Re: [emacs-bidi] mirroring of glyphs

[Eli Zaretskii]

> From: "Yair Friedman (Jerusalem)" <address@hidden>
> Date: Tue, 20 Nov 2001 14:56:19 +0200
> 
> Eli Zaretskii <address@hidden> writes:
> 
> > I'd expect LRO...PDF to also work, but I agree that LRM is a better 
> > choice.
> 
> I don't believe that humans have much use for all bidi codes except for
> LRM and RLM.

They are needed on rare occasions, when you have nested utterances,
like "i said `HE SAID ``she said'' ...' ..." type of phrase.  People
rarely talk that way, but it's possible in theory.

(Ironically, this highly improbable use is the reason that we need to
live with the 64-level stack which adds a lot of overhead to the
storage requirements.  Oh, well, what can you expect from an MS
brainchild?..)

> > I think the Hebrew input method should insert LRMs and RLMs 
> > automatically in cases like this one, if we can reasonably detect them.
> 
> Keyboard language change either by toggle-input-method command or by
> <language-change> event might be a good candidate for this, we need to
> wait for the next input character at that position to know if auto
> LRM/RLM is needed.

You cannot rely on language-change events, I think, since the input
method needs to DTRT even if you move the cursor into the middle of
text you didn't type, and then start typing.

I think the only way to work reasonably well in such cases is to look
at preceding and following characters, and implement some of Ehud's
wisdom to fix UAX#9 lossage.