ietf-822
[Top] [All Lists]

Re: plain-text checksums considered useless

1991-10-29 01:52:47
On Mon, 28 Oct 1991 20:49:01 -0500 (EST) you said:
OK, I'm convinced:  checksums belong in base64.

Any objections to an optional checksum at the end of base64-encoded
data?

It should definitly *not* be optional (sorry).

In a previous life, I did devise a base-64-like system with checksums. Another
'safe' character (there is still some room in the invariant ASCII/EBCDIC) was
reserved for the purpose. This character could appear after any full quantum,
and was followed by a quantum encoding the checksum of the portion of the
binary data starting after the last checksum (or at the head of the encoded
object for the first one). Since the encoder can put several checksums, many
of them for really prudent use, a very powerfull algorithm is not needed.
The checksums, once in the encoding, stay there across mail gateways, and even
resist cutting/pasting of bodies in simple UA's, where a header-based checksum
will likely disappear. The decoder, in the presence of frequent checksums,
could tell where a problem had been detected, so (a knowledgeable user, at
least) was enventually able to only re-get one or a few parts of a multipart
object.

Any decoder had to be able to absord and verify the checksums, while any
encoder was required to put a checksum at least every 3x encoded bytes,
and another one at the very end. To be sure that the complete object had
been received without imposing two passes to the encoder, yet another safe
character (there is room) was reserved to mark the end of the encoding.
Last detail : since that system was able to delimit records (like base-64
with the comma), the record markers were somehow included in the checksum.
                                                            /AF

<Prev in Thread] Current Thread [Next in Thread>