On Thu, 15 Mar 2001, Michael (michka) Kaplan wrote:
> There is no one single, intuitive algorithm that could
> be used, which everyone would like.
I highly agree. This input method, or what you may call it, should be
locale-dependent. European numerals, which are just normal numerals in a
Hebrew context, are considered foreign numerals in a Persian context, and
should have different behaviour.
But there's also no one single collation algorithm. Or let's say there's
one, but it uses some locale-dependent information to work. This
bidirectional editing algorithm thing can use the same concept, so it can
get standardized and remain customizable.
I would also like to tell that it should also be application dependent.
Just try to write a bidi HTML page with FrontPage 2000, and you'll
understand what I say. All the "<"s and ">"s get this way and that way,
and the markup gets unreadable. A bidi-aware HTML editor should consider
HTML tags somehow atomicly, let's say like an embeddeding.
> With that said, I am sure that there are improvements that could be made
> here. But they would have to be made by people who do mind the complaining
> that will follow from all the people who have other ideas on what makes
> sense.
Yes. And we have tried it. You should wet your hands at the "emacs-bidi"
mailing list if you are interested. I can't remember where the archives
are, but a good search, let's say with google, should reveal that.
Many details got discussed there. But simply because we did not have the
resources to test many different implementations that would next go to the
trashcan, we did not come to a good conclusion. Also, to understand the
mechanism of the UAX #9, we tried to convert the algorithm to a more
state-machine thing, but again it was harder than what we expected. One of
my colleagues is working on this currently, but I'm not sure when he
expects results.
I think a good research team should be created to work on this, with
access to good QA. This is really needed for bidi users.
--roozbeh
This archive was generated by hypermail 2.1.2 : Tue Jul 10 2001 - 17:21:20 EDT