On 6 Mar 2003, Derek Atkins wrote:
Hey, I still use a v3 key... And I have a LOT of mail encrypted with
it. I would highly object to taking compat out of the spec.
Then use an implementation that speaks both RFC 1991 and OpenPGP. Don't
add cruft into OpenPGP because you have an emotional attachment to a dead
key format.
Adding verbiage to the extent of "you MUST NOT generate a v3 key" is
fine with me, however implementation SHOULD (if not MUST) be able to
parse a v3 key.
All of this is beside the point. "PGP Desktop", or whatever it is being
called today, could implement both RFC 1991 support, and OpenPGP support,
and not violate OpenPGP even if v3 keys weren't in the spec. It's just
doing two different protocols. (Just like OpenPGP says nothing about disk
encryption, but it's in PGP Desktop.)
The only thing that would have to change, functionally, is that people may
have to start encrypting messages twice if they are to a large number of
users: once for the people with v3 keys, and once for the people with v4
keys. Unfortunately, that is the state of the world now in some cases,
where IDEA is the cipher for v3, and 3DES is the cipher for v4. Not to
mention the v3 interop bugs in GnuPG, which, while resolved now, still
linger in old versions. Better that the protocol not try to handle these
cases, and instead leave it up to the application. (This change could be
implemented invisibly to the user.)
--Len.