My impression is that MIME has done an adequate job of handling I18N
in message bodies and in Subject: lines. The remaining ASCII-only
parts of the message are the headers that include addresses. They
present a difficult I8N problem because the MIME encodings provide
multiple ways to represent a string, but mail addressing demands an
exact match, other than case folding in domains.
I presume that we'll eventually use some version of punycode since
it's essentially the same problem as I18n DNS, but it's way too early
to say anything about that.