Somewhere on the net, I learned that LOCALE should be a combination of Language/Country with a form of la_CO with the country part optional.

That is common, indeed. See below for the relevance of encodings, though.

So, there should be zh_TW and zh_CN instead of big5 and gb. Am I right?

Not sure what mailman is doing, however, either appears to be wrong, based on the following rationale: - a locale/catalog should not, per se, be based on the encoding. Instead, at run-time, the program should convert from one encoding to another. Is it true that big5 and gb can be converted forth and back for all practical purposes? - The locales zh_TW and zh_CN may differ, but the catalogs should *only* differ if the translations are actually different. There is no point in duplicating work if the Chinese and the Taiwanese translator come up with the same translation (perhaps in a different encoding). If you want to put the encoding into a locale, it normally goes after the country, i.e. zh_TW.big5 and zh_CN.gb; other combinations should also be possible, like zh_TW.gb or zh_CN.UTF-8. Regards, Martin