On Tue 2016-04-12 04:34:09 -0400, Vincent Breitmoser wrote:
Daniel Kahn Gillmor(dkg(_at_)fifthhorseman(_dot_)net)@Mon, Apr 11, 2016 at
08:40:22PM -0400:
* it should be cheap to compute from a given key -- you shouldn't need
a gig of RAM or a minute of CPU to calculate the fingerprints of any
key.
Strictly speaking, we can be slightly less restrictive: It must be
cheap to verify, given a fingerprint, that it's the correct one for a
key. This distinction does not make a difference unless we store the
fingerprints as part of the data format (which we probably shouldn't),
so this is more of an academic point.
Right, i don't think we should store the fingerprint as part of the data
format, so we still need to be able to rapidly generate it, not just
verify it.
* it should be strong enough that we do not believe anyone can create a
key with a fingerprint that collides with another key's fingerprint
Quite importantly, this should be "another *independent* key's
fingerprint", i.e. the requirement is preimage resistance, not
collision resistance. Creating two keys with colliding fingerprints
is fine, at least noone could come up with a attack scenario where it
mattered.
This clarification also matches my understanding. Thanks for the
precision, Vincent.
--dkg
_______________________________________________
openpgp mailing list
openpgp(_at_)ietf(_dot_)org
https://www.ietf.org/mailman/listinfo/openpgp