I'am agree with Geoff :
MODS should be the "next generation bibliographic standard".
"An important design consideration within MODS is to simplify MARC".

About the "unicode vs. transliteration" issue, the answer for me is UNICODE.

>Allowing researchers to use Pinyin at the keyboard rather than
>forcing them through an alternate keyboard is still considered a "service"
>by some.
if you want to provide such facility, ok you could provide a optionnal field
that say that original datas where written in Pinyin (Pinyin is a language,
not a charset ?).
But all MODS data should be in unicode.

>The fact is that people using MARC *do* have both kinds of fields in their
>records, so by not including them we run the risk of making MODS less
>useful and therefore less used.
Currently, MARC softwares doesn't support yet MODS.
So when softwares engineers will provide import/export modules for MODS,
they will provide automatic (if possible) transliteration from/to unicode.

In western europe, a lot of ILS don't use Unicode and use MARC with
transliteration, but this is a headache for software engineers and probably
for librarians and end users.