I would like to point out that the debate on the LICR and related matters
has mainly only delt with what one might call the LaTeX Text Character
Model (LTCM), but there is another character model in current LaTeX which
should also be given some thought: the LaTeX Math Character Model (LMCM).
Possibly one could also distinguish a LaTeX Verbatim Character Model (LVCM)
(sorry about all these acronyms), but I'm less certain about that one.
Luckily matters may be easier in these models because there we don't have
do deal with that multilingual complex of problems which noone completely
understands because noone knows all the languages.
Concerning the LMCM, I believe the expressed opinion was that greek and
cyrillic letters (as input characters) should be allowed in math, but that
symbols outside ASCII should not (except when necessary for compability
reasons). I suspect user demands may make the latter problematic if the
input encoding becomes Unicode (in some form), especially if they get the
math characters well sorted out, but that is a distant problem. In the
world of 8-bit encodings a restriction of input symbols in math to ASCII is
probably the right things to do.
Allowing greek letters does however raise some interesting problems. Many
of the greek letters have var-forms in the current math fonts, so which
form should the input letter select? E.g. \epsilon and \varepsilon are
hardly distinct enough to count as different letters/symbols, they are
merely different glyphs, so which one should it be? I for one much prefer
\varepsilon, so I would like to have some interface which lets the user
In a more general view, one should perhaps try to clear up the LMCM so that
the user commands select characters (or character plus math class) rather
than glyphs. This could make it easier to provide new math fonts in that
one wouldn't have to concentrate on providing precisely the same set of
glyphs as the CM math fonts do, but could provide more (very tricky these
days, as new glyph forms require new commands that make documents which use
them incompatible with other math fonts) or fewer (possible by duplicating
the glyphs) forms of the characters as it suits the design.