[Fwd: Re: Request for review and advice on wqy-bitmap-fonts fontconfig settings]

Tue Dec 4 19:49:50 UTC 2007

Le mardi 04 décembre 2007 à 11:40 -0500, Qianqian Fang a écrit :

> Unfortunately the Chinese user community is quite weak in communicating with
> the upstreams, majorly due to language reasons. More than half
> of the users do not like to use English to discuss their problems, the
> vast majority of the feedback and problem-solving were done at various 
> Chinese
> Linux forums, BBS (bulletin board system) and instant messaging.
> Even those who are able to describe their problem clearly in English,
> only a small fraction went through all the culture training
> and practicing and become a contributor.

This small fraction needs to organise itself, identify problems in
Chinese font/text support, possible fixes, and relay them to upstream
projects. For example every year freedekstop.org organises a text summit
where pretty much every project that counts in FLOSS text rendering is
represented:

http://unifont.org/TextLayout2007/

It would be extremely helpful if the Chinese FLOSS user community sent
someone to next year's summit to list the main problems affecting
Chinese users, and what the Chinese community feels needs to be done in
projects like fontconfig or pango to fix them.

The rest of the year having clearly identified Chinese relays people can
ask questions to (like "if I do this will it break Chinese apps") may
probably help too.

> The bad lucks of propagating patches back to upstream is also another
> reason that discourages Chinese to get involved. Chinese is one
> of the most complicated scripts and is always challenging to
> get what people expect without altering the default Latin handling,
> therefore, the upstream developers are very cautious about any change
> related to Chinese (or CJK).

CJK is difficult sure but do not underestimate the part lack of
communication plays. Any maintainer will be ultra-cautious about making
CJK changes when he knows that if he makes a mistake users are likely
not to report back but suffer silently for years while cursing his name.

> This also negatively impacts the situation.

As a counter example you may have noticed there's been a lot of
Greek-related activity on the list lately. It's not because Greek is
easier or more interesting that other languages, but because the Greek
community managed to organise itself. As a result they're getting good
support from every distribution, Fedora included.

> As a result, Chinese users HAD to find out work-arounds to meet their
> day-to-day needs. You may be supprised that almost all Chinese linux
> forums have a board called "Font Beautification", it sounds ridiculous but
> this is true. People used to spend days or weeks trying to fix their
> Chinese font settings for all applications.

I'm not surprised at all this is typical workaround culture.

>  That is also my motivation
> to create the Wen Quan Yi project, just trying to save people's time
> and make Linux easier to use by Chinese.
> 
> I can do my best to help pushing the fontconfig scheme that you mentioned,
> but I am not supprized if that still not implemented after years.

I'll be honest even if someone actively pushes fontconfig changes it may
take a year for them to be integrated and yet more time for the changes
to percolate in distribution. Getting fontconfig to change is not easy.
However if you don't try I'm pretty sure nothing will have changed in 5
years.

And you probably need fixes at other levels too. Just like the Wen Quan
Yi project is part of the solution, but not the whole solution, fixing
fontconfig will probably not be sufficient. For example even if
fontconfig selects the perfect Chinese font when told to render Chinese,
apps still need to detect they are rendering Chinese, which is not
possible basing yourself only on unicode points, or the session locale
(though for this particular problem you are better of than Latin
languages since you only share codepoints with Japanese)

> But there are immediate needs to use Linux in a Chinese-friendly
> way and a good work-around can really build up the expanding
> user community and likely developer group, and that could
> make the life easier in the future. That's my rationale to
> push a reasonable fontconfig file for my font.

The work-around limits as you've found out is they get removed as soon
as someone else complains of them. No one really wants to choose between
Latin, Chinese or Japanese users at Fedora, so if two user communities
conflict the one stepping on the other loses.

By selecting  a fontconfig priority of 61 you pretty much removed
everyone relying on fonts with a less than 61 priority from the picture.
That leaves people relying on fonts with a more-than-61 priority to
complain. I suspect some of them, most likely Japanese users, can still
be negatively affected by your changes but I'm no Japanese speaker so
that's up to Jens to confirm (or infirm). And even if Jens greenlights
there is still the possibility of later complains causing to remove your
changes.

Lastly for Fedora ≥ 9 we'll probably use DejaVu full not DejaVu LGC as
default, so you may want to adapt your fontconfig file accordingly in
Fedora-devel (you'll note DejaVu full as default got blocked for several
releases due to the same kinds of conflicts you're encountering, and is
only pushed now we're more confident in its non LGC parts. We try to be
fair to everyone)

Regards,

-- 
Nicolas Mailhot
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 197 bytes
Desc: Ceci est une partie de message
	=?ISO-8859-1?Q?num=E9riquement?= =?ISO-8859-1?Q?_sign=E9e?=
Url : http://lists.fedoraproject.org/pipermail/fonts/attachments/20071204/8fb745ac/attachment.bin 