W3C home > Mailing lists > Public > www-ws-desc@w3.org > July 2004

Re: Action Item 2004-07-01 Solution to 168/R114

From: Sanjiva Weerawarana <sanjiva@watson.ibm.com>
Date: Tue, 13 Jul 2004 21:21:36 +0600
Message-ID: <058e01c468ed$19732190$f24e4109@LANKABOOK>
To: "WS Description List" <www-ws-desc@w3.org>

Wouldn't this scenario be best solved by defining operation A
(the one that's gone out to lunch) to be one that returns an
element whose content model is a choice between Y and Z? That
is, I don't think this is a good example to motivate leaving
dispatching out of band.

I'm +1 to leaving dispatching out of band on the basis that
its the server's business to know how to dispatch and the WSDL
is what the server has decided to tell *the client*. There's no
need for the server to tell the client how *it* does its internal
work.

The only case I've seen to justify including that is for "industry
standard" WSDLs which are intended to be implemented by different
service providers. That is, the WSDL there does not describe
a single service offered by someone, but the structure of services
to be offered  by others. In that case however I imagine the 95%
scenario will be unique GEDs or some other form that is patently
obvious to the service implementor.

My proposal supports both these to work happily. I'm not mandating
that the SOAPAction based solution be used, but it does provide a
mechanism to make "dispatching" clear when necessary. At the same
time, we are not precluding other mechanisms (like the one Gudge
mentioned) from doing what they want.

Sanjiva.

----- Original Message ----- 
From: "Martin Gudgin" <mgudgin@microsoft.com>
To: "David Booth" <dbooth@w3.org>; "Jeffrey Schlimmer"
<jeffsch@windows.microsoft.com>
Cc: "Umit Yalcinalp" <umit.yalcinalp@oracle.com>; "WS Description List"
<www-ws-desc@w3.org>
Sent: Tuesday, July 13, 2004 4:00 PM
Subject: RE: Action Item 2004-07-01 Solution to 168/R114


>
>
> Let's take an interface with operations B and C both of which have the
> same input message, X. Operation B has an output message Y, while
> operation C has a different output message Z. Both B and C use the
> In-Out pattern.  Whether you get message Y or Z back depends on the
> content of X. Let's for the sake of argument say that if a particular
> value in X is over 1000 you get Z, otherwise you get Y.
>
> I believe that this is a coherent (if somewhat simplistic) example in
> messaging systems. I also understand that it does not fit particularly
> well into the RPC style. And that the WSDL does not describe the details
> of how the server determines whether to send Y or Z. C'est la vie. There
> is still enough information in the WSDL to construct messages that the
> service will accept and to deconstruct messages the service will emit,
> that is to interoperate with the service.
>
> Some of you are wondering what happened to operation A. But that's
> another story...
>
> Gudge
>
> > -----Original Message-----
> > From: www-ws-desc-request@w3.org
> > [mailto:www-ws-desc-request@w3.org] On Behalf Of David Booth
> > Sent: 08 July 2004 17:40
> > To: Jeffrey Schlimmer
> > Cc: Umit Yalcinalp; WS Description List
> > Subject: RE: Action Item 2004-07-01 Solution to 168/R114
> >
> >
> > At 02:30 PM 7/7/2004 -0700, Jeffrey Schlimmer wrote:
> >
> > >WSDL 2.0 should not require identifying the operation name
> > because doing
> > >so will unnecessarily limit the applicability of WSDL 2.0.
> >
> > Can you give an example?
> >
> > >R114 mandates that the WSD language define a way to uniquely
> > map, but it
> > >does not mandate that each WSDL document must uniquely map.
> >
> > The current wording of R114 is indeed ambiguous ("R114: The
> > description
> > language MUST allow unambiguously mapping any on-the-wire
> > Message to an
> > Operation.").  It isn't clear whether the "MUST allow" verb
> > applies to the
> > _mapping_ or the _writer_of_a_WSDL_document_, i.e., whether
> > it MUST allow
> > any message to be mapped to an operation (this would be the stronger
> > interpretation), or whether it MUST allow a WSDL document to
> > be written
> > such that any message can be mapped to an operation (this
> > would be the
> > weaker interpretation).  Also, the wording of this
> > requirement somehow
> > changed (weakened) after the WG agreed to it on 4 April 2002,
> > though I
> > can't find anything in the minutes to justify the change.
> > (Here is the
> > chronology that I found:
> > http://lists.w3.org/Archives/Public/www-archive/2004Jul/0021.html )
> >
> > However, I think the precise wording of R114 is somewhat
> > irrelevant.  The
> > real question is what does the WG think we need.
> >
> > Jeffrey, are you suggesting that you think Scenario X (
> > http://lists.w3.org/Archives/Public/www-ws-desc/2004Jun/0300.html )
> > is an acceptable situation and is not a interoperability
> > problem that we
> > need to solve?
> >
> > >The RPC style (http://www.w3.org/2004/03/wsdl/style/rpc)
> > defines a way
> > >to uniquely map and therefore satisfies R114. Nothing else is needed.
> >
> > Again, that depends on your interpretation of R114.  Unique GEDs also
> > provide a way to uniquely map.  Personally, I think the weak
> > interpretation
> > of R114 would render R114 somewhat pointless, because the
> > author of a WSD
> > can always simply write the WSD to use unique GEDs -- nothing
> > special is
> > needed in the WSDL 2.0 spec to facilitate this.
> >
> >
> > -- 
> > David Booth
> > W3C Fellow / Hewlett-Packard
> > Telephone: +1.617.253.1273
> >
> >
Received on Tuesday, 13 July 2004 11:22:15 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:58:32 GMT