John's message suggests an obvious compromise on the richtext front:
Keep the 8859-* identifiers in richtext, since we all know what they
mean, but not the identifiers for any character set which isn't a
superset of ASCII, since we don't know what they mean, at least with
respect to parsing richtext. Is this a reasonable comproimse? It will
allow shifting between European languages, in particular, with the ease
that David wants, without any of the alarming confusion of switching to
a character set in which "<" and ">" don't mean what richtext wants them
to mean...