Is this inconsistent with either your expectations or RFC 1113?
Your diagram is fine and is consistent with my expectations. Note I
suggested that base64 was a candidate canonical encoding, so I have no
trouble with someone inventing another.
However, if I understood what you said correctly, your principal objection
to using base64 encoding as the canonical form was that it did not require a
specific line length. I do not think this is an issue, since the canonical
form could be a byte stream, ie compute the checksum before breaking it into
lines for inclusion in the message. Does this address your concern?
Jim