W3C home > Mailing lists > Public > www-xml-schema-comments@w3.org > January to March 2012

[Bug 1974] Our published names for datatypes etc. don't resolve

From: <bugzilla@jessica.w3.org>
Date: Wed, 15 Feb 2012 22:17:51 +0000
To: www-xml-schema-comments@w3.org
Message-Id: <E1RxnAd-0007Bx-S6@jessica.w3.org>

--- Comment #11 from Henry S. Thompson <ht@inf.ed.ac.uk> 2012-02-15 22:17:47 UTC ---
> If a statement from the owner of a namespace NS that a particular name N in
> that namespace names a particular thing T does not suffice to make it the case
> that name N in namespace NS names thing T, then what kind of action can
> make it the case?

In the absence of anything else, yes, it probably suffices---I don't think I
said anything to the contrary.

The particular problem I failed to take account of does not arise from anything
we might write in English (or any other natural language) in the namespace
document.  It arises from the proposal I made to add anchors (that is,
id="...") for each datatype to (the (X)HTML version of) that document.  Why is
that a problem?  Because the (X)HTML media type registrations say that in such
a case, the corresponding URI _identifies an element_.  And that contradicts
what we say in the spec, which is that it identifies a datatype.

> If  the plan described in comment 6 is no good because HTML elements are not
> datatypes, then how is any namespace owner to document the namespace?  The only
> URIs whose meaning can conveniently be explained by means of a human-readable
> HTML element would then appear to be URIs denoting elements in HTML documents
> -- an interesting but not exhaustive subclass of URIs.

There are two ways to cut the gordian knot:
 1) Don't say that NS#name identifies a datatype, say it identifies definitions
    of datatypes.  Then my proposal would have been good on _two_ counts, back
    when I made it:
     a) It would have provided anchors in the (X)HTML which identified things
        which _did_ define the datatypes, albeit briefly;
     b) It would have made the content negotiated alternatives
        in that the 2001-vintage XMLSchema.xsd has anchors already, and they
        identify definitions too.
 2) Make the NS#name actually identify a datatype, using the only technology we
    have available to take URIs out of the web and into the world, namely RDF.

(1) is almost certainly closed to us: if we tried to change the relevant bit
of Datatypes now, the whole RDF community would push back, since they have been
taking us at our word, and using the advertised URIs to indentify datatypes
since 2001.  Note that (1b) would also no longer be true, because the 1.1
schema document for schema documents no longer contains the relevant anchors.

I'll shortly make a proposal which would implement (2), while still also
providing some comfort for human beings as well.

Regarding your final question, about the alleged incompatibility between
descriptions and representations:  that is at the heart of an ongoing and
extremely contentious debate.  It's known as the httpRange-14 problem, and the
TAG is currently gearing up for a major effort to find a position which will
attract consensus on both the theoretical and practical fronts.  I don't hold
out much hope for a resolution here in the short term, hence my interest in
finding a solution which doesn't fall foul of any of the known challenges in
this space.  Please bear with me, until I can actually frame my new proposal.

I'm sorry to be doing this in such an extended fashion, both in time and in
volume of prose, but we are for better or worse in the position of those
pioneers identifiable as such by the arrows in our chests, and I'm trying to be
sure I get every one of them out.

Configure bugmail: https://www.w3.org/Bugs/Public/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the QA contact for the bug.
Received on Wednesday, 15 February 2012 22:17:56 UTC

This archive was generated by hypermail 2.3.1 : Wednesday, 7 January 2015 14:50:12 UTC