I agree completely with Craig's message that:
-- BASE64 should have a checksum.
-- That the checksum be included with the BASE64 data.
-- That the checksum not be in the header.
-- That BASE64 checksums is a seperate issue from message checksums.
But for a different reason than Craigs.
One factor of whether BASE64 becomes an accepted encoding ids whether
publicly available encoders/decoders are available and widely
distributed. This is especially important during the/any transition
period in which XXXX capable MTAs and UAs are rare so that end-users
can take apart XXXX style messages.
By adding the checksum to the output of b64encode (proposed name) it
becomes much more useful in the same way that uuencode+checksum is more
useful than plain jane uuencode.
If the checksum is seperate from the BASE64 text the checksum probably
will not be carried along & not be available to b64decode.
Finally.. I am not at all certain that checksumming is needed or useful
over an entire message, it probably is but this isn't certain. I *am*
certain that it will be useful for certain body parts. If there were a
checksum over the entire body it should *obviously* be in the header.
There may be times when it is appropriate to add a checksum for a
particular bodypart to the mini-header for that body part. This is not
one of them.
David