Re: [openpgp] v5 sample key

[Daniel Kahn Gillmor <[e-mail]>]

On Thu 2019-04-25 10:11:33 +0100, Jonathan McDowell wrote:
> A move to easily enable key material > 2^16 bytes seems to be in
> conflict with dkg's work on trying to reinvigorate the usefulness of key
> servers + the suggestion to limit key material packets to < 8383 bytes.

For existing OpenPGP key types, i do hope that sane implementations will
keep to the limits i've suggested in
draft-dkg-openpgp-abuse-resistant-keystore.  But i think Werner's change
to 4-octet length representation during fingerprint and signature
calculations is intended to allow for significantly larger keys if there
are any post-quantum algorithms that need it.

If we get to a world where PQ keys of that size are necessary, the
guidance on abuse-resistant keystores will need to be updated to
accommodate the new algorithms, of course.  But presumably any
cryptographically-validating keystore will need significantly more work
to implement the PQ algorithms, so a change to those suggested limits is
pretty minor in comparison -- not too big a deal.

And to be clear: this isn't mandating a 4 octet length on the wire for
all keys -- keys in transport can still use OpenPGP's weird
packed-type-length encoding and shave off a couple bytes that way when
they're aiming for smallness.  We're only talking about 4 octets in RAM
when calculating the key fingerprint or the signature, right?

so i agree with Jonathan that it's not tightly aligned with the work to
set forward simple guidance for abuse-resistant keystores, but i don't
think it's in direct conflict with it either, and i support the move to
a 4-octet length in fingerprint and signature calculations for v5.

     --dkg