RE: [wsrf] Singleton Resource Pattern

Hey Steve,

Please excuse my non-understanding of the terms that you use but I wonder why “implied” is used. There is nothing implied about the existence of a resource and the fact that it becomes the logical recipient of the messages. The introduction of the “singleton resource pattern” term does not help either.

Here’s what I mean...

When you create a resource through the factory pattern, the resource is not implied! You get back an EPR. At the architectural level you reason about the resource. The destroy message is not to the service; you are not asking for the service to be destroyed, you are asking for the resource to be destroyed. Now, how is the resource implied here? The same with the resource properties.

So, when you suggest that there is a thing as a “singleton resource pattern”, grid service instances come to my mind and we’ve been there. Even if you suggest that there is no 1-1 association between a Web Service and a resource, that’s how WS-RF users are going to treat them. A destroy message to the “singleton resource pattern” service will mean that it’s a request for the resource to be destroyed and, hence, there will be no need for the service to exist (dynamic creation of grid service instances anyone?).

I understand that this is not your intention but from an architecture point of view, this is exactly how it looks like.

But it’s probably just me.

Regards,

--
Savas Parastatidis
http://savas.parastatidis.name

From: Steve Graham [mailto:sggraham@us.ibm.com]
Sent: Monday, July 12, 2004 4:37 PM
To: John DeCarlo
Cc: wsrf@lists.oasis-open.org
Subject: Re: [wsrf] Singleton Resource Pattern

Hi John:
In the case where there is no disambiguation, then this is a trivial subset of the "singleton resource pattern" wherein there is no further "decoration" on the wsa:Address part of the EPR.

It doesn't seem to make sense to me to separate out this use case from the "singleton resource pattern". For example, the various xsd:element definitions that describe EPRs as component of message response payloads, parameters on message requests etc. are well served as being typed as wsa:EndpointReferenceTypes. Help me understand if there is a benefit to typing these components as: "wsa:EndpointReferenceType" or "anyURI". I would observe that an EPR containing just a wsa:Address component contains the information necessary, and has little overhead. It would be good to isolate requestor from the decision made by the service provider: 'full implied resource pattern' or 'singleton resource pattern' or 'no disambiguation at all'. We can do this by using EPR.

sgg

++++++++
Steve Graham
(919)254-0615 (T/L 444)
STSM, On Demand Architecture
Member, IBM Academy of Technology
<Soli Deo Gloria/>
++++++++

wsrf message