[
Lists Home |
Date Index |
Thread Index
]
At 12:14 23/01/2002 -0800, Bill Lindsey wrote:
At 1014 23/01/2002 -0500, Elliotte Rusty Harold wrote:
> >
> > <DocSet rddl:doctype="http://whatever/DocSet";>
> > <doc1 rddl:doctype="http://whatever/doctype1";>
> > :
> > </doc1>
> > <doc2 rddl:doctype="http://whatever/doctype2";>
> > :
> > </doc2>
> > </DocSet>
>
>What's wrong with:
>
><DocSet
> xmlns:xsi="http://www.w3.org/2001/XMLSchema-instance";
> xmlns:rxds="http://whatever/DocSet";
> xsi:type="rxds:doc">
> <doc1
> xmlns:rxdt1="http://whatever/doctype1";
> xsi:type="rxdt1:doc">
> :
> </doc1>
> <doc2
> xmlns:rxdt2="http://whatever/doctype2";
> xsi:type="rxdt2:doc">
> :
> </doc2>
></DocSet>
Actually I wrote that (and I suspect that Elliotte disagrees with it).
I was not trying to debate the exact form of the attributes which associate
a doctype with an element (sub-)tree, but making the point that a PI based
solution is unsatisfactory if you ever need to be able to package documents
together.
On the more specific question of the form of the attribute, I think it
would be best if it were added at a more fundamental level, e.g.
xmldt="http://whatever/some-doc-type";
The notion of a doctype (if it's useful at all, and I'm still open minded
on that) is independent of RDDL, just as the notion of a namespace is.
RDDL provides a good way of associating meta-data with a resource named by
a URL, whether that resource be a namespace or a doctype (or anything else,
for that matter).
--
Cheers,
John
|
