On Mon, 4 Dec 2000, Michael (michka) Kaplan wrote:
> http://www.microsoft.com/globaldev/drintl/001/br_sniff.asp?fa
>
> in x86 NN 6.0 (and Mozilla for all milestones 13-18), it will format the
> currency value or 3943.23 as 3,943/23. This is identical to the behavior on
> Windows 2000 and Arabic Win98.
First of all, I should clarify that no Persian reader puts any difference
between the currency separator and decimal separator. If anyone has
evidence against this, I'll be more than interested to hear.
That page gets many things wrong in my IE 5.5 on Windows 2000. Specially,
"NO" Persian reader accepts / as a currency or decimal separator between
European digit shapes. That more than unacceptable. If anyone has seen any
use of it in print, please show me and I will stop bugging ;)
With current European digit shapes of that page, it should render the line
as: "3,943.23".
I don't comment about Mozilla or Netscape, because they do not claim bidi
yet. The bidi you see is there only because of Windows native support.
> On IE 5.0 and 5.5, it will format the same number as 23/3,943.
Which is more than broken.
> Is this a bug? It seems uncertain and would depend on your definiton of
> "bug." The only thing we know for sure is that anyone who has a particular
> opinion can unintentionally take advantage of the diversity of opinions to
> prove that they were right -- no evil motive required.
This is a bug, but the question remains: How can I prove this? I can even
try this to be approved by a national standard approved by ISIRI (Iranian
national standard body). I'm not talking about my opponion, I'm talking
about common practice and standards.
BTW, I have scanned parts from the standards:
http://sina.sharif.ac.ir/~roozbeh/farsiweb/mom/3342.gif
The relevant table from ISIRI 3342, the Iranian character set
standard. Red bullets mark the lines that define Momayyez and Slash.
http://sina.sharif.ac.ir/~roozbeh/farsiweb/mom/style1.gif
Is from the book "Shivenaame" (The Book of Style), Second Edition, Iran
University Press, 1993, ISBN 964-01-8127-7, the section "Style rules for
Mathematics and Physics texts". Look at the red bullet for the line
defining the Momayyez, and blue ones for usage of Slash and Momayyez.
http://sina.sharif.ac.ir/~roozbeh/farsiweb/mom/style2.gif
Some other examples for use of Slash from the same source.
> This is the danger of trying to rely on "de facto" national standards (i.e.
> depending on common usage), since according to common usage, everybody is
> right (and everybody is wrong). Exactly how many people have to be using a
> particular convention before one changes the opinion from "they are wrong"
> to "the other convention is wrong or incomplete", exactly?
I am talking about "real" standards. See above.
> I think the Mozilla/Win2000 behavior is better since it properly handles the
> people who *do* use the slash character as a currency separator, and I do
> not see any cases that are broken for its usage either (note the date format
> usage and how it is not negatively impacted for any Arabic script locale).
> With this change, AFAICS, both camps are well treated and neither is
> punished for their preferences.
I think bidi algorithm has nothing to do with what the user wants. The
application should use the overrides, etc to get what she wants. She
should then be able to give her file to someone else and he should be able
to see it exactly like she wanted. The bidi class of the slash character
is not that much important if it gets rendered the same way across all
platforms.
> Unfortunately, in the Unicode Bidi algorithm case and the IE 5.0/5.5 case,
> at least one group is punished.
>
> The change would be to treat the Slash character as a Common Separator
> instead of a European Separator. This change is (I think) at least worth
> consideration for the Unicode Bidi algorithm.
I do not comment on The Standard's Bidi algorithm, since it is different
from a Persian user's expected behaviour in many ways, and it can't get
everything right, since different Arabic-script users expect different
things. And the Unicode algorithm does not claim that either. But I think
applications should simply conform.
We (the Arabic writers) have had a really bad nightmare with different
renderings of the same text during all these years of computing. Would
everyone that claims Arabic support please fix their application? Why are
we here?
Sorry for the flames,
Roozbeh.
This archive was generated by hypermail 2.1.2 : Tue Jul 10 2001 - 17:21:17 EDT
