On Fri, May 02, 2008 at 10:05:05AM +0200, David Teller wrote:
> Actually, I remember Xavier Leroy saying that there was no problem, as
> soon as we were able to decide exactly what's a symbol, what's a letter,

Luckily Unicode defines such classes already.  There's a very high
level overview in this Wikipedia article:

  http://en.wikipedia.org/wiki/Mapping_of_Unicode_characters

We'd probably want to concentrate on a subset of the mathematical
operators first:

  http://en.wikipedia.org/wiki/Unicode_Mathematical_Operators

There's even support for this in OCaml already:

  http://camomile.sourceforge.net/dochtml-0.6/UCharInfo.html

Rich.

-- 
Richard Jones
Red Hat