> My vote is definitely for (3), though (1) needs to be an
> option for compatibility.
>
> But rather than pass through the URN of the namespace in
> each call, I would pass an integer identifying it within the
> set of namespaces used in the document, with a separate
> function to map that to a URN.
>
> I guess people will also want to know the prefix used.
> Although the prefix is theoretically arbitrary, there are
> likely to be many conventional prefixes in use and
> applications may want to leave the prefix unchanged in an
> output document.
I strongly agree (3 with 1 for compatibility). In addition to the two niceties
Michael suggests, I would like the parser also to parse the namespace PIs. Not
only does this save all applications from writing the same code, parser writers
probably already have low-level functions they can adapt/use for this purpose.
The parser would use a separate callback (not processingInstruction) to return
the namespace information.
-- Ron Bourret