Lists Home |
Date Index |
Micah Dubinko wrote:
> For general-purpose XML linking, I don't think anyone has a comfort level
> for what works and what doesn't, otherwise, this discussion wouldn't be
> necessary. :-)
We have some areas where we have some knowledge on what works
or what is needed.
- The universal (more or less literal) including of another
document. With a properly modularized core the issues on how
this interacts with validation, adding default values and
so on should be solvable.
We have this already as xinclude, xsl:include, xsd:include
and certainly in a host of other vocabularies. There's
certainly potential for reusing a common syntax and quite
a bit of common semantics here.
- The idea of "importing" another document, with a somewhat
vocabulary specific idea of "overriding declarations" or
something, as expressed in xsl:import, xsd:import or
Ant's import task. The potential of reuse is more limited
but the idea might be worth a look.
- The idea of a user interface link vocabulary, with possible
choices for the user on the target (let's say, a book title
links to a variety of online book shops), or where the user
agent might use a bunch of different URLs pointing to the
same content for establishing parallel downloads to overcome
server bandwidth limitations.
Instead of attacking the general problem, as interesting as it
seems to be, some more specialized work could lead to earlier
As for "general purpose XML linking":
- Aren't ID/IDREFS links?
- Does the XSD key construct establish links?
- Does xsl:key/key() denote a link?
- Does the XSLT document() function represent a link?
It would be interesting to have a reasonable formal definition
of in what a link differs from other relationships.