Bidi proposal, implementation

["Chahine M. Hamila" <mch at chaham dot com>]

After all the resistance to change I received, I came to the conclusion
that no proposal could work if it requires any change to what has been
done before.  This complicates our work a bit since we have to keep in
mind both the fact that the loss would be great not to be able to use
existing monodirectional programs, and at the same time we have to manage
with legacy unicode algorithm.

The idea has then been to introduce the change as an extension to the
existing unicode alg implementations. I will change the text of the
proposal accordingly.

I will just say a few words about it now. Limiting yourself to the
existing unicode algorithm means you have to write pure English text
(not one single word of Arabic) or use software with bidirectional
capabilities (impossible in some cases). If you want to write Arabic on
a non-bidirectional capable program (i.e. most existing software today),
you will *have to* use the extensions mentioned when reading/editing it
from a bidirectional capable program. By using them as described in the
proposal, you will actually be able to write any kind of text.

The implementation is based on the fribidi library.
Two functions have been added:
   - fribidi_enable_bidi_ext
 which enables the use of the extensions,
   - fribidi_assumed_orientation,
 which changes the assumed direction when the extensions are enabled.

By default, for an Arabic user, we strongly recommend that the bidi
extensions would be enabled and the assumed orientation would be right
to left. It should be possible for a user to enable disable these
parameters manually.

This will be the only way to make it possible to read correctly Arabic
stuff written with VI, emacs (etc...) which don't have bidirectional
capabilities. When doing this, Arabization could be summarized to
inverting the X coordinates on low level display drivers and translation
of documentation.

All the existing fribidi API has been preserved and it should be
possible to simple change the library with the patched one and continue
using the usual fribidi based soft without any modification (Pango for
example).

A new idiosyncracy (in a unicode encoding) will be introduced to force
the assumed display (in order to avoid me some research, people with
unicode knowledge have a suggestion on which code to use?).

Unfortunately, as you can notice with Akka, things are very slow since
I am very busy. If someone wants to continue developping the proposed
extension to the fribidi library, please go ahead, I'm all available for
directions (this would be a nice work for a newbie as the fribidi
library code is fairly easy to master).

Salam,
Chahine