It has a number of Mule character sets that correspond
explicitly to ranges of the UCS, mule-unicode-2500-33ff,
mule-unicode-e000-ffff, etc. This is practical for the moment
as people don't use too may exotic characters, but will still
fail in the same way XEmacs does once Unicode characters
outside that set are encountered. Anyway, UCS as our internal
encoding isn't obscenely difficult.
We should just do that.