Re: Updated portTypeExtension proposal

Hi Gudge,

> > I prefer:
> > 
> >     <portType name="ncname" extends="list-of-qnames">
> >         ...
> >     </portType>
> > 
> > rather than the "bases" attribute. 
> 
> Yeah, we talked on the call about working on the names later in the life
> of the spec so that we can come up with a consistent set when things are
> more stable.

Understood. 

> > Also, similarly for the service:
> > 
> >     <service implements="list-of-qnames">
> >         ...
> >     </service>
> 
> Err, there is no such attribute on service ( only has a name attributes
> ). The mapping for the port types property now drills down via the
> binding attribute on the port element. This makes sure things are
> internally consistent thus avoiding possible mistakes where the list of
> port types in an 'implements' attribute doesn't match the list of port
> types you can actually get to via the bindings. It also avoids having to
> state such a constraint in the spec.

I understand - this is basically what's in WSDL 1.1 too. However,
we sould like to have the service indicate explicitly what its
portTypes are - that's a cleaner model than having to go down via 
the bindings and discover the abstract functionality of the service.

Could we please add this attribute? The value would be a list
of QNames of portTypes and the service element MUST provide at
least one binding for each portType. If there's more than one
binding then they are to be considered alternate access mechanisms.
(That's all the same semantics we have now I believe.)

If you don't want to add it to the proposal please do make it
an issue against it. 

Thanks,

Sanjiva.

Received on Monday, 14 October 2002 12:26:02 UTC