Re: ISSUE : Extensible message exchange patterns

Perhaps I'm being overly simplistic, but this just doesn't seem that hard to
me.  WSDL 1.1 has hard-wired concepts of "input" and "output" message types.
If we accept the fact that those type names shouldn't perhaps be in the WSDL
spec directly, but associated with an MEP, we might get something like this:

<operation name="foo" mep="http://.../simple-req-resp">
  <message type="request" name="FooRequest"/>
  <message type="response" name="FooResponse"/>
</operation>

The MEP spec determines the names and disposition of the various allowed
message types.  We could do something like this and mandate one or more MEPs
which must be understood by all WSDL processors....

--Glen

----- Original Message -----
From: "Jean-Jacques Moreau" <moreau@crf.canon.fr>
To: "Christopher Ferris" <chris.ferris@sun.com>
Cc: "Sanjiva Weerawarana" <sanjiva@watson.ibm.com>; "Jeffrey Schlimmer"
<jeffsch@windows.microsoft.com>; "Glen Daniels" <gdaniels@macromedia.com>;
<www-ws-desc@w3.org>
Sent: Thursday, June 06, 2002 7:49 AM
Subject: Re: ISSUE : Extensible message exchange patterns


> Hi Chris,
>
> This sounds good. Would the next step then be to handle the existing
> WSDL MEPs (e.g. request-response, sollicit-response) the same way, i.e.
> to outsource their definition and give them a URI?
>
> Jean-Jacques.
>
> Christopher Ferris wrote:
>
> > I'm at a loss here. While some vocabulary that defined
> > MEPs with angle-brackets might be a good thing(tm), it isn't
> > at all clear to me that it is needed. SOAP1.2 defines
> > MEPs and assigns URIs to these formal definitions. It also
> > recommends in the binding framework that MEPs be named
> > with a URI[1].
> >
> > I think that rather than treat this through extensibility
> > in WSDL, that a binding identify the MEPs it supports
> > with a URI (or qname I suppose, but we may need to do some
> > coordination on that).
>

Received on Thursday, 6 June 2002 10:58:24 UTC