Around 12 o'clock on Jul 8, Edward Lee wrote:

> > The question is whether we should mark certified GB18030 fonts as suitable 
> > for zh-TW as well as zh-CN.  I have the GB18030 varient of SimSun here in 
> > TrueType and it does not have the traditional Chinese codePageRange bit 
> > set, but does cover the codepoints found in Big5.
> 
>  There are `two' Traditional Chinese fonts here. In zh-tw the
>  radical/stroke of some glyphs are differrent with the TC glyphs
>  in GB18030 fonts.

Then the goal should be to identify GB18030 fonts as only zh-cn.  For 
users who have only a GB18030 font and no traditional Chinese fonts, the 
zh-cn mark will cause that font to be preferred over a Japanese or Korean 
font, but for people with fonts marked zh-tw, they'll always get the zh-tw 
font instead of a zh-cn font.

For GB18030 fonts that come in TrueType format, the codePageRange bits 
already set the appropriate hints -- they indicate the font is suitable 
for simplified Chinese, but do not indicate that the font is suitable for 
traditional Chinese.  Clearly, I should use the codePageRange bits to 
help select Han language targets -- when any of the Han codePageRange bits
are set, exclude languages which aren't listed in the codePageRange bits.

>  Not only typographic style, but different radical/stroke.

That's really bad; I know that looking up Chinese characters in my 
Japanese dictionary is often an adventure.

>  Is there any other solutions?

Yes, we have lots of possible solutions, I'm just searching around for the 
right one that will generate the best possible result for the greatest 
number of people.

What might be idea is to be able to do is order the available Han languages
and let people configure the 'fallback' order.  A user might prefer a
Japanese font for Chinese text when they have no Chinese fonts.  

Right now, all I've got is three levels: matching language and country,
matching language and matching neither -- the user in this case will
essentially get a random font with Han support when they don't have one
suitable for the target language; the font selection will be driven 
strictly by Unicode codepoints.

Obviously, I'd like to avoid introducing arbitrary configuration 
complexity if this solution isn't adequate; we've already got a pretty 
sophisticated system that will take some considerable understanding before 
people can really get the results they want that it is capable of 
producing.

Keith Packard        XFree86 Core Team        HP Cambridge Research Lab


_______________________________________________
I18n mailing list
[EMAIL PROTECTED]
http://XFree86.Org/mailman/listinfo/i18n

Reply via email to