Dan Hughes

@fuserleer

76 Following

21 Followers

Dan Hughes pfp

There's also the case that a quorum isn't found and block+N can not achieve finality. In that case you might want to include the votes on chain as a fall back measure (perhaps aggregated). In a healthy network that should be rare so optimizing for the honest majority case has net-benefit.

0 reply

0 recast

0 reaction

Dan Hughes pfp

In a non-compromised network those "bad votes" will be seen by honest actors who can embed them in blocks they generate (and also broadcast to other validators). Worst case, "bad votes" are happening and no one is seeing them, therefore the network is compromised, probably by a strong adversary. Mandating all votes are recorded on-chain gives very little, as the adversary likely has influence of the network, preventing the votes from being witnessed, or can censoring blocks including them.

1 reply

0 recast

0 reaction

Dan Hughes pfp

Vitalik stated the reason is to be able to punish/reward. But punishments and rewards are not real-time. You do need to record "bad votes" on chain as evidence which new validators will need to "replay" the state and ensure punishment was just, but that evidence being deferred to block N+2 or N+3 etc doesn't cause any issue.

1 reply

0 recast

0 reaction

Dan Hughes pfp

A better solution seems to be broadcast of votes to each other via gossip, collect them, and when the 2/3rds+ threshold is met, produce a quorum certificate proving agreement. Whoever is producing the next block enshrines that quorum certificate. This is equivalent to an "all to all" messaging model, but so is recording them on chain anyway so should not be any net-detriment. For the validator who creates the next block to record the votes/signatures they have seen, they must have been communicated to them by some means anyway. TLDR if a quorum is found, you don't need to record the votes.

1 reply

0 recast

0 reaction

Dan Hughes pfp

Vitalik posted this article a few days ago regarding TTF reductions. It has a section about "Single slot finality" and how the cost of recording all the votes from all stakers/validators on chain is a technical challenge to solve. I think the approach is wrong (unless I'm missing some detail) and that the votes from all stakers don't need to be recorded on chain in all cases. https://ethresear.ch/t/orbit-ssf-solo-staking-friendly-validator-set-management-for-ssf/19928

3 replies

0 recast

1 reaction

Dan Hughes pfp

It won't be Eth because it stepped away from the best solution to enable a unified, opaque developer and user experience. State sharding.

0 reply

0 recast

0 reaction

Dan Hughes pfp

https://open.substack.com/pub/danxrd/p/why-sharding-splits-security-is-a?utm_campaign=post&utm_medium=web

1 reply

0 recast

2 reactions

Dan Hughes pfp

https://open.substack.com/pub/danxrd/p/why-sharding?utm_campaign=post&utm_medium=web

2 replies

0 recast

2 reactions

Dan Hughes pfp

2 replies

0 recast

13 reactions