W3C home > Mailing lists > Public > www-tag@w3.org > December 2002

Re: uri-comp draft necessary?

From: <noah_mendelsohn@us.ibm.com>
Date: Thu, 19 Dec 2002 09:50:02 -0500
To: Miles Sabin <miles@milessabin.com>
Cc: www-tag@w3.org
Message-ID: <OF16E3EA8E.4E06086B-ON85256C94.00518E2B@lotus.com>

Miles Sabin writes:

> noah_mendelsohn@us.ibm.com wrote,
> > Would it make sense to augment the normative Namespaces rule on
> > character-by-character comparison with a health warning:  "NOTE: 
> > although namespace identity is determined on a character match basis,
> > users are strongly discouraged from intentionally creating distinct
> > namespaces named by URIs that would be considered equal per RFC 2396
> > or other governing specifications."  Or some such.
> 
> In what _practical_ way is this different from changing the Namespaces 
> rule from character-by-character comparison to RFC 2396/deployed 
> network infrastructure comparison?

Very big difference, I think.  My proposal is:
namespace-aware processors still compare
character-by-character.  For purposes of schema
(targetNamespace), deciding whether two attributes have
the same name on a given element, etc. it's still
character-by-character.  My proposal is merely that we
warn users away from creating namespace names that are
likely to conflict.  Changing namespaces to mandate RFC
2396 processing would require an incompatible change to
already-deployed code (and one that is not
necessarily practical to deploy when you consider that
comparison rules can depend on scheme names that a
given processor might know nothing about...presumably,
a user inventing a namespace name SHOULD know the rules
for any scheme used, no matter how obscure that scheme
might be.)  So, big difference I think.


> This is the core of my complaint here: the push for namespace documents 
> will inevitably lead to a de facto revision of the Namespaces REC, and 
> that's too big a change to introduce by the back door.

Well, I think the question is: do the revisions mandate
a change to the normative processing and conformance
rules for documents, or are they merely health
warnings.  A change to mandate RFC 2396 makes the following
document not namespace well-formed:

<e xmlns:p1="http://example.com" 
   xmlns:p2="HTTP://example.com" 
   p1:attr="1"
   p2:attr="2" />

I really, really, really don't want to go there.  I
think it would have been a mistake in the first place,
and I certainly don't want to do it now that it would
imply loss of backwards compatibility.  This decision
was considered to the tune of something like 3000
emails, and I suggest we not reopen it.

> Cheers,
> 
> 
> Miles

------------------------------------------------------------------
Noah Mendelsohn                              Voice: 1-617-693-4036
IBM Corporation                                Fax: 1-617-693-8676
One Rogers Street
Cambridge, MA 02142
------------------------------------------------------------------
Received on Thursday, 19 December 2002 09:53:16 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Thursday, 26 April 2012 12:47:14 GMT