The WSRP use case is definitely not
the resources talking to each other. Rather it is a class of web services
that require access to information stored in stateful entities in order
to properly handle incoming messages. While some use cases may have the
web service and the stateful entity being one and the same, the current
WSRF semantics and syntax do not require it.
Rich Thompson
OASIS WSRP TC Chair
"Savas Parastatidis"
<>
08/12/2004 06:29 AM
To
"Anish Karmarkar"
<>
cc
Steve Graham/Raleigh/IBM@IBMUS,
<>, <>
Subject
RE: [wsrf] WS-Addressing
submitted to W3C as input
Hi Anish,
>
> But taking a step back, I am not sure what EPR is trying to do --
is
it
> trying to provide a referencing mechanism for a Web service/endpoint
or
> provide such a reference with additional bag of coordination/context
> related data. Both seem like a useful thing to do, but my concern
is
> that EPR construct does not provide a clear separation of the two.
I.e.,
> provide a referencing mechanisms at the base layer and build on it
by
> providing a way to including coordination/context related data (thru
ref
> properties/ref parameters). It seems like architecturally it makes
sense
> to separate the two.
>
You probably know that I agree. I would go a step further though...
Architecturally, we all agree that we need a message-level mechanism for
addressing _services_. I believe the reference properties/parameters
elements are misused by WS-RF to convert the addressing mechanism for
services to addressing for _resources_ at the architectural level. There
is nothing wrong with that, if that's what you want to do, but we should
be honest and say that services are irrelevant in this architecture. All
we have is resources talking to each other and we have addresses for
them. (Please note that I am talking about the architecture and not how
that is implemented with existing WS tooling and specs).
I can be persuaded to see the usefulness of ref properties/parameters
(for example in the way they can be used in Ws-Eventing) in general but
not as a mechanism for introducing architectural changes at the
realisation level which is where Web Services reside.
But then again... I am only the minority voice here :-)
.savas.
