From: Antoine Leca (Antoine10646@leca-marti.org)
Date: Wed Jun 15 2005 - 05:55:45 CDT
On Wednesday, June 15th, 2005 09:12Z Dominikus Scherkl wrote:
> But the discussion reveals, that the markup-topic "colorize ligatures"
> seems to become easier to solve at a rendering-engine level,
OTOH, this very need for attributes to be associated with characters become
a problem for certain class of engines. For example, Freetype intend to
deliver a simple engine to be use to render text. It can be used to render
glyphs (and is used this way by e.g. Pango, often quoted here), but it also
"targets" more simple applications, which feed it with characters, and
expect to receive bitmaps in exchange.
We are not presently able to deliver something that handles complex scripts;
we tried something for Freetype 1.x for Arabic OpenType, but it becomes a
dead end, and it certainly cannot be scaled (easily) to handle Indic
OpenType. For Freetype 2 we had (and are still having) a deep reflexion
about how to shape the API to allow the engine do the shaping of complex
scripts, while at the same time allow the applications to provide attributes
attached to the characters, since it appears to be a common demand overall.
The result is pretty ugly (too big, too complex), and among other things it
overwhelms our resources, so we redirect the interested parties to more
ambitious but also more strictly-targetted rendering engines (Pango, Qt,
ICU/ICJ is an interesting thing by the way, since they also target a wider
audience, not only X11 (to shape a manicheist, untrue, painting). I do not
know how it reacts to that class of requests, however.
This archive was generated by hypermail 2.1.5 : Wed Jun 15 2005 - 05:57:33 CDT