W3C home > Mailing lists > Public > www-tag@w3.org > January 2009

Re: Near final draft of TAG finding on the Self-Describing Web (Self-describing XML Documents)

From: Dan Connolly <connolly@w3.org>
Date: Mon, 19 Jan 2009 09:15:05 -0600
To: Mark Baker <mark@coactus.com>
Cc: noah_mendelsohn@us.ibm.com, www-tag@w3.org
Message-Id: <1232378105.30129.1308.camel@pav.lan>

On Sat, 2009-01-17 at 11:31 -0500, Mark Baker wrote:
> On Fri, Jan 16, 2009 at 4:04 PM,  <noah_mendelsohn@us.ibm.com> wrote:
> > Mark Baker wrote:
> >
> >> My concern is with the mention of the application/xml media type
> >> there, which I feel inappropriate because RFC 3023 doesn't license,
> >> for example, that an XHTML document delivered as application/xml is
> >> intended to evoke XHTML semantics.
> >
> > Mark, thank you as always for your comments.  We have discussed this
> > extensively at several face to face meetings, and I believe we came out
> > comfortable that RFC 3023 [1] does support inferrence of XHTML semantics,
> > albeit indirectly.  For example, it says:
> 
> Oh, I didn't realize you'd had this discussion before.

I suppose we could/should have solicited a bit more review at the
time...

`cvs annotate` says this bit comes from 1.8 May 12 2008...

Shane replied at the time and commented on the reference
to the XML media type spec, though it was only editorial.
http://lists.w3.org/Archives/Public/www-tag/2008May/0070.html

Steven Pemberton also replied
http://lists.w3.org/Archives/Public/www-tag/2008May/0156.html

I suppose neither of those directly represents the work
on RFC3023.


I asked the question as point-blank-ly as I could
manage in Nov 2006 ...

follow your nose from XML documents to namespace documents?
xmlFunctions-34, nsMediaType-3, RDFinXHTML-35
http://lists.w3.org/Archives/Public/www-tag/2006Nov/0086.html

... but I didn't manage to solicit a response; I suppose
I could have written to the xml-types list.


> > "An XML document labeled as text/xml or application/xml might contain
> > namespace declarations, stylesheet-linking processing instructions (PIs),
> > schema information, or other declarations that might be used to suggest
> > how the document is to be processed.  For example, a document might have
> > the XHTML namespace and a reference to a CSS stylesheet.  Such a document
> > might be handled by applications that would use this information to
> > dispatch the document for appropriate processing."
> 
> Right.  That's a paragraph that Larry and I worked on together in
> response to my raising exactly this issue.  It wasn't as clear as I
> would have liked, but it's all we could agree upon.  Its intent, from
> my POV, was to serve as a warning - through the use of the word
> "might" - to those who might assume that publishing an XML document as
> application/xml would necessarily be interpreted the same way as if
> the format-specific media type were used.

OK, well, the intent fades over time, and the text remains, right?
Perhaps the text admits multiple readings, but one of them
is that consumers MAY dispatch on namespace.

Given that, as you point out below, the cost of updating 3023 
seems to exceed available resources, I'm content that the current text
suffices for the purposes of the self-describing web finding...



> > I think we convinced ourselves that, particularly in the case where a
> > namespace document is available at the namespace URI, 3023 does allow one
> > to infer that the semantics can be inferred from the qualified names (or
> > more specifically, from the corresponding expanded names) of the elements
> > used in the document.  This was quite crucial to resolving a difficult
> > discussion, so I have some reluctance to reopen it this late in the
> > publication cycle unless there's pretty clear evidence that we've
> > misunderstood 3023.

I don't think it's a question of whether the TAG misunderstood 3023;
we were aware this was a bit of a stretch; the question is: does
the community agree that 3023 is sufficient justification?

I gather we have some push-back.

>   It certainly seems desirable to me that user agents
> > receiving a document of media-type application/xml be allowed, in the
> > cases where they recognize the qualified names of the root and other
> > elements, to apply the semantics of those names to the document.  If there
> > is any ambiguity, I'm tempted to suggest that it be resolved by clarifying
> > 3023, rather than by changing the SDW finding.
> 
> Work was started on 3023bis a few years ago, but there wasn't enough
> interest in it.
> 
> But there is ambiguity in practice with such an interpretation of
> 3023, in that XML language designers aren't required to ensure that
> the root namespace is their own, and indeed some haven't.  Two
> examples;
> 
> http://www.w3.org/TR/xslt#result-element-stylesheet
> http://www.w3.org/TR/rdf-syntax-grammar/#start

In the RDF case, I'd expect to be able to follow my nose
thru the namespace document for the root element to the
RDF spec, eventually. But yes, XSLT literal result elements
are a strange beast.
(as was noted as early as Feb 200
http://lists.w3.org/Archives/Public/www-tag/2002Feb/0178.html )

We mostly talked about those under XMLFunctions, so I had
left that aside in my thinking, but perhaps it's not
technically orthogonal.
 http://www.w3.org/2001/tag/group/track/issues/34




> And then there's the surprisingly large number of languages which
> don't even use namespaces.

Since using namespaces is more work than not using them, I'm
not at all surprised that there are many cases where people
don't use them.

[...]
-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/
gpg D3C2 887B 0F92 6005 C541  0875 0F91 96DE 6E52 C29E
Received on Monday, 19 January 2009 15:16:13 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Thursday, 26 April 2012 12:48:11 GMT