Ralph writes:
I don't think these ‘Preferred MIME names’ would introduce extra NULs.
It's basically any single-byte stateless encoding which only maps NUL to
NUL.
US-ASCII
ISO-8859-{1..10,15}
Windows-1252
And many others.
I'd rather not enumerate the many others. Maybe another approach
could be the following?
1. Attempt to decode the body, given any encoding.
2. If unable to decode or if any NUL octets in the decoded text,
leave the body encoded. If unable to decode, that's allowed by
RFC 2047 § 6.2. If any NULs, that seems to be in the same spirit.
David