So how about a new MIME field like Content-Transfer-Compression?
It's still an incompatible change to MIME. MIME UAs that didn't
recognize that header would try to treat the body part as if it
were uncompressed, and that would be worse than just defining
a new content-transfer-encoding.
It was an explicit design decision to NOT allow multiple layers
of encoding in MIME, in the name of simplicity. (It was also
an explicit decision to require standardization of new c-t-e's,
because the definition of new c-t-e's will decrease interoperability.)
Even though "gzip64" seems inelegant, it is a better "fit" into
MIME than trying to define a new mechanism which separates encoding
from compression.
Keith