W3C home > Mailing lists > Public > www-ws-arch@w3.org > August 2003

Re: Definition for a Web Service

From: Anne Thomas Manes <anne@manes.net>
Date: Mon, 11 Aug 2003 17:45:26 -0400
Message-ID: <00b101c36053$524a29d0$7101a8c0@TPX21>
To: "Francis McCabe" <fgm@fla.fujitsu.com>
Cc: "Cutler, Roger \(RogerCutler\)" <RogerCutler@chevrontexaco.com>, "Jean-Jacques Moreau" <jean-jacques.moreau@crf.canon.fr>, <www-ws-arch@w3.org>

+1.

Exactly right.

The way I see it, the service URI *names* the service (effectively
equivalent to the wsdl:service @name). It is not and should not be construed
to be an endpoint of the service. After all, a service may have multiple
endpoints, and you may want to move the service at some point in the future,
so you don't want the name dependent on a URL. And in terms of
targetResource (which I believe has been dropped), the service URI should
represent the service, not the thing that the service acts upon.

----- Original Message -----
From: "Francis McCabe" <fgm@fla.fujitsu.com>
To: "Anne Thomas Manes" <anne@manes.net>
Cc: "Cutler, Roger (RogerCutler)" <RogerCutler@chevrontexaco.com>;
"Jean-Jacques Moreau" <jean-jacques.moreau@crf.canon.fr>;
<www-ws-arch@w3.org>
Sent: Monday, August 11, 2003 3:58 PM
Subject: Re: Definition for a Web Service


>
> +1
>
> However, I don't think the omission was deliberate.
>
> There are some issues with associating a URI with a service:
>
> IMO the primary purpose of a Web service URI is for *inference* -- we
> can use the same identifier in two different places/times in order to
> *assert* that the service being used is the same service.
>     This is different to the Web service's endpoint information, and it
> is also different to the targetResource.
>
>     Where is can probably *not* be used is in any messages to/from the
> service!!!! Or at least, not without considerable consequences:
> a. If a message to a service mentions the Service URI, and the agent is
> expecting a different URI what gives? In any case, presumably, a
> Service agent already knows its URI and doesn't need telling.
> b. In a composite service, where a reply may come from a different
> agent than the requested agent, none of the entities may be aware of
> the service URI
>
> Frank
>
>
> On Monday, August 11, 2003, at 11:02  AM, Anne Thomas Manes wrote:
>
> >
> > I raised a discussion on the WS-Desc list suggesting that they really
> > should
> > identify a Web service by a URI rather than just a Qname. I was a
> > little
> > surprised by the resistence to such a concept. I got the sense that a
> > lot of
> > people didn't understand what in fact the URI was meant to identify.
> >
> > I don't know what the end decision on the discussion was. I believe it
> > was
> > discussed at the last meeting.
> >
> > But I do think that the architecture group should have some influence
> > on the
> > discussion. If the architecture group believes that a Web service
> > should be
> > named by a URI, then the WS-Desc team should provide a means to
> > capture that
> > name in the WSDL description.
> >
> > From my perspective, a Web service is an "important" resource, and as
> > the
> > Web Architecture says, all "important" resources should have a URI. I
> > also
> > expect that a Web service may be described by a variety of description
> > languages (WSDL, DAML, text documents, etc.) and so there ought to be a
> > means of referring to the Web service that doesn't depend on just one
> > description language (a URI derived from the wsdl:service Qname).
> >
> > Anne
> >
> > ----- Original Message -----
> > From: "Cutler, Roger (RogerCutler)" <RogerCutler@chevrontexaco.com>
> > To: "Jean-Jacques Moreau" <jean-jacques.moreau@crf.canon.fr>;
> > <www-ws-arch@w3.org>
> > Sent: Monday, August 11, 2003 10:47 AM
> > Subject: RE: Definition for a Web Service
> >
> >
> >>
> >> I think that this happened because of all the confusion about URI's
> >> and
> >> QNames.  As I understand it (and I am very willing to admit that I
> >> understand this imperfectly), just about everyone concerned would be
> >> VERY happy to say that Web services are identified by URI's -- except
> >> that currently in WSDL they are identified by a Qname -- which is not
> >> exactly a URI but can be mapped to a URI.  This, at the least, adds a
> >> layer of confusion to any conversation on this subject.  I think that
> >> the basic thinking was that the "Web-related standards" would lead one
> >> sort of inevitably to URI's, and that the detailed issues could be
> >> dealt
> >> with ... in the detailed sections, I guess.
> >>
> >> -----Original Message-----
> >> From: Jean-Jacques Moreau [mailto:jean-jacques.moreau@crf.canon.fr]
> >> Sent: Monday, August 11, 2003 3:45 AM
> >> To: www-ws-arch@w3.org
> >> Subject: Definition for a Web Service
> >>
> >>
> >>
> >> Thanks for the new draft; obviously, this is the result of a lot of
> >> efforts!
> >>
> >> Regarding the new definition for a Web Service: apart from being more
> >> specific (WSDL, SOAP, HTTP), which I like, the other major difference
> >> seems to be that a Web Service is no longer identified by a URI. Is
> >> this
> >>
> >> intentional? Shouldn't this be added back?
> >>
> >> <previousDefinition>
> >> A Web service is a software system identified by a URI [...].
> >> </previousDefinition>
> >>
> >> Comments?
> >>
> >> Jean-Jacques.
> >>
> >> Champion, Mike wrote:
> >>
> >>> Update from the W3C publication team:
> >>>
> >>> New WD of "Web Services Architecture" Document is available at :
> >>> http://www.w3.org/TR/2003/WD-ws-arch-20030808/
> >>
> >>
> >>
> >>
> >>
> >
>
Received on Monday, 11 August 2003 17:55:56 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 3 July 2007 12:25:22 GMT