From mboxrd@z Thu Jan  1 00:00:00 1970
From: David Tolpin <dvd@davidashen.net>
Message-Id: <200403092039.i29KdQDk094407@adat.davidashen.net>
To: 9fans@cse.psu.edu
Subject: Re: [9fans] bidi
In-Reply-To: <101e1d3477ddc2d90dba62df1f391f86@vitanuova.com>
Date: Wed, 10 Mar 2004 00:39:26 +0400
Topicbox-Message-UUID: 2781c0ac-eacd-11e9-9e20-41e7f4b1d025

> > You get discontinuous selection in the underlying file. If it
> > does not work with the current model, then it is a limitation
> > of the current model.
>
> it is indeed a limitation of the current model.  for instance sam and
> acme's command language assumes that dot, the current selection, can
> be represented by two numbers, and changing that would change many
> fundamental assumptions in the code.
>
> acme(4) exposes its selections externally too; how would that
> work?

1. That means that the code should be changed.

2. Untill it is changed, there are choices, and the choices are
well explored:
- limit selection to spans with one direction.
- limit selection to spans with starting and ending points with
  the same direction

> i wouldn't say it'd be impossible, but it might be easier to code the
> whole thing from scratch. would it be worth it 

Untill bidi is implemented, it is not very usable for roughly a half
of the Earth's population.

> (against, say,
> having a heuristic to determine directionality for the whole
> of a window)?

1) It is not a heuristic, it is a setting. One does have now
this setting, it has one value - ltr. It should be possible to
add the other one, rtl.

2) It is not an issue of bidi. The basic bidi can be implemented
without it.

> > What is it about?
>
> it?
>

Use of glyphs  for markup. Bidi is independent of use of direction
marks. Bidi can be implemented without direction-override marks. And
still be very usable. 'embed' as the only mode is almost the answer.

It lets people write and read in their languages.

?tfel ot thgir morf hsilgnE daer ot deirt uoy evaH

David