Re: Implementing Complex Unicode Scripts

Date: Tue Feb 20 2001 - 09:37:07 EST

On 02/20/2001 06:21:09 AM "Charlie Jolly" wrote:

>Should an open source script processing engine be part of the standard? As
>understand it if you want to develop Unicode solutions for complex scripts
>then you either have to do it yourself or rely upon Uniscribe or ATSUI.

Whether or not the Standard should include an open source rendering engine,
I can't say. I can comment, though, that there have been a lot of requests
(especially from Unix / Linux folk) for us to make our Graphite rendering
engine available as open source, and we are leaning toward doing that.

>Is the following a correct assumption?
>Unicode > Engine > Screen

Yes, if I know what you really mean. It's not completely clear.

>Do fonts have to tie themselves to a script engine.

Yes, but it's possible for a single TrueType font to simultaneously include
tables to make it work with either OpenType, ATSUI (it's AAT, actually) or
with Graphite.

>Will an Opentype font
>for lets say Hindi such as MS Mangal work on an Apple OS or Linux?

If developers write the code to make it happen, then it will. I can't speak
for whether developers intend to or not.

>Or is
>this font tied to Uniscribe?

Ah! Here's an interesting question. One shortcoming of OpenType is that it
assumes *some* software that lies between it and the Unicode character
representation with the result that at the interface to the OpenType tables
in a font you don't speak in Unicode but in a related but different lingo:
it's re-ordered, tagged Unicode. Potentially, different engines could use
different variants of that intermediate lingo. The tags used are a
registered set, and each is intended to have a particular meaning. It
remains to be seen if every implementation that's ever done will maintain
consistency with the others. (Uniscribe is not the only one - there is also
Adobe's CoolType engine, some work is happening on engines for Unix/Linux,
and I believe Apple may be working on something. Currently Uniscribe is the
only one available that handles Devanagari, so for the moment this font is
tied to Uniscribe.) So, hopefully this font will not be tied to Uniscribe
in the long run if other engines remain compatible with Uniscribe's
implementation. It's perhaps not likely that other implementations would
deviate, but I think it is possible.

>If this is correct then shouldn't there be a
>better solution?

The best solution is the one that supports every writing system and that
works everywhere. The next best is the one that comes closest to that. So
far, Uniscribe/OpenType is in the lead because it works with the largest
amount of software. The other OpenType implementations fall behind the
Unscribe one because they currently handle fewer scripts, work with a more
limited selection of software, or both. AAT and Graphite use
general-purpose processing engines that don't need to be updated in order
to extend them to handle additional scripts, which gives them an edge in
one regard over Uniscribe and other current OpenType implementations.
However, there is as yet hardly any software that makes use of either. So,
I'm forced to conclude that Uniscribe/OpenType is *currently* the best.
Whether or not it is in some academic sense the best is probably a matter
for debate and politics. It's not perfect, but none of the alternatives is

- Peter

Peter Constable

Non-Roman Script Initiative, SIL International
7500 W. Camp Wisdom Rd., Dallas, TX 75236, USA
Tel: +1 972 708 7485
E-mail: <>

This archive was generated by hypermail 2.1.2 : Tue Jul 10 2001 - 17:21:19 EDT