On Tue 2015-03-17 11:55:33 -0400, Wyllys Ingersoll wrote:
Right, but they don't. The application doing the encryption really doesn't
know what the original character encoding was, so doing the conversion can
result in lost or corrupted messages.
Character encodings are a problem for message signatures too. The fact
that the messages don't embed any knowledge of this means that it's
possible to subtly alter some textual content.
See the "message tampering through header substitution" section here for
a quickly-contrived example (someone with better knowledge of existing
character sets can probably craft a more clever substitution):
https://dkg.fifthhorseman.net/notes/inline-pgp-harmful/
I think the only way to get it right is if the message composer (i.e. mail
client usually) is tightly coupled with the PGP engine so that it can pass
along the character encoding when the encryption request is made.
Or have the composer only send well-structured data that includes
framing information (and have the reciever know how to parse the framing
information).
This has to happen *inside* the signature (and inside the encryption,
for that matter). For PGP/MIME, that work has already been done: the
MIME structure is the framing information.
If you pick some other sort of framing, you're kind of on your own
afaict -- i don't know how many other standards there are out there.
--dkg
_______________________________________________
openpgp mailing list
openpgp(_at_)ietf(_dot_)org
https://www.ietf.org/mailman/listinfo/openpgp