RE: i0001: EPRs as identifiers - alternative proposal

Hi Tom, all:

Tom sez:
> I do not understand why Reference Parameters would be 
> acceptable but Reference Properties are not?
> If either is present, they are used to identify the instance 
> of what is being communicated to?

Actually, I don't see it that way at all.  The way I interpret it is
that the provider of the EPR has included some "extra" information that
needs to be conveyed to the endpoint.  The spec should say HOW that
information should be propagated, but not anything about WHAT the
information is being used for.  So if your particular system needs to
use this stuff to identify different "instances" (which might mean a lot
of things, including a) different Java objects, b) different sets of
state in a database, etc), that's fine - but I don't need to know
anything about that as a consumer of your service.  If I simply agree to
the contract in the EPR, which in this case includes an "echoback" of
some sort, you can do your work without my needing to look "under the
kimono" and therefore become more tightly coupled to your particular
implementation strategy.

This seems to me a good thing, and very in line with the Web Services

Incidentally, this also seems to me to be analagous to Action.  Some
systems require or desire a separate piece of information, in the Action
case outside the envelope (or at least outside the body), in order to do
their work.  This often means "dispatch", but it could be used for just
about anything.  If I require you to send me a particular Action value,
I'm going to TELL you that you need to do so - just like I would tell
you I need a particular ReferenceP, or for that matter just like I would
tell you you need to use WS-Security.  This information is in the
metadata/WSDL/EPR.  As long as anyone reading the metdata understands
the contracts implied therein (send an Action value of http://foo to
this guy, but this other guy doesn't care), everything will work, and
the senders do NOT need to know what the other guy is using the data
for, unless that information has been exposed explicitly in the

OK this is getting quite long enough, that's all for now.


Received on Thursday, 9 December 2004 15:56:01 UTC