[XeTeX] Problems with some "exotic" RTL scripts

[Jonathan Kew]

On 9 May 2007, at 1:28 pm, François Charette wrote:

> I have tested N'Ko with Code2000, and it turns out that contextual
> substitutions are not performed, even though they seem to be properly
> implemented in the font.

I'd think that it is impossible to judge "properly implemented"  
without a shaping engine specification.

> This occurs with XeTeX+ICU but also with the
> latest version of Pango on my Linux box. Does anyone on this list have
> any experience with the N'Ko script?

I'm not surprised by this; the ICU layout library has no shaping  
engine for N'Ko, as far as I know. Presumably Pango doesn't either.  
Nor do I see a specification for OpenType features for N'Ko on the  
Microsoft typography site; do you know if there is any information  
available? It would presumably be a fairly simple extension to the  
Arabic engine, but without a font spec, any implementation is going  
to be an arbitrary, hit-or-miss affair.

While I'd love to have complete support for everything :-) the  
reality is that N'Ko is not at the top of my personal priorities for  
now. But if someone cares to offer a patch, either for XeTeX or  
(better) directly to the ICU project, I'd be happy for it to be  
included in due course.

>
> The other problem concerns Old Italic (Etruscan): even though the  
> script
> is declared in the fonts Code2000 and MPH B2 Damase, it is not  
> processed
> as right-to-left scripts as it should (I have to type \char"202E in
> front of each word to have them displayed correctly). Any idea?

Old Italic is encoded in Unicode with left-to-right directionality;  
see The Unicode Standard 5.0, p 474. Right-to-left rendering will  
require directional overrides (and currently, you'll have to apply  
these to each word in XeTeX; this is a shortcoming of the  
implementation).

JK