Re: some uses of relative URI as namespace names

Quick points re the Xalan examples cited:

Note that Extension support is very much _experiemental_ at this time. It
is hoped that the XSLT WG will eventually specify how stylesheets create
and bind to extensions. In the meantime, anything done in this area should
be considered experimental and subject to change... so if we decide that
this solution has to be reworked to track the evolution of Namespaces, I
don't think anyone will be hugely bent out of shape.

Note that this solution involves directly dereferencing the Namespace name
-- sometimes as a URI, sometimes not! -- but that equivalent behavior can
certainly be achieved in other was.

       An arbitrary (but unique) string that matches the prefix
       attribute of an lxslt:component element in the stylesheet.
       Example: xmlns:ext1="xyz"

This is just the "Literal" interpretation straight from Namespaces 1.0. I
don't think there's any real dependency on relative URIs here; "xyz" is
just a sloppy placeholder, and could just as easily have been absolute.

       [class:]FQCN where FQCN is a Java fully qualified class
       name. If the extension only involves static class method
       calls (no instance constructors or instance method calls)
       precede the class name with class:.  Example:
       xmlns:ext2="java.util.Hashtable"

Note that this is a special case, _NOT_ a relative reference, since it's
hardly likely that the stylesheet's base URI starts with "class:"
pseudo-scheme. Making the implied "class:" work requires being able to
retrieve the literal string. Of course, having to explicitly code "class:"
would not be a disaster.

       The file name or URL for another document that contains the
       lxslt:component element.  Example:
       xmlns:ext3="my-component.txt"

This too is a special case; it's doing browser-like guesswork about whether
what the user has typed is a local file or a relative URI reference, which
is something that URI References themselves do not support.

Again, making that guess requires that we be able to retrieve the literal
string.

______________________________________
Joe Kesselman  / IBM Research

Received on Thursday, 15 June 2000 11:39:59 UTC