Re: Text widget comments

[Roozbeh Pournader <roozbeh sharif edu>]

On Mon, 12 Mar 2001 dov imagic weizmann ac il wrote:

> * How can I make xkb input a special character? E.g. I would like 
>   Shift-A in Hebrew mode input LRM (Ux200E)? 

Use 0x100200e as the key code.

> * I created a file that contained a RLM character, which worked 
>   without any problem and modified the BiDi output as it should. But
>   I then tried to use the mouse and keyboard to mark this zero width
>   character in order to copy it and paste it somewhere else. I never
>   managed... I managed in the end to position the cursor in order
>   to erase it through hte keyboard, but it took me a while to figure
>   out where I was in logical and visual spaces... How much fun it
>   will be if someone actually uses the recursive 63 levels of BiDi
>   override, that the Unicode specification allows. 8-)

I believe that this weird (il)logical cursor and mouse behaviour should
only be considered a temporary solution. Add explicit bidi to that, and it
will explode. I don't know how The Right Thing should be done yet, but I
believe that a bidi widget can be much easier for the user. She should not 
tackle with LRM and RLM insertion, or things like that. I believe that
such expertise, should be moved into software. I wonder when I see the
bidi widgets in MS Windows 2000 having almost the same behaviour as
MS Windows 3.1 versions. MS Word 2000 does a better job at that, but it's
only a little better.

We had deep discussions regarding this in the emacs-bidi mailing list, and
everyone's invited to take a look at the archives, but most importantly
because no one had the time to convert the Unicode bidi algorithm to
something automaton-like, put aside generalizing it, the discussuisons did
not converge. 

[Please note that to my best knowledge, even a single conformant bidi
engine does not exist yet, the best candidate may be IBM'S ICU, but that's 
because they have written the bidi spec themselves... The implementation
guidelines at the end of the Unicode bidi technical report will also make
the engine non-conformant, something that the spec writers themselves
didn't know, because they didn't use their own recommendations when
implementing the reference code.]

I believe that some real time is needed for this: a good time playing with
tables and specs, a better time implementing different ideas, and
specially, QA by some dozens of bidi typists. The team working on this
should have the ability and courage to throw away the intermediate code,
and restart from scratch many times. Add to this, the time needed for
writing an implementable spec based on the research.

> * A final comment is regarding interaction between the widgets and 
>   the keyboard direction. Consider a clerk sitting somewhere in the
>   world having to fill in the following form:
> 
>    Name:                     ________
>    Name in Latin characters: ________
> 
>   where the native language is a non-latin character language. The
>   clerk is tabbing between the input fields and is entering the
>   requested information. As the clerk is in a hurry it would be very
>   nice if the tabbing to the Latin character entry would
>   automatically change the keyboard to a Latin character keyboard
>   setup that was specified by the programmer.
> 
>   As far as I understand this interaction is not possible right now.
>   What would it require to make it work? Is it desirable?

It will be needed. What do you mean by desirable? By whom? The user will
have a better time with this feature.

--roozbeh