Re: [emacs-bidi] improve visual-to-logical

[Alex Schroeder]

"Eli Zaretskii" <address@hidden> writes:

> There are two types of Hebrew characters in Emacs 21: those which
> belong to the hebrew-iso8859-8 charset, and those which belong to the
> mule-unicode-0100-24ff charset.  My point is that you should use with
> each one of these the LRM and RLM that belong to the same charset.

I understand.  But the nature of Ehud's idea applies to any script --
Arabic, for example.  Thus, the question is not wether to choose LRM
from hebrew-iso8859-8 or from mule-unicode-0100-24ff, but any of the
potentially numerous LRM's around (perhaps only two, who knows...).

Anyway, I think that I should use the most general one (Unicode) and
rely on unification (Dave Love's effort).

Perhaps we should define the problem again:  Ehud's idea requires the
insertion of LRM characters.  When saving such a file, it would be
very annoying if Emacs complained about unsave coding systems just
because a LRM character from another charset was used.

There are two solutions to this: Pick a "safe" LRM character, or rely
on unification.

I prefer the second approach because I think unification will happen
anyway -- especially since Hebrew and Arabic are in 8859 so the work
might already be done for those charsets.  Additionally, I don't know
how to pick a "safe" LRM character.  This would involve the same
machinery used when deciding wether the current coding system is safe
when writing the file.  I don't think it should be necessary to invoke
that machinery just to decided which LRM to use.

Alex.
-- 
http://www.emacswiki.org/