[XeTeX] On combining diacritics again

[Adam Twardoch]

Will Robertson wrote:
> Just getting a bit off-topic here -- I can understand that for 
> backwards compatibility this is totally the way to go here, but are 
> new fonts being implemented with AAT tables at all now? 
Mostly because the OpenType Layout support in Mac OS X 10.4 is still 
very incomplete. Only very basic "Western" layout features (ligatures, 
small caps, old-style numbers etc.) are supported. Contextual lookups 
are not supported, and more importantly, there is no language-sensitive 
or script-sensitive support. This means that, for example, OpenType 
Layout features that take care of Arabic of Devanagari shaping won’t 
work at all.

So for those scripts, one still needs to develop AAT tables. Developing 
these is pain in the neck because there are only buggy command-line 
tools from Apple available with *very* scarce documentation.

At ATypI in Helsinki, Apple announced that they’re actively working on 
adding support for all OpenType lookup types and all script-sensitive 
processing to Mac OS X. This means that at some point (perhaps in Mac OS 
X 10.5, perhaps later), we’ll see more support for OpenType Layout 
tables and less need for creating AAT tables.

The nice property of the OpenType format is that you can have both types 
of layout tables in one font (OTL and AAT). Such fonts are rather tricky 
to build but it can be done. With such a font, you can get e.g. Arabic 
shaping in Cocoa applications using AAT and in Mellel or InDesign ME 
using OTL, all from the same font.

Jonathan was on the SIL team that built the free SIL Arabic fonts 
(http://scripts.sil.org/ArabicFonts ). They have chosen to build two 
separate fonts for each design, one in OTL and one in AAT. This had 
surely some practical reasons, one of them being that the AAT font 
needed more precomposed glyphs due to the lack of mark attachment 
support in AAT. But I guess one could theoretically take the glyphset 
and the features of the AAT font and basically just add the OTL features 
of the other one, and it would principally work (the redundant 
precomposed mark-to-base glyphs would not be used by the OTL code that 
would use dynamic positioning instead).

BTW, kudos to Jonathan and the rest of the SIL team for the great work 
on those fonts!

Adam


-- 

Adam Twardoch
http://www.twardoch.com/