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

Re: Use of fragment identifiers in XML

From: Dan Connolly <connolly@w3.org>
Date: 31 Oct 2002 17:07:55 -0600
To: Chris Lilley <chris@w3.org>
Cc: www-tag@w3.org, "Simon St.Laurent" <simonstl@simonstl.com>, Brian McBride <bwm@hplb.hpl.hp.com>, Dan Brickley <danbri@w3.org>, Aaron Swartz <me@aaronsw.com>
Message-Id: <1036105680.7448.1855.camel@dirk>

I hope this discharges ...

"Another issue: Use of frags in SVG v. in XML
Action DC: Describe this issue in more detail for the TAG."
http://www.w3.org/2002/08/26-tag-summary

It relates to an obligation I had to keep the RDF Core
WG informed about TAG proceedings relevant to
  http://www.w3.org/2000/03/rdf-tracking/#rdfms-fragments

I told the RDF Core WG that this was already on the TAG
issues list, but when I look more closely, I think I should
raise a new issue (or agree with Simon's suggestion
to raise an issue).

On Wed, 2002-10-30 at 14:46, Chris Lilley wrote:
> 
> On Tuesday, October 29, 2002, 10:42:11 PM, Simon wrote:
> 
> 
> SSL> In the TAG minutes for the 28 October teleconference, I find this
> SSL> exchange:
> SSL> ------------------------------------------
> SSL> 2.1.1 Use of fragment identifiers in XML
> 
> SSL>       * Action DC 2002/09/26: Describe this issue in more
> SSL>         detail for the TAG
> 
> SSL>     CL: The fragments refer to elements.
> 
> SSL>     DC: Unclear from text of SVG spec (indicates that
> SSL>     might be circle element or circle abstraction).
> 
> SSL>     CL: Please don't use SVG frags as an example of using
> SSL>     a URI + frag to refer to an abstraction (circle).
> SSL> ------------------------------------------
> 
> SSL> There isn't much context there, so I'm not entirely clear what the TAG
> SSL> is considering,
> 
> I assert that in this SVG document toy.svg
> 
> <svg xmlns="http://www.w3.org/2000/svg">
>  <circle r="2" id="foo"/>
> </svg>
> 
> that toy.svg#foo refers to a circle element.
> 
> Dan cites some wording in the SVG spec to the effect that no, this is
> an example of pointing to an actual circle on screen, or the
> mathematical definition of a circle, or the abstract notion of a
> circle, or something.

No, not that example. Consider this one:

  MyDrawing.svg#svgView(viewBox(0,200,1000,1000))

from the spec
http://www.w3.org/TR/2001/REC-SVG-20010904/linking.html#FragmentIdentifiersSVG

"This form of addressing specifies the desired view of the document
(e.g., the region of the document to view, the initial zoom level)
completely within the SVG fragment specification."


> I offered to clear up the loose wording in the spec as an erratum;

Do you consider the quoted paragraph above in error?

Or do you disagree with my interpretation of it, i.e. that
MyDrawing.svg#svgView(viewBox(0,200,1000,1000)) identifes
a view of the drawing, and not any particular XML element (nor other
syntactic structure) in the document.

> Dan
> was against that, presumably as fodder for the pointing-to-a-car line
> of argument.

It's related, but not the same issue; the pointing-to-a-car
issue, 15, is about what http://foo can point to, where foo doesn't
have a hash.

This issue is about what anyscheme:stuff#frag can point
to, especially when anyschema:stuff is an XML document,
i.e has an XML representation, or a representation in
some XML-derived media type.

>  Myself, I find the whole argument akin to counting the
> number of angels dancing on the head of a pin.

Well, yes, we postponed issue 15 on the grounds that
we couldn't find sufficient observable impact
on interoperability.

Maybe this one will go the same way.

But I'm sure I've seen test cases where this one matters.
At least: I was convinced at the time that it mattered.
I took the action from the 26Aug meeting on the expectation
that I would be able to remember the test case/scenario.
But I can't seem to bring it to mind. It was something
about using the same URI in XLink and RDF, in such
a way that the XLink seems to refer to an element
and the RDF seems to refer to what the element is
talking about.

Hmm... here's a particularly relevant part of the
history of this issue in my mind...

[[
But maybe this is a different sort of reference
than the one used in <a href="...">. Hmm...
]]
-- Yours truely to Masinter and company in rdf-logic, May 2001
http://lists.w3.org/Archives/Public/www-rdf-logic/2001May/0309.html

But re-reading that message, I'm pretty satisfied
with the position I took; I'm not sure I convinced
Larry, but I don't have my finger quite on the issue.

I recall, though only in a fuzz way, discussions
with Aaron Swartz where I was more convinced that
there's an open issue here.

I just can't seem to swap it all back in.

Meanwhile, back to the present thread...

I'm having trouble understanding this next paragraph...

> URIs point to Web resources. If people want to make derrived
> assertions, of the form "I am talking about the car described on this
> web page" then that is an arc,

_what_ is an arc?

> or an arcrole, or whatever - it is not
> a property of the UIR,

which URI?

> but an additional assertion that uses the URI
> as part of its arguments. Both XLink and RDF have the concept of arcs
> being separate from endpoints - I really don't see what the big deal
> is here.
> 
> 
> SSL> but this recent post of mine to uri@w3.org might be
> SSL> relevant:
> 
> SSL> http://lists.w3.org/Archives/Public/uri/2002Oct/0027.html
> 
> SSL> URI references and fragment identifiers seem to have grown more
> SSL> complex over the years, especially in XML contexts, and it seems
> SSL> worthwhile at this point to discuss how they work once again.
> 
> Pointing to a bare id does not seem over complex to me. Your general
> point is likely valid but the relevance to this example is not shown.

-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/
Received on Thursday, 31 October 2002 18:07:53 GMT

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