Identifying WS-Resources: ReferenceProperties or ReferenceParameters?

[A Djaoui <a.djaoui@rl.ac.uk>]

Hi Steve

I copied and pasted into a new thread since it looks this topic belongs on its own.

Steve Graham wrote:

Hi Abdeslem:


"Djaoui, A (Abdeslem) " <A.Djaoui@rl.ac.uk> wrote on 08/13/2004 07:34:34 AM:


> The introduction of Reference Parameters and the clarification on EPR comparison
> will benefit WSRF and help eradicate some of the question on how WS-addressing is
> used in WSRF. I never was comfortable with the way ReferenceProperties were used to
> identify WS-Resources. Reference Parameters seem to be more appropriate for that.
This is where we need further debate.
First, the EPR comparison section is somewhat helpful, but not terribly so.  For example,
it says nothing on reasoning about EPRs that are different. It is quite possible that
two EPRs differ, but still refer to the same resource. This situation is much more interesting
and happens quite a bit in systems management scenarios.  This is where the heart of
the matter actually lies.

Now, on the preference for Ref Parms to identify (disambiguate) resources, I am not sure
that is the right approach.
The WS-Addressing (Aug 2004) spec says (section 2.1):
[reference properties] : xs:any (0..unbounded).
A reference may contain a number of individual properties that are required to identify the entity or resource being conveyed.
So, (avoiding the grumbling about the use of the word identify) the spec suggests Ref Props,
not Ref Parms, are more appropriate for the sorts of things we are doing in WSRF.

sgg

Isn't this then in contradiction with the new WS-Addressing spec that suggests "A consuming application should assume that different XML schema, WSDL definitions and policies apply to endpoint references whose address or reference properties differ"? The way it is used now, we have  different referenceproperties associated with the same schema, wsdl and policy.

I concluded from the above statement that referenceproperties are used to allow the use of the same address by many different services (endpoints) (potentially different in schema, wsdl or policy). An example is a load balancing service that is a front for a server farm where different requests are routed to different machines depending on their policy requirements for example.
Now where does the reference parameters come in.
The spec says

[reference parameters] : xs:any (0..unbounded).
A reference may contain a number of individual parameters which are associated with the endpoint to facilitate a particular interaction

Each enpoint in the server farm can additionally use reference parameters to identify/disabiguatre WS-resources (subscriptions for example). So we have some sort of two level referencing. The first allows the use of one adress by many endpoints and the second allows the use of one endpoint by many ws-resources.

Abdeslem
///////////////