[
Lists Home |
Date Index |
Thread Index
]
robin.berjon@expway.fr (Robin Berjon) writes:
>> I don't think that changes the argument about the value of generic
>> linking at all. To put it bluntly, "one linking mechanism to rule
>> them all" is ridiculous on its face.
>
>Blunt is fine, but this particular instance is tragically unhelpful.
>Why is it ridiculous on its face, does it have a big red shiny nose or
>something?
Ah, "tragically unhelpful". I take it that means that we're getting
close to the heart of the problem.
WHY "ONE LINKING MECHANISM TO RULE THEM ALL" IS RIDICULOUS
Broadly speaking, there are two ways to look at links.
The first way abstracts the notion of "link" from a wide variety of
practices. Any connection between two resources can be a link, whether
the connection is annotation, a sign of traversal, an informational
cross-reference, a key relationship in databases, or whatever else seems
to be convenient at the time.
The second way doesn't seek out "links" per se. Instead of a grand
unified theory of links, it encourages people to connect information as
they find convenient to the projects on which they work. Generalization
can and will occur as people share information which uses links, but
there's no notion that all of these things are necessarily the same. XML
is in many ways an opportunity for the second way, because the enormous
vocabulary flexibility it provides gives enormous room for
experimentation.
XLink took the first route. There's a general notion of a "link"
containing resource identifiers, paths between them, and metadata. The
role, arcrole, title (sort of), show, and actuate attributes let
developers customize their own links from a single standard notion. Role
and arcrole act as URI-based escape hatches (a favorite W3C approach)
which are basically "say what you really mean here".
In its early drafts, while XLink used this approach, it at least
preserved some degree of vocabulary flexibility. As namespaces came
into common use, XLink lost that flexibility as well, and now we have a
situation where XLink is not merely a common model but a fixed syntax,
and that URI escape hatch isn't enough to let people do the things they
want with links. The HTML IMG and the HTML 2.0 "href everywhere" cases
make that painfully obvious.
Much of the opposition to XLink has come from its top-down imposition of
a single way of looking at links. The many battles over show and
actuate and whether they should even be in the spec, the fights over
links as sets vs. links as traversal paths, and concerns about
integrating this with existing vocabulary and practice all look to me
like well-earned friction. XLink's thorough failure to make friends,
even within the W3C reinforces my suspicion that its approach was simply
bone-headed to start with.
"There's more than one way to do it" seems especially important in
linking, where even just the hypertext field of linking is marked by
incredible diversity and a wide range of conflicting opinions. I'm not
sure what gave the XLink working group the moxie to think that they
could develop the one true way for linking XML documents or the W3C TAG
the gall to expect the HTML WG to conform to that way, but I regard it
as perhaps the saddest chapter of XML history.
Perhaps some day we'll have enough experience to say that "this
abstraction of linking is good enough to cover all cases and be used
this way". Unfortunately, I don't think that specifying an abstraction
and a syntax now gets us anywhere near that. Even if you really think
there's a Platonic Form of linking out there, it's hard to claim that
assembling a committee to develop an XML vocabulary is going to get
anyone near there.
If you think linking is a simple enough field to be modeled this way,
I'm afraid I have to suggest that you haven't taken linking nearly
seriously enough. Even in the simplest cases, though, I have a really
hard time finding benefits to XLink that aren't outweighed by the costs
of using it.
Junk it.
--
Simon St.Laurent
Ring around the content, a pocket full of brackets
Errors, errors, all fall down!
http://simonstl.com -- http://monasticxml.org
|
