> > This part would have to be done using
> > a choose-when expression, I think.
> 
> or applying templates over a large, if somewhat repetitive, xml file
> 

Or you generate a lookup document of the form

<c x="&#x1234;" n="1234"/>
<c x="&#x1235;" n="1235"/>
...

and then you can get the character code for a character using key().

Better still, you could generate the lookup document to contain the %HH
representation of the octets making up the UTF-8 representation of the
character.

If this lookup file is generated using Java or XSLT 2.0 (a one-off task),
does it still meet the requirement of being a "pure XSLT 1.0 solution"?

Michael Kay
http://www.saxonica.com/