Lists Home |
Date Index |
On Thu, Aug 12, 2004 at 02:41:21PM -0400, Michael Champion wrote:
> So, the question seems to me to boil down to whether this
> needs standardizing, in which case XML seems like the obvious choice,
> or whether it should remain opaque implementation detail, in which case
> a URI seems like the obvious choice.
I look at it a bit differently.
The URI is the identification mechanism of the Web, and an EPR is an
identifier. So from that perspective, I think it's quite clear that an
EPR should be a URI. The question then, I believe, is whether or not
Web service agents should be peeking into URIs to pull out "reference
properties and parameters". The TAG has made its view clear on this
question, at least;
"Agents making use of URIs SHOULD NOT attempt to infer properties of
the referenced resource except as specified by relevant specifications."
> IMHO, standardization of this stuff makes sense, even though *at
> present* the different implementation components that need to exchange
> routing information tend to be proprietary or custom-written. After
> all, you have a bunch of fierce competitors coming to the W3C and
> saying that they would like to see a Recommendation for this so that
> they can interoperate better. I'm inclined to give them the benefit of
> the doubt.
Hmm, I guess I'll be nice and just say "no comment". 8-)
> Also, at long last we seem to be getting past the
> "technical" debates that for some inexplicable reason :-) tended to
> line up on industry alliance boundaries. The absolutely last thing the
> W3C should do, assuming they want to ever get a web services submission
> again, is say "thanks, but no thanks, we don't think that needs to be
Well, I can see it both ways, but I agree with Bill that they should be
accepted or rejected based on their technical merit. So far, the W3C
has been doing, IMHO, a very poor job at submission evaluation.
P.S., there's a great example of an EPR in the spec which is just
begging to be a URI;
<wsa:EndpointReference xmlns:wsa="..." xmlns:fabrikam="...">
Compare with one of these, or another of your making ...