- From: Jungshik Shin <jshin@i18nl10n.com>
- Date: Fri, 10 Oct 2003 09:08:49 +0900 (KST)
- To: Richard Ishida <ishida@w3.org>
- Cc: public-i18n-geo@w3.org
On Thu, 9 Oct 2003, Richard Ishida wrote: > Congratulations to ourselves ! This announcement has just gone out on > the W3C home page. Thank you for your great work ! I just skimmed over it quickly and came up with a couple of comments. In 4.1 Choosing & specifying fonts, it has to be noted that whenever localized font names (family names) are specified (in CSS), all-ASCII font names should be specified alongside because localized font names (in characters outside US-ASCII) are not available on all platforms and nor recognized by all user agents. For instance, the name of a font returned by Win32 font-related APIs depends on the default system locale so that on fonts whose name are specified in Japanese (Chinese/Korean) [1] wouldn't be matched by most [2] user-agents running on non-Japanese(Chinese/Korean) Windows. Another example is fontconfig library used by both Mozilla and Konqueror on Linux and other free Unix. It doesn't yet support international localized font names. Yet another example is XLFD (X11 Logical Font Description) still widely used on commercial Unix despite a number of shortcomings. AFAIK, XLFD doesn't support anything other than US-ASCII in all its fields (including foundry and family name fields). Given all these, it's essential that family names be specified in US-ASCII only strings (along with localized names). In 14.1 Date & Time, it's adised that 'words for the month be used and an example (02 mar 2004) is given. I'm not sure of the wisdom of this. For speakers of European languages, that may be a good advice, but for East Asian users, that doesn't seem to as good (and I'm not sure of other regions/languages). At the minimum, YYYY-MM-DD format (as specified in ISO 8601) should be mentioned as a rather universal and culture/language-neutral alternative. Jungshik [1] Win32 APIs return localized family names only on CJK windows. [2] It's certainly possible that user agents themselves can look 'deep' inside a font to figure out ASCII or localized family name even though the OS API doesn't.
Received on Thursday, 9 October 2003 20:11:00 UTC