Re: xsl:number

Here be dragons.

I agree with you that the specification of numbering sequences is very
weak. In my view it's a classic case of "benign cultural imperialism" -
the spec authors wanted to make it fully international and localisable,
but since they were a bunch of Americans plus the odd expatriate
European, they didn't really have much idea in detail how to go about
it. This situation hasn't really changed in the 2.0 working group, and
the same problem has also made it difficult to agree a spec for
format-date().

Added to which, the implementors of the operating systems we have to use
don't even seem to be able to agree on how to implement "locale" at the
OS level.

As regards the specific questions, I think the result is that
implementors have a pretty free hand to do whatever they think is right.

When I first saw it, I assumed it meant "A-Z as specified in ASCII
or ISO 646RV", this being the most obvious sequence of characters
known to work on all machines, despite it being entirely the wrong
sequence for at least 30% of the work I do.

I think having xsl:number open to arbitrary reinterpretation on the
basis of the locale would be a serious error: there would need to be
some way to specify up front which sequence is to be used.

So far as I am aware, ordered lists which use alphabetic labels in
non-English Latin-alphabet languages never use diacriticals (but I'm
happy to be corrected). This is probably a cultural inheritance from
an older period when text was written in Latin, where there was a
relatively stable alphabet.