You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I talked to Isaac and Chris about this. Isaac opinion was that one of the block proposer needs to be honest but can be anyone from the main chains and not only the acceptors in the overlaps of the quorums if we are ok with loosing liveness.
Chris and I however are not sure about this. Chris brought up that there will be a problem with data availability. I think there is also a problem with delaying the consensus on the main chains if we allow this. If there is a fork in the chimera chain caused by another non-overlapping acceptors from the other main chain creating a forking chimera block, only the acceptors in the overlap would know about it and blocks in A or B can build on these. It will be live (but just delayed): Heterogenous Paxos guarantees that somehow, but it will result in the validator in the quorum overlaps having to vote blocks down that everyone else thinks are ok.
Moreover, if we allow all acceptors to produce blocks networking becomes more critical.
The text was updated successfully, but these errors were encountered:
We discussed this issue in sync. While for two chains, the easiest and straightforward solution is to only allow proposers be from intersections of quorums, for three or more chain cases this gets complicated, since there will never be any acceptors who is on an intersection of all chains necessarily. Heterogeneous Paxos allows 3 or 4 chain cases if there is only pairwise intersections. The questions is is such cases what do we do for block proposers? Do we make the condition more strcit and require intersection of all chains or have a different block producing rule?
Don't we just need for there to be a block proposer with the appropriate data availability who we can somehow elect? In theory, couldn't the block proposer be part of none of the quorum as long as they have (a) data availability (b) the ability to sign messages (c) something at stake and (d) the ability to communicate with the acceptors?
(though I agree that for practical reasons we should have only acceptors be block proposers in all likelihood)
I talked to Isaac and Chris about this. Isaac opinion was that one of the block proposer needs to be honest but can be anyone from the main chains and not only the acceptors in the overlaps of the quorums if we are ok with loosing liveness.
Chris and I however are not sure about this. Chris brought up that there will be a problem with data availability. I think there is also a problem with delaying the consensus on the main chains if we allow this. If there is a fork in the chimera chain caused by another non-overlapping acceptors from the other main chain creating a forking chimera block, only the acceptors in the overlap would know about it and blocks in A or B can build on these. It will be live (but just delayed): Heterogenous Paxos guarantees that somehow, but it will result in the validator in the quorum overlaps having to vote blocks down that everyone else thinks are ok.
Moreover, if we allow all acceptors to produce blocks networking becomes more critical.
The text was updated successfully, but these errors were encountered: