- From: Steve Tuecke <tuecke@mcs.anl.gov>
- Date: Fri, 20 Jun 2003 08:49:08 -0500
- To: "David Booth" <dbooth@w3.org>
- Cc: www-ws-desc@w3.org, Mark Baker <distobj@acm.org>, "Sanjiva Weerawarana" <sanjiva@watson.ibm.com>, "Champion, Mike" <Mike.Champion@SoftwareAG-USA.com>
+1 Note that we had a very similar issue in OGSI (see section 7.5.2.1 of [1]), and reached the same conclusion described by David. -Steve [1]. http://www.gridforum.org/ogsi-wg/drafts/draft-ggf-ogsi-gridservice-29_2003-04-05.pdf At 02:14 PM 6/19/2003, David Booth wrote: >It looks to me like there is some misconception of what the targetResource >means, partly (I think) because some of the statements in this discussion >have been a little imprecise. > >The targetResource attribute has nothing to do with describing a >service. It is used to indicate a relationship *between* services. Its >purpose is to allow two WSDL <service> descriptions, d1 and d2, to assert >that (behind the scenes) the services s1 and s2 that they describe are >actually "manipulating" the same resource. In other words, if d1 and d2 >both state "targetResource='u'", where u is some URI, then they have >asserted that s1 and s2 "manipulate" the resource r that is identified by >URI u. > >As we know from RFC2396[1], a resource can be anything -- a physical >object, an abstract concept -- anything. So what does it mean to say that >s1 and s2 "manipulate" the same resource r? Without knowing the semantics >of d1 and d2 you don't know. That is not defined by the WSDL 1.2 >specification. (Nor should it be, IMO.) Until you know the semantics of >d1 and d2, the only concrete thing you can conclude is that s1 and s2 are >somehow related to each other through r. > >(Just in case there is confusion about this, the "targetResource='u'" >attribute is NOT asserting that s1 and s2 are the same resource as each >other, nor is it asserting that s1 and s2 are the same resource as r.) > >Does this vagueness present a problem? No. Different applications will >know what they wish to do with this. (The canonical example is a >printDocument service s1 and a managePrinter service s2, both manipulating >the same physical printer r.) The reason the WG described the >targetResource as "manipulating" the same resource was to give users >guidance about its intended use, even though the precise meaning of the >word "manipulate" is impossible to nail down in this context. > >Regarding the name "targetResource", u does identify a resource, so the >"Resource" part of the name definitely is appropriate. Furthermore, >proponents of the targetResource attribute like to think of r as being the >ultimate "target" of interactions with s1 or s2; hence the name >"targetResource". > >I hope this helps to clarify the situation. > > >1. RFC2396: http://www.ietf.org/rfc/rfc2396.txt > > >-- >David Booth >W3C Fellow / Hewlett-Packard >Telephone: +1.617.253.1273
Received on Friday, 20 June 2003 09:49:56 UTC