Awesome stuff and super excited to see some stronger language against correlation, especially the idea of using publicly known clusters (which is where this makes the sense). Some of my iniitial thoughts:
- I think the whole discussion around client diversity is enlightening in this regard. The inactivity leak penalty is massive for correlating failures but yet it’s not doing much to actually jump start client diversity. Whether we’re in this state or not is worth looking into (I’m guessing it’s a curve of adoption that once we’re past, they don’t mind)
- As others have pointed out, a CEX that owns 90% of the stake but that distributes it amongst various LST’s/DVT’s is still a piss poor validator set when it comes to decentralization.
- Addressing systemic risks – discouragement attacks against centralized validator sets - #2 by aelowsson. Wrote this post on a discouragement attack that seems relevant here. This would make this particular discouragement attack even more severe as a large proposer (who also has lots of aggregators) could just ignore attestations of parties they claim are correlated.
For a question, If we don’t like the public mapped correlation, why are we not just coming to consensus on it and just limiting rewards/penalizing for size?