Re: Generic link handling

Chris Lilley writes:
>And as I said, a link would consist of a URI plus
>associated metadata.

That statement raises some interesting and difficult problems, many of
which are germane to both XPointer and XLink.  I have no problem with
the URI part - URIs are well-known to be useful identifiers, though they
have the comparison problems described in a concurrent thread.

There are some enormous problems, however, once you start dealing with
URI references rather than URIs.  XLink chose to stuff all the target
information into attribute values, so we're pretty well trapped into
using lightly-structured attribute-compatible forms to describe what it
is we're really connecting. 

If you consider fragment identifiers - and the MIME Media Type
information that makes them meaningful - as associated metadata, then
what you say above makes sense.  Unfortunately, since XLink crams that
particular metadata into the same value space as the URI itself,
XPointer has evolved into a very strange set of complex parts operating
in a very constrained space.

It may be time to seriously reconsider URI references and their proper
role in linking.  While I have few qualms about using URIs as part of a
Web-oriented linking structure, we need to ask some hard questions about
what part URI references should play.  From my perspective, they're
stunted abbreviations whose limitations are only recently becoming
clear, not a central part of any rational definition of a link.

See, for instance:

XHTML hasn't hit these issues as directly, mostly because the original
HTML fragment identifier syntax was conveniently simple while acceptably
limited for its context.  These are likely issues that XHTML will have
to think about in the future as well, especially if out-of-line linking
becomes any kind of priority for XHTML.  (Some days I think Tim Bray and
I are the only ones who find that notion fascinating, though we
certainly disagree about how to get there, but there are probably

>I don't see anything contentious there; I don't see anything radical
>or difficult or dangerous, and I don't see any mention of XLink or
>chess in there either.

In dealing with XPointer, I often feel that it's the result of some
basic mistakes in XLink, and that, much as Ann suggested, the game of
chess has been redefined so that the queen moves like the rook. 
Simon St.Laurent
Ring around the content, a pocket full of brackets
Errors, errors, all fall down! --

Received on Monday, 2 December 2002 17:08:16 UTC