[HarfBuzz] Hackfest report
behdad at behdad.org
Mon May 28 10:15:57 PDT 2012
On 05/28/2012 03:32 AM, Pravin Satpute wrote:
> Recently i had conversion with Wikipedia team on the word list and
> answer received was use wikisource, it provides digitized book. These
> contains mostly validated words from books.
Interesting. I'll ask Roozbeh to take a look.
To be fair, we *do* want to have the garbage in the corpus too. After all,
there's lots of garbage on the net. We just want to pick and choose which
kind of garbage we want to support and which to ignore, and the number do
exactly that for us.
> Excellent achievement, congrats :)
> I will go through 560 words and will update you if anything interesting.
I'm trying to streamline the testing process and get the test data public to
get collaboration going.
> I think Punjabi and Tamil should also give good results with current fix
> for Devanagari script.
I ran these through. Tamil is at 0.87%, which is really nice. There's 806
Punjabi / Gurmukhi, is more like 10%, but majority of cases are positioning
differences around uni0A71.alt. It's probably one change from going down to
sub 1%. I'll take a look.
> I will say this is Best approach. Thinking all script at once is pain.
Indeed. And having a dashboard up, testing them continuously so we won't regress.
> Might be more people cant join face to face meeting, at least i will try
> to be more active on IRC during same time.
The nature of the meeting makes it hard to interact over IRC, but I'll make
sure that next time it's announced properly on the list.
In the mean time, and after I get the suite online, what helps is people
running it, checking the failures, and put to words what it is that's failing.
>> PS. I'm leaning towards shutting down the harfbuzz-indic list and using the
>> main list for all communication. Any objections?
> Might be not required presently but when we will deploy harfbuzz-ng in
> various project, i..e pango, icu might be that time we will need it.
Right. For maintenance period it's useful. Right now, I would be too worried
about main development not reaching that list.
> Best Regards,
> Pravin Satpute
More information about the HarfBuzz