[HarfBuzz] Indic support in Harfbuzz

[Eric Mader]

प्रविण सातपुते wrote:
> 
> 
> 2009/12/14 Eric Mader <emader at icu-project.org 
> <mailto:emader at icu-project.org>>
> 
> 
> 
>     Jonathan Kew wrote:
>      > On 14 Dec 2009, at 16:26, kevin & siji wrote:
>      >
>      >> On Mon, Dec 14, 2009 at 12:39 PM, Behdad Esfahbod
>     <behdad at behdad.org <mailto:behdad at behdad.org>> wrote:
>      >> On 12/13/2009 11:03 PM, Parag Nemade wrote:
>      >>> Hi Behdad,
>      >>>
>      >>> On Friday 11 December 2009 01:57 PM, Jonathan Kew wrote:
>      >>>> On 10 Dec 2009, at 21:29, Parag Nemade wrote:
>      >>>>
>      >>>>> Hi,
>      >>>>>       I would like to know whether current harfbuzz code can
>     be used for testing the rendering of Indic scripts?
>      >>>> Not easily, I think. A little while ago, I started work on
>     Indic support (initially concentrating on the "new" Devanagari
>     standard as used in Vista/Windows7), but AFAIK I don't think Behdad
>     has committed any of that code to the repository yet. At that time,
>     there were still some changes happening in the internal
>     shaper/features APIs. I've been busy in other areas for the past few
>     weeks, so have not been pressing to get this integrated.
>      >>>>
>      >>>>
>      >>>         Can you check if Jonathan's code can still be
>     integrated in harfbuzz?
>      >> It's not.  Will get to it this week.
>      >>
>      >> Is it worth reinventing the wheel? Is it not possible to use the
>     code from ICU or Pango? Already a bunch of complex script rendering
>     engines (ICU, Pango, Uniscribe) has made the Indic rendering scene a
>     mess. Currently fonts should have separate ligature rules according
>     to different rendering engine behaviours to create a single expected
>     result universally.
>      >
>      > AFAIK, neither ICU nor Pango code has support for the new Indic
>     script tags and shaping behavior, so we'd need to
>     adapt/extend/rewrite/something for that. And the shaping code
>     interacts closely with the glyph buffer and with the feature
>     controls, all stuff that differs between engines. So "use the code
>     from ICU or Pango" is nowhere near as straightforward as it sounds.
>     We may do that, but it would still need considerable work.
> 
>     ICU has code to support 'dev2'.
> 
> 
> does ICU code rendering matches with Uniscribe?

I didn't do this implementation, so I don't know how much testing was done.

    > In view of all that, I decided to experiment with writing a
>     'dev2' shaper directly from the spec, to see how that would work out
>     (and to learn more about the details of the new spec, in the
>     process). If we decide in the end that the old Pango or ICU code is
>     preferable, that's fine. I'm not particularly interested in
>     reinventing any wheels, if the old wheels can be upgraded to meet
>     today's need, but it was not immediately clear to me whether that
>     would actually be any easier or lead to a better result.
>      >
>      > JK
> 
>     Microsoft has claimed that the Indic specs needed to be rewritten
>     because the old specs weren't sufficient to implement all the required
>     behavior for all the scripts. I don't know the details, but I'm inclined
>     to believe them.
> 
> 
> hmm
> 
> All the effort on harfbuzz are going on to make a unified rendering 
> engine and that is must things today. And it is better to try to match 
> it with Uniscribe (i guess it is with latest OT specs). so after that 
> users should not complain again fonts is working on Linux but not on 
> windows/windows fonts not working on linux. All opensource community can 
> put effort in fixing issues only at one place not like bug for ICU, 
> pango etc.

I agree that there should be a single OpenSource layout engine that the 
whole OpenSource community uses. That is, as I understand it, the goal 
of HarfBuzz.

Given that Microsoft will continue to use Uniscribe, I think it's 
important to make sure that all the fonts that work on Windows work w/ 
HarfBuzz and visa versa. This is what users will expect, and it seems to 
me to be a reasonable expectation. Plus, it will mean that there are a 
lot more fonts available on OpenSource platforms. ;-)


> Pravin S


Regards,
Eric