The difference the WS-A working group seems to be expousing between Reference Properties and Reference Parameters is that Reference Properties comprise part of the identity of the resource (along with the address), where as Reference Parameters do not constitute any part of the identity. In other words, Reference Parameters are like cookies on the Web that are used to identify transient sessions, for example. Given this viewpoint and the remove of Reference Properties, the "correct" place to push this information is into the Address, rather than the Reference Parameters. But I agree that Reference Parameters could technically be used for what we are currently using Reference Properties. In either case, the information still shows up in the SOAP header of the request message -- either in the wsa:To header, or in the Reference Parameters element. Note that this who discussion does not suggest to me that we should move away from WS-A. The reason I've always liked the move to WS-A was that it clearly defined how information from the EPR was required to show up in the SOAP header of a message -- the EPR address must be put in the wsa:To header, and the Reference Properties must be copied into the header. This allows dispatch to be performed solely on the contents of the SOAP message. Previously the address to which a message was being directed would only show up in the http post header of the message, which made message dispatch somewhat more complex. So the practical ramification of this WS-A change on GT, for example, is simply that we get the service address and resource identifier from the wsa:To SOAP header, rather than from some other SOAP header that came from the Reference Properties. -Steve On Jan 22, 2005, at 6:47 AM, Tom Maguire wrote:
I have to agree with Savas w.r.t. Reference Parameters. The issue at hand in the WS-A working group was the 'identity' semantic of Reference Properties vs URIs. I do not think that reference parameters will be removed from the spec and reference parameters can be used for service-side dispatch.
Tom
Savas Parastatidis wrote:
Dear all,
I agree with some of what Steve said bellow :-))
The change does make a difference in implementations because if the information that was meant to be included in the Reference Properties is now encoded in the Address property of an EPR (the URI) it cannot be echoed back as SOAP headers. This was part of the semantics of Reference Properties.
For a resource-oriented specification like WSRF, however, using URIs to identify resources is a good solution. On this I agree with Steve. The semantics of WSRF do not have to change even if it wasn't abstracted away from WS-Addressing. What is lost is the ability to echo back as SOAP headers any resource-specific information that may be required by a SOAP intermediary to identify the targeted resource. If having SOAP headers is a requirement, though, I don't see why Reference Parameters couldn't be used. So, while Reference Properties have been removed, Reference Parameters are still there and while I would personally like to see them go too, I don't think it'll happen. The semantics of Reference Parameters are slightly different but I don't see why WSRF couldn't use them, if it needed to, instead of Reference Properties to carry resource-specific information.
Best regards, -- Savas Parastatidis http://savas.parastatidis.name
-----Original Message----- From: owner-ogsa-wg@ggf.org [mailto:owner-ogsa-wg@ggf.org] On Behalf
Of
Steve Tuecke Sent: Friday, January 21, 2005 9:55 PM To: Ian Foster Cc: 'Stephen Pickles'; 'OGSA-WG'; Steve Tuecke; Djaoui, A (Abdeslem) Subject: Re: [ogsa-wg] FW: Issue #1 proposed resolution
As far as I can tell, this decision has basically no effect on WSRF. The argument being made by some in the WS-A working group is that it
is
equivalent and more true to the Web to carry a resource identifier as part of the EPR address, rather than in a separate ResourceProperties field -- that is, the resource reference should all be in the URI, rather than split between a URI and separate resource properties. Implementation-wise it certainly makes very little difference. And
the
WSRF working group had already abstracted the WS-Resource reference
and
access pattern, so that it is not tightly coupled to WS-A and
reference
properties anyway, so WSRF specification-wise it makes no difference.
-Steve
On Jan 21, 2005, at 1:59 PM, Ian Foster wrote:
I think the technical term is "carefully architected set of specifications" not "house of cards" (-:
Regards -- Ian.
At 05:10 PM 1/21/2005 +0000, Djaoui, A (Abdeslem) wrote:
Well, it probably doesn't, because WSRF is now decoupled from WS-Addressing through the definition of the "abstract" resource Access Pattern, which defines different embodiments for different ways of accessing
state.