[HarfBuzz] [Indic] Inspecting the font for consonant position with Free Sans

Behdad Esfahbod behdad at behdad.org
Tue Mar 5 23:33:34 PST 2013

On 13-03-06 01:38 AM, Pravin Satpute wrote:
> On 03/06/2013 10:40 AM, Shriramana Sharma wrote:
>> >
>> > Why not just ask the Free Sans guys to fix it?
>> >
> Indeed, still i have not checked the FreeSans.
> Uniscribe has some specific things handled which are not documented
> anywhere, because of those handling even some wrongly written OT tables
> work.
> IMHO this does not mean in harfbuzz-ng also we have to handle such
> invalid cases, i think rather we should try to fix the font.

As we all know, the OpenType Indic specs are insufficient for a full
implementation.  As such, "what Windows does" is de facto standard.  If fonts
work with Windows but not HarfBuzz, people call it a HarfBuzz bug.  So we try
to support all weird cases that Uniscribe does.

That said, there are cases where what Uniscribe does is clearly wrong, or
inferior to what HarfBuzz does.  In those cases, we feel free to diverge from

So, it's not that we want to match Uniscribe bug to bug.  But given a font and
a sequence, if what Uniscribe shows is expected of the Unicode sequence, then
we want to match it.


More information about the HarfBuzz mailing list