[HarfBuzz] Question about HarfBuzz-ng OpenType Implementation

Ed ed.trager at gmail.com
Tue Sep 14 11:32:39 PDT 2010

Hi, Behdad and Everyone,

I'm not sure if I know the best way to word the following question
--which is also a sort of “feature request” too-- but I will try
anyway:  Once you find time to implement support for complex script
OpenType layout stuff in HarfBuzz-ng, will you insure (1) that the
various OpenType features are *always* available *regardless* of the
language or script tag, and also (2) always work correctly regardless
of whether the font is "ttf" or "otf" format ?

One of the problems with the currently-available layout engines is
that even though a font designer may implement certain OpenType
features within a font, the layout engine may not process them for
certain languages or scripts, or simply may not process them at all,
for reasons that are usually not clear.

For example, I just learned from Vernon Adams that my GPOS
mark-to-base lookups that I began working into a Tai Tham font I am
designing  will not work in a lot of existing
software and platform combinations:

>   <<< excerpt of email from Vernon Adams: >>>
> Your gpos lookups in OpenType fonts (.ot) do not work in all
> applications. E.g. Scribus, OpenOffice dont work. I'm guessing this
> maybe same with Inkscape. However a TrueType font with opentype lookups
> does work, at least in Scribus, OpenOffice and Firefox (the 3 apps i
> have tested).
> So basically, use opentype flavoured Truetypes instead of pure
> OpenTypes.
> Also... your GPOS mark to mark and/or mark to base Lookups will only
> show correctly in Firefox 3.6+ on Linux. Firefox 3.6+ on  OSX or Windows
> will not read those gpos lookups. Chrome, IE explorer, wont either.
> Safari 4+ on OSX seems to partially read something i think but basically
> it doesnt.

And of course, that is not just my problem: lots of people, like
Vernon who is working on a Laos font, are constantly implementing
imperfect, laborious, and footprint-augmenting work-arounds in fonts.

A lot of people like myself are hoping that the emerging webfont
technologies are going to make it much easier to publish on the
internet (inter alia) using some of the less-common scripts which may
have only recently been encoded in Unicode, such as Tai Tham.

Achieving such a goal is not currently very easy, for many reasons.
Continuing to use Tai Tham as an example, right off the bat I see that
a script tag for Tai Tham is not yet published:


... and furthermore, even if Microsoft does publish a script tag for
Tai Tham, how long will it be before Uniscribe knows how to deal with
it?  Forever?  Yes.

Tai Tham is only one of many scripts that appear doomed to suffer
miserably under OpenType's less-than-stellar implementations.
Especially all of those persnickety scripts of South and South East
Asia, of which Tai Tham is but one.

Best - Ed

More information about the HarfBuzz mailing list