W3C home > Mailing lists > Public > xml-uri@w3.org > May 2000

Re: Are *relative* URIs as namespace nemes considered harmful?

From: <keshlam@us.ibm.com>
Date: Tue, 16 May 2000 11:06:11 -0400
To: xml-uri@w3.org
Message-ID: <852568E1.0052F3E0.00@D51MTA03.pok.ibm.com>
I'm joining this discussion late, apparently. I hope everyone has seen a
good summary of the discussion that occurred on the Plenary list, so we can
avoid re-re-rehashing the same ground. I'm going to restate my own position
_once_, and then try to shaddup until I have something new to say...



For me, the critical question -- which _nobody_ was able to answer -- was
"How can one write an application that performs namespace-aware document
processing if namespace names are compared after absolutizing." Until that
is answered, I won't be able to consider that a reasonable approach.


Note that this isn't to say it's unreasonable to want to bind a namespace
to a specific location in the web. I just don't think that the namespace
name itself should be required to perform that role; this was explicitly
_NOT_ part of the design scope for Namespaces, and as XML Schemas have
demonstrated it's easy enough to achieve this effect via other mechanisms
(eg another attribute that maps the namespace into a  URI Reference).

As Tim Bray has pointed out, there was no intention in the namespace spec
to support relative references in the first place -- their intent was to
support #locators, and they missed the fact that switching to URIReference
also opened this can of worms. I don't think institutionalizing an admitted
mistake is a good idea.

As things stand, the specs really do conflict. I really believe the Plenary
List's straw poll came up with the best possible compromise:
     Leave the Namespace Name as a literal string in URIReference syntax.
     For purposes of Namespaces, compare it as a literal string.
     Folks who want the absolutized form can calculate it when they need
it.
     Warn users that, while relative URIRef's will work, they really don't
fit the Namespace metaphor.
If this discussion was happening within a single WG, I suspect we'd have
accepted that rough consensus by now. It may not be exactly what everyone
wanted, but I really do think it's a practical solution that everyone could
live with.


However, there is a significant minority which actively disagrees with this
majority position. I don't know if there's a single rationalle, or if it's
several subgroups, and I'm not going to attempt to summarize their
positions because I'm not sure I understand them well enough to do so
fairly.

______________________________________
Joe Kesselman  / IBM Research
Received on Tuesday, 16 May 2000 11:06:25 GMT

This archive was generated by hypermail 2.2.0 + w3c-0.30 : Tuesday, 12 April 2005 12:17:22 GMT