[HarfBuzz] Handling OpenType layout across itemisation boundaries

[Khaled Hosny]

On Mon, Nov 09, 2015 at 09:11:28PM +0000, Jonathan Kew wrote:
> On 9/11/15 18:31, Khaled Hosny wrote:
> >While reading page 13 of John Hudson’s IUC39 presentation [1], I was
> >wondering if HarfBuzz can handle script, language and direction
> >properties the same way it handles optional features; i.e. instead of
> >sending HarfBuzz buffers that has a single script, language and
> >direction, one sends a single buffer for the whole paragraph with
> >arrays of script etc. ranges similar to hb_feature_t ranges and let
> >HarfBuzz do its magic on it, allowing for any cross boundary lookups to
> >be applied as the required context will not be lost.
> >
> >WDYT?
> >
> >1. http://tiro.com/John/Hudson_IUC39_Beyond_Shaping.pdf
> >_______________________________________________
> 
> While this might be nice in principle, I suspect it would be of limited
> value in most real-world use cases, because it's very common that script-
> and language-run boundaries also turn out to be font boundaries. (This is
> much less likely to be the case for language changes within a single script,
> of course; but it's also relatively uncommon for content to actually be
> marked-up for language at this level.)

Language is not as important, I just included it for completion sake.
But interaction across script and direction boundaries is more common,
there is even a case of Unicode changing the bidi type of some symbols
(Arabic subtending marks?) because it was nearly impossible to properly
support them in OpenType otherwise.

Regards,
Khaled