Michael Haardt wrote:
Shouldn't the implementation be free what to convert them to? It may
chose a different unicode representation. Do we need to enforce UTF-8?
I realize the language in 3028 clearly implies implementations are
required to convert to UTF-8 but if an implementation wants to use UCS-4
or UCS-2 or UTF-7 internally, that must be allowed. The specification
has no power to specify behavior that can't be externally observed, and
the text you cited is just wrong.
Perhaps the text should have been simply "MIME parts SHOULD be
canonicalized to Unicode prior to the match". I'm sure there's a way to
phrase it in a way that's completely encoding-neutral, though.
Tim