Re: [Sks-devel] Well connected?

[Andrew Gallagher]

On 01/09/15 16:38, Kristian Fiskerstrand wrote:
> On 09/01/2015 04:42 PM, Andrew Gallagher wrote:
> 
>> One thing that does concern me about the current arrangements is 
>> how manual (and ad-hoc) the peering system is. I can foresee 
>> scalability problems...
> 
> How so? What kind of vectors are you aiming to protect against? if a
> server drops behind on updates it is dropped from the main pool,
> additionally it is fully possible to run stand-alone keyservers that
> doesn't synchronize with any other keyserver.

For one, I'm worried about split-brain - particularly when it comes to
revocations. The pool (for understandable reasons) measures deltas in
terms of numbers of keys, but not their specifics. And you say yourself
that care and feeding of the pool's latency profile is a manual process.

> Peering protocol should be manual as there is a level of trust (not
> necessarily a very high bar, but one there still) required between the
> operators.

Indeed. Although I can think of plenty of ways to fsck with the pool
without requiring a recon connection...!

A

signature.asc
Description: OpenPGP digital signature