RE: [wsrf] Modelling sessions

["Savas Parastatidis" <Savas.Parastatidis@newcastle.ac.uk>]

Dear Thomas,

 

Thank you for your message.

 

As Sastry pointed out, the document was not meant to come into any conclusion. We were given the task of investigating the possible approaches that could be taken and leave it to the working group and community to decide what is the more suitable approach for meeting requirements.

 

My questions to this list is an attempt to continue the investigation and try to understand in which situations WS-RF should be used and whether there are any implications when doing so. As you rightly suggest, the resource-oriented view of the world that WS-RF suggests certainly does not apply in every situation. However, we have to identify those situations and try to identify the issues related to the application of the resource-oriented pattern in a service-oriented world. For example, what are the implications of using a resource-oriented pattern when used with a protocol like WS-TransactionManagement that only knows about services?

 

It was the case that in the OGSI world session-based interactions were modelled using service instances. Since WS-Resources are seen as the equivalent of service instances I think both the investigation by the DAIS WG and the questions here are valid. The answers so far have shown that WS-RF is not meant to be used for sessions. Instead, a protocol like WS-Context is the more natural choice and since WS-Context defines its own lifetime management requirements and behaviours, there is no need to use WS-RF to manage the sessions themselves.

 

Best regards,

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

---

From: Thomas Freund [mailto:tjfreund@us.ibm.com]
Sent: Thursday, May 27, 2004 5:13 PM
To: Savas Parastatidis
Cc: David Snelling; Sastry Malladi; Simon Laws; wsrf@lists.oasis-open.org
Subject: RE: [wsrf] Modelling sessions

 

Savas & all,

Reading the referenced article http://forge.gridforum.org/projects/dais-wg/document/draft-ggf-dais-mappings-ggf11/en/1
by Savas, Sasty & Simon it is clear what you are is citing scenario 2, 'Sessions'. However, the article clearly points out that a range of possibilities exist (i.e. after all the WS-* standards are meant to be composable). So scenario 2 in the article illustrates that a.) sessions & resources can be represented entirely using WS-RF (5.2.3.1) or b.) sessions can be represented using either WS-I or WS-Context along with a WS-RF resource (5.2.3.2)

The article makes no value judgement. It also apparent from some of the responses that others have repeated this message (that depending on the requirements) a vendor has the flexibility to implement in any fashion he choose. I believe the important thing here is to emphasis that this is not an either/or proposition (i.e. the answer it's not necessarily all WS-I/WS-Context nor is it all WS-ResourceFramework).

My preference for scenario 2 would be to model sessions as WS-I/WS-Context and stateful resources using WS-RF. It follows a 'pattern' which we are very familiar. However I'm not sure that answer necessarily satisfies everyone & certain it does not apply to every scenario (or as I said before that one pattern does not necessarily 'fits all' situations).

Regards,
Tom