSDL should interpret the 
> absence of this information as being semantically equivalent 
> to the information being unrecognized, then that would be 
> sufficient to address my concerns.

I don't have a problem with that, providing that we're not prescriptive
about what to do in the "unrecognised" case.

If "unrecognised" is accepted without adding further semantics then
surely it covers both cases - those like Jacek has exemplified [1] where
the developer wants (and indeed needs) to advertise the dispatch
mechanism, and those like mine [2] where I see the collection of
operations collectively describing message exchanges (and nothing more).

Jim
--
http://jim.webber.name 

[1] http://www.jacek.cz/blog/archives/000038.html 
[2]
http://jim.webber.name/2004/08/16/f39b2516-1244-48c1-a814-7165d00dddc2.a
spx

Received on Wednesday, 15 September 2004 03:11:10 UTC