Re: Changes in latin-ltx.el

[Eli Zaretskii]

On Tue, 11 Dec 2001, Karl Eichwalder wrote:

> > We already have a Hebrew input method, but it inserts hebrew-iso8859-8
> > characters, not mule-unicode-0100-24ff characters.
> 
> This is something I'll never understand: Why do you want to distinguish
> at _edit_ time whether your characters will be encoded as xxx or yyy?

Who said I _want_ to do that?  I surely don't want this dichotomy, but
it exists whether I want it or not, because the two kinds of Hebrew
characters are already part of Emacs.  A user who visits a UTF-8
encoded file with Hebrew characters will get mule-unicode-0100-24ff
characters, while visiting an iso8859-8 encoded file will decode them
into hebrew-iso8859-8 characters.

> When you save the file determine the encoding.

I believe that's currently impossible in general, unless and until we 
turn on Dave Love's changes by default.  Unless you run the CVS head 
version and have unify-8859-on-encode-mode turned on, Emacs cannot encode 
mule-unicode-0100-24ff into iso8859-8.

And even if you do turn on unify-on-encode, it doesn't solve the
problem of searching text that could have come from different sources.
For that, you need to unify on decode as well, thus effectively using 
only one out of several possible character sets for each given ISO8859 
character.