Le samedi 31 janvier 2009 à 23:42 -0800, Roozbeh Pournader a écrit :
I was trying to avoid fontconfig, but it caught me at the end.
Trying
to figure out which languages in comps are supported by which font, to
be able to include them in the language group, I compared the list of
languages in F11's comps file with the orthography lists fontconfig
supports.
To my surprise, some were missing/problematic. The list is here:
http://fedoraproject.org/wiki/I18N/LanguageSupportCriteria/Missing_fontco...
I'd be very careful to redefine tagalog for example. The tagalog script
is definitely not latin and has a specific unicode block, iso 15924
script tag, and specific supporting fonts
http://www.unicode.org/iso15924/iso15924-codes.html
http://fedoraproject.org/wiki/Pmorrow_Tagalog_Doctrina_1593_fonts
IMHO we've been conflating too many different notions (country, region,
language, script) in a few short locale tags and what you see is just
the system breaking appart for non-mainstream languages/scripts.
I also found quite a few problems with existing orth files in
fontconfig that I'm working on fixing:
http://tinyurl.com/dbk6a8
This is something for Behdad. I hear he's cutting a new fontconfig
version right now, you may want to catch him before he's done and the
projects wents back to its usual 6-8 months sleepiness ;).
Since fontconfig support is critical for any language Fedora claims
to
support [1], I think we should remove the language groups from comps
file if we don't have a fontconfig orthography file for it.
That's not really helpful, all the different parts of language support
are done by different groups with different agendas and different time
scales so of course initial support is going to be incomplete. Expecting
one contributor to provide all the parts in one go is illusory.
What you want is to help the different contributors to a language group
to:
— identify other bits are missing
— ping in the right place so they get added
Otherwise everyone will just wait for everyone else.
Some people will claim that « full » language support means a system
dictionnary and thesaurus BTW, completeness is a slipery slope.
I went and
updated the language criteria page we have here, adding a fontconfig
step:
https://fedoraproject.org/wiki/I18N/LanguageSupportCriteria
For the specific language cases, I went and filed upstream bugs
against fontconfig for all I could find, except for Berber, which is a
bit problematic by nature (language code used is actually for a family
of languages, glibc locales are incomplete, Latin/Tifinagh/Arabic
script division is not along country lines...).
This again is because iso-639 is not well suited to identify scripts.
This is a report. I would appreciate help and feedback, especially
your thoughts about fontconfig .orth requirements for claiming
language support in Fedora.
--
Nicolas Mailhot