recommendation for key servers

[Andrew Gallagher]

On 06/07/2021 20:59, Daniel Kahn Gillmor wrote:
> On Mon 2021-06-28 18:42:02 +0100, Andrew Gallagher via Gnupg-users wrote:
>> It’s not clear, but it may be due to a lack of canonical ordering of
>> packets.
> 
> There are no published specifications for how to canonically order
> OpenPGP packets, but i sketched a proposal here:
> 
>      https://dev.gnupg.org/T3389
> 
> Adoption of such a canonical ordering would reduce the amount of
> computation for synchronizing keyservers, once they all adopted the same
> one.

That's an interesting idea, and it has merit in itself, but from a 
keyserver point of view I think a more general solution is to explode 
TPKs into atomic components, sync them separately, and reconstruct the 
TPK on demand at query time. This addresses not just reordering of 
packets, but also differential filtering, simultaneous updates, etc.

See https://github.com/hockeypuck/hockeypuck/issues/137

-- 
Andrew Gallagher

-------------- next part --------------
A non-text attachment was scrubbed...
Name: OpenPGP_signature
Type: application/pgp-signature
Size: 833 bytes
Desc: OpenPGP digital signature
URL: <https://lists.gnupg.org/pipermail/gnupg-users/attachments/20210706/f6e3e074/attachment-0001.sig>