Chinese font families

From Internationalization

At a recent workshop organized by the Beihang Host of the W3C, representatives of Chinese industry requested consideration of some significant issues faced by Chinese users and developers.

There is a Chinese version of this information at http://www.chinaw3c.org/issue-chinese-fonts.html

The Problem as Originally Stated

With 7-8 major font types, Chinese fonts don't map very well onto the generic font families described in the CSS3 Fonts spec.

Furthermore, there is a particular issue related to the file size of Chinese fonts, which can easily be 10Mb in size. If a user doesn't have the right font on their system, downloading a font with a web page is often not very practical, especially given the high proportion of people accessing the Web via mobile devices in China, some of whom have to pay by bandwidth used.

Vendors in China see it as important to be able to substitute existing fonts on the user's system, where needed, in a way that matches Chinese font types, rather than just those generic font families provided by the current specification.

Picking up on hints such as substrings like song or hei in the font name is often inadequate.

The question is, how to make it possible to map fonts used in Web/eBook content to system fonts, where needed, in as predictable and effective a way as possible.

Additional notes

Presumably, this problem also affects Japanese and Korean users, and it is possible that the solution may also cover their needs.

There may be other scripts that require attention with regards to the suitability of current generic font families. For example, Arabic distinguishes between font styles such as naskh, nastaliq, ruq'ah, kufic, diwani, sini, etc. which have very different characteristics and uses (and in some cases use different typographic rules for things such as justification). They also do not map well onto the generic font families defined in the spec.

Possible Solutions

The following are just brainstormed ideas, which need to be explored using email or other forms of communication.

[1] Extend the generic font families currently defined by CSS to cover Chinese/EastAsian needs, ie. re-using existing generic font family names where possible.

[2] Define a complete set of new Chinese/EastAsian generic font families in the CSS spec.

[3] Establish a registry of Chinese/Asian fonts and how they map to the asian font families, so that developers can produce predictable mappings, and do so quickly.

[4] Automatically classify the fonts on the user's system into the generic font families by analysing and comparing the font glyphs to those expected for a generic font family.

[5] User-friendliness of existing generic font names may be helped by allowing local aliases. For example, the serif generic font family may be more meaningful to Chinese users if they can use something like song. However, Japanese users may prefer the alias mincho, and Koreans batang.

Note that if we add new generic font family names to the spec, we need to find identifiers that won't be confused with font names.

The Chinese community needs to propose what generic font family names are needed for Chinese. It would be useful if experts in Japanese and Korean could reach consensus on names that would serve all three scripts.


Questions

Preliminary information suggests that there is not a great deal of font diversity in Japan, and that content authors are able to address the problem mostly by listing in their CSS font-family values fonts that are known to be supported on user's operating systems. What makes the difference in the Chinese case?

What are the generic font families that would be required for Chinese support, and which are needed to go beyond what is currently available, ie. serif, sans-serif, cursive, fantasy, monospaced?

What is the average size of webfont downloads for Chinese fonts? Is subsetting generally practiced?


Latest Views on the Issue

The Chinese IG has been discussing this issue. This section attempts to represent current thought on the matter.

The CSS Fonts spec mentions Kai font-family in connection with the serif generic font family (as well as Song). The Chinese IG thinks that it should actually be treated as a cursive generic font, as should Fangsong. If this is the case, it should fix the majority of the problems.

There is a great variety of fonts in China, so making a comprehensive list to allocate each to a generic font family is probably not helpful. However, it will be necessary to ensure that browsers are able to produce a Kai substitute font if the cursive generic font family is specified by the content author. A test case will be created for that, and the Fonts spec should probably be updated at some point.

For the issue related to webfont size, the feeling is that this is not likely to be a major issue in most cases, since special fonts that users may not have on their system are likely to be used only for titles or other short pieces of text. With subsetting, the webfont size should be ok for such uses.