RE: [ws-caf] [WSRF][Fwd: [OASIS members] Press Release to Announce OASIS

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

All,

> 
> After spending a great deal of time yesterday with the Globus and IBM
> folks behind this spec, my feelings are generally unchanged though I
do
> sympathise with what they tried to achieve. In the application domains
> they're thinking about, resources are really important and so they
> figure in their specs. I think I am just concerned about the notion of
> resources permeating down the stack where it would class with the
simple
> services-plus-messages view.

Like Jim and after loooong discussions with the WSRF authors, I think I
have a better understanding of what WSRF is trying to achieve. The use
of WS-Addressing to contextualise interactions with (logical or
physical) resources behind a service boundary is not the main issue.
However, I believe that the conceptual model that is suggested and is
used by some of specifications that are built on top of WSRF may lead to
fragile applications.

I believe that the model encourages the design of distributed
applications around the concept of resources and sharing of those
resources. Instead of documents being exchanged, endpoint references
(EPRs) are exchanged. I have just seen how the WS-Agreement has been
designed. Instead of exchanging agreement documents the entire protocol
is based on the exchange of EPRs. Agreements are not agreed for
conversations (distributed units of work or activities) but, instead,
agreements are used for the creation of resources. Needless to say
agreements themselves are resources. It's only the interactions with the
newly created resources that fall under the terms of an agreement.

This model does not scale and, in fact, they have chosen to model only
bilateral agreements. They suggested to me that multi-party agreements
are not used in real life!!! (yes, I know!). I told them about how easy
it would be to use coordination or transactions to model the negotiation
process for an agreement and then use the agreement to as part of a
contextualised interactions but that doesn't really fit well with their
resource-based model.

So, to answer one of Joe's questions... I don't see a problem with
identifying a file through a WS-Address + some local-to-the-service
information but I see a problem with building scalable applications that
extensively utilise network-wide pointers to resource beyond
organisation boundaries. WSRF encourages just that. Of course you can
choose to not use WSRF in that way but that's its intended use from what
I see here in GGF.

Best regards,
.savas.