> >Furniss: Aren't there also cases where the BPEL process requires that
> >the communication be via a correlating protocol, but doesn't mind
which
> >one ? The requirement from the process is just that the communication
> >be "connection-oriented" at application level - the application is
not
> >going to use its own data values to discriminate, and isn't
interested
> >in what the underlying protocol uses, so long as the conversations
can
> >be distinguished. Which underlying protocol is actually used is
> >deferred to the binding.
> >
> >Or is this out in abstractional hyper-space ?
> >
> When we move outside of WSDL- and Schema-described XML
> messages, don't
> we lose some control? Doesn't this opacity bring risk with it? What
> other issues have we not identified (plus allowing use of proprietary
> specifications)?
I didn't read Peter's remark as requiring to "move outside of WSDL- and
Schema-described XML messages". I think his point includes cases where
the correlation protocol is a standard, WSDL- and XML-based WS
correlation protocol, but such that the BPEL language does not get to
directly manipulate it. The analogy would be the use of a standard WS
reliability protocol: even if it's there, the BPEL language itself does
not get to manipulate it.
Ugo
