Re: Why not XHTML+RDF? was Re: Links are links

Norman Walsh wrote:
>...
> 
> So stand-off markup is a requirement in your view. No global attribute
> names, no namespaced element names, only something that puts all of
> the link descriptions "out of line".
> 
> Others have the view that links are sufficiently fundamental that they
> should be inline, recognizable in the surface syntax without a
> specific schema or other external description.

I am actually sympathetic to both views and think that individuals will 
have to choose. I am EVEN somewhat sympathetic to the view that it is 
best to decide between these two extremes rather than provide both 
options. I don't have a single right answer that is optimal along all axes.

But XLink is pretty near pessimal along all axes. If links are so 
important that they should be burned deep into the syntax then the 
appropriate namespace is "xml" and the specification should be VERY 
SHORT, VERY SIMPLE and VERY APPLICATION AGNOSTIC. Surely it isn't the 
case that links are "fundamental" to hypertext and somehow unimportant 
to database-style applications. They don't call it the "relational" 
model for nothing.

If XLink were a defacto extension of XML, and one that was simple enough 
to be fairly uncontraversial and universally understood, then people 
would be more willing to accept syntactic limitations as they do for XML 
(structured attributes anyone?). And of course it would be crucial to 
add linking support to schema languages, DOMs, XPath/XSLT, etc. etc.

On the other hand, if XLink is "just" a layer that will be

  * supported by some arbitrary subset of W3C-sponsored, XML vocabularies,
  * hardly supported at all throughout the family of specifications
  * and ignored by tool vendors,

Then please don't make it painful to use nobody will want to use it! It 
is all cost and no payoff. My XSLT's don't work better. My schemas don't 
do anything better. My DOM doesn't work better. My link checkers don't 
work better (because they ignore important links like object and 
stylesheet links). And my browser isn't even much more functional than 
if I used CSS to declare my link behaviours. All pain, no gain.

  Paul Prescod

Received on Wednesday, 2 October 2002 18:18:01 UTC