Re: BIDI and IBM

From: Yung-Fong Tang (
Date: Sat Dec 11 1999 - 03:35:26 EST

I have look many Arabic fonts from NT. Most of them do have most of the
Presentation Form-B but not Presentation Form A. And although Unicode do have all
different shape, it is strongly discourage user use them in data source. They are
there for compatability w/ some legecy charset- such as DOS Arabic. It think it is
safe to convert the Basic Arabic characters into presentation form B and then as
the os to display them. Actaully, if you read my shaping code in mozilla. My code
do take account some Presentation form B glyph may missing. In that case, it will
fallback to the basic glyph.

Michael Kaply wrote:

> You are correct, the shaping code involves mapping the glyphs into the proper
> Unicode representations (initial, medial, final) and then they are displayed.
> My understanding (I am not the expert) is that the Unicode Arabic set contains
> all representations of the Arabic characters (standalone, initial, medial,
> final), whereas some platform codepages only contain the standalone form and
> the operating system renders the entire word properly based on the position of
> the characters.

I don't think the OS support for Arabic shaping is available in the following OS
Win95 US
Win95 FR
Win95 C/J/K
Win98 US
Win98 FR
Win95 C/J/K
WinNT4 C/J/K
I think it only available on Arabic/Hebrew version of Win95/98/NT4
I have no idea about Window 2000

So... why don't we do Arabic shaping by ourself ? My code is sort of working
there. The rest is need bi-di expert to QA and fix bugs....

> So any machine that supports Unicode and actually has the Arabic characters in
> its Unicode font should work.
> How many of the Mozilla platforms actually map the Unicode character codes to
> platform fonts rather than this displaying in a Unicode font? I assume this is
> done because most of the platform fonts do not have Unicode glyph codes in
> their default fonts?

Mac will do the right thing.
As I said, I do have Logical Arabic Unicode to LangBox -8 font conversion code in
my private build.

> Mike Kaply
> Erik van der Poel wrote:
> > Michael Kaply wrote:
> > >
> > > The work we will be doing will be cross platform. Obviously the platform
> > > will have to have support for the proper codepages, but things like
> > > shaping for Arabic, etc. will be done in the cross platform code, not
> > > relying on the operating system.
> >
> > Arabic shaping involves mapping Arabic characters to initial, medial and
> > final glyphs, right? Are those glyph codes in Unicode?
> >
> > Mozilla's cross-platform code deals with Unicode. We map Unicode
> > character codes to font glyph codes in the platform-specific code in
> > gfx. If those glyph codes are the same for Windows and OS/2 fonts, you
> > may be able to share that code (or just copy and paste it), but I wonder
> > whether the glyph codes would be the same on Linux.
> >
> > Erik

This archive was generated by hypermail 2.1.2 : Tue Jul 10 2001 - 17:20:56 EDT