Re: XP Service URIs

On Tue, Nov 21, 2000 at 04:22:23PM -0500, Ray Denenberg wrote:
> Mark Nottingham wrote:
> 
> > Can we assume that every XP service has an identifiable URI?
> 
> It's been a fundamental premise in my thinking, except for the case where
> two parties have prior agreement.
> 
> > If so, perhaps
> > we should consider requiring this?

Sorry, not requiring. I meant formalizing it in the model, as expressed by
the glossary, etc. 

re: when the parties have prior agreement; yes. I think that the 'prior
agreement' scenario has many implications and possible optimisations, but
IMHO we shouldn't plan for it. I.e., let's design a conservative protocol
and if people want to cut corners, etc. when they own both ends, so be it.
At some point there should be a discussion about how this can be expressed;
it may be explicit along with each protocol mechanism, or it may be
summarized in a separate section (my preference). But I digress...


> Depends what we mean by "requiring".
> 
> > .... it would also be good to require the envelope to identify
> > the URI, ........
> 
> I think it would be good to require the envelope to *support*
> identification of the URI. I don't necessarily think that the URI should
> be a required element.

I know that it's a concern that we not stuff too many things into the 'core'
specification, but IMHO service identification is critical for a number of
reasons; it is likely to be the basis of most XP modules that are created
(e.g., with caching, it helps define the namespace; with authorization, it
identifies the realm, with routing, it identifies the target, etc.).
Additionally, it makes a 'virtual hosting' style service much simpler.

I'd prefer something something more detailed than just supporting it (taking
into account previous discussion about "supported" functionalities).



-- 
Mark Nottingham, Research Scientist
Akamai Technologies (San Mateo, CA)