On Fri, 20 Jun 2008, Werner Koch wrote:
You assume arbitrary OIDs. I am talking about changing a one byte value
to a multi byte value. They both get assigned by IANA for OpenPGP use
and thus it does not make a difference.
Actually, my concern has to do with the fact that OIDs are of arbitrary
length. X.509 got it wrong; we might too.
How was PGP 7.5 handling this? Just one specified curve, or?
I don't know about PGP 7.5.
I've got some 7.5 ECC keys I could look at, though I suspect it was just
one curve. Jon or Hal could answer this better than I, though.
The practical problem with DSA 2048 is that you create a key, ask people
to sign it and the majority of people are not able to sign it because
there software can't handle that. We do not have a way to specify
preferences for algorithm lengths (that would in theory only be possible
for subkeys).
I agree that's a problem, but isn't the solution "upgrade the client that
can't handle the larger keys?"
--Len.