Thanks for the clarification! As is so often the way, now that I have a
better sense of what you're doing, I would propose something almost but
not quite entirely different...

On Thu, 25 Apr 2002, Charles Muller wrote:

> I am glossing a Chinese, Japanese, or Korean word (term, temple,
> personal name, etc.) with its Chinese characters. Thus:
> english words <foreign>romanized Japanese, Chinese, or Korean
> words</foreign> <cjk>chinese characters</cjk>

> english words <term>romanized Japanese, Chinese, or Korean words</term>
> <cjk>chinese characters</cjk>

The key point here is that the stretch in Chinese (or CJK) characters is
in fact a gloss for the preceding <term> or <foreign> element, using a
different writing system. In which case, I would recommend it be tagged
as ..... <gloss>!

> english words <term>romanized Japanese, Chinese, or Korean words</term>
> <gloss>chinese characters</gloss>

As Ralph points out, the LANG attribute combines writing system and
language. Strictly speaking, therefore, if you tag something as e.g.
lang='ko' I am not sure that everyone would assume you meant romanised
Korean. The way the TEI currently defines the LANG attribute, you should
probably be using something like lang="ko-rom"  for romanized Korean and
lang="ko-cjk" for Korean in CJK.

Note that the ISO 639 code can be specified on the <language> element. To
quote from the spec for the ID attribute on <language>, to which LANG
attributes generally refer:

"Its value will usually be a language identifier from ISO 639, possibly
extended to indicate the writing system in use (e.g. heb-hel for Hebrew
written in Greek.)"  (