Re: BIDI: possible fix

From: Roozbeh Pournader (roozbeh@sharif.edu)
Date: Sun May 20 2001 - 09:58:38 EDT


On Sun, 20 May 2001 matial@il.ibm.com wrote:

> Underlining or overcircling a separator within a number hardly qualifies as
> "plain text".

Well, it's plain text. My definition of good plain text: a stream of
undeprecated Unicode characters each treated exactly as Unicode says.
Unicode encodes those combining characters, doesn't it?

And I had not mentioned "plain text" even once. I also need the change for
HTML text, for my own fine homebred markup language, and for everything
else that respects Unicode bidi.

> If the user is in the business of adding special marks in
> the text, he/she might as well put a pair of LRMs around the number, and
> the desired result will be achieved.

Well, do you know any other standard way for overcircling something, (or
overlining it), other than putting a U+20DD COMBINING ENCLOSING CIRCLE
after it? If you don't like my circle, replace it with anything you want
from the 20D0..20FF or the 0300..036F block. If you ask me, I prefer that
U+0332 COMBINING LOW LINE thing, it will be good for everything from me
emphasizing it to others, to underlining effect in future Unix UTF-8 text
terminals.

> I still don't see enough justification for changing the algorithm
> (although I would not have minded if it had always conformed to
> Roozbeh's proposal ).

Ok, I agree that changing the algorithm may hurt. I believe that you are
also trying to be conformant to the algorithm with minimum effort, but I
think other conformant implementors should also speak on this. We are
still far from fully conformant implementations, with only few existing
ones except the references.

And let's see, you could not see that one may need a non-spacing character
over a separator. You a bidi guru, implies that there will be little (if
any) existing data that wants the NSMs change the separator underneath.
There's not much existing data.

Finally, a question from everybody: What are enough justifications for
"any" change to the bidirectional algorithm? The current algorithm is
working against its own sense, and the change is so small that will escape
from anyone not serious enough. Is "But I do not want to change my
implementation" a good reason for remaining with something broken? I think
we need comments from the people inside, the authors, like Mark Davis or
Asmus Freytag.

--roozbeh



This archive was generated by hypermail 2.1.2 : Fri Jul 06 2001 - 00:18:17 EDT