[
Lists Home |
Date Index |
Thread Index
]
> >There are substantial communities - remarkably unrepresented here on
> >xml-dev - who have been complaining vociferously because
> neither W3C XML
> >Schemas nor RelaxNG shows any sign of addressing the entity
> problem, and
> >they claim they really need them, mostly for this problem of naming
> >characters.
Trying to summarize the objections that people are raising, it is that
using a very similar mechanism at two different stages of processing
will work quite happily a lot of the time, but will be extremely
confusing in corner cases and when things go wrong, because it will be
even less clear to users what is actually going on. We already have
enough trouble explaining to users that entities are expanded at XML
parsing time and not an XSLT transformation time, explaining that they
might also be expanded at octet-stream decoding time is not going to be
much fun.
I think the most confusing aspect of it is the coupling to UTF-8. Users
who need this facility are most likely using text editors that work in
iso-8859-1. They want to be able to use this facility without disabling
the characters that actually appear on their keyboards, such as "£" (in
the UK) and "äöü" (in Germany).
This macro expansion should be something that happens after character
decoding and before XML parsing. Java got it right (in this respect)
with \uuuu.
Michael Kay
|
