4 ms·

> If any m of n stakers submit their part of the signature, a random number is created; no smaller subset of stakers can predict what that number will be, and t

> If any m of n stakers submit their part of the signature, a random number is created; no smaller subset of stakers can predict what that number will be, and the number is the same regardless of which stakers submit. so the number is known just from who is participating in the scheme? who controls the `n`? > Any staker can see the evidence that you signed conflicting histories what if there is a temporary network split? everybody who signed on chain A is now malicious actor if they sign on chain B and vice versa.

DennisP 8y ago
The number isn't known until all m submit their signatures. I think n is a protocol parameter, so the community will decide by the software they agree to run. If you sign conflicting blocks you get penalized, but "conflicting" is tightly defined. E.g. signing two different blocks at the same block height is a conflict. That's not something you can do if you're on one side of a network split, and it's certainly something you can avoid doing.
- keymone 8y ago
  > signing two different blocks at the same block height is a conflict i don't need to sign at same height to be a participant in an attack.
  - DennisP 8y ago
    That's true. There are several specific slashing conditions, and I forget the exact details, but they're all similarly well-defined and limited. There are proofs that the network converges properly as long as most stake obeys those conditions, and that's why only violations of those conditions are penalized.
    - keymone 8y ago
      > There are proofs that the network converges properly as long as people don't violate those conditions do you have a link to those proofs? i've seen proofs already where underlying assumptions were something like "majority of nodes are honest" or "network is reliable", etc.
      - DennisP 8y ago
        I'm not up to date but here's the Casper FFG paper, which says: > Accountable safety means that two conflicting checkpoints cannot both be finalized unless ≥1/3 of validators violate a slashing condition (meaning at least one third of the total deposit is lost). Plausible liveness means that, regardless of any previous events (e.g., slashing events, delayed blocks, censorship attacks, etc.), if ≥2/3 of validators follow the protocol, then it’s always possible to finalize a new checkpoint without any validator violating a slashing condition. https://arxiv.org/abs/1710.09437 https://arxiv.org/abs/1710.09437 So if more than a third of the stake violates a slashing condition, that does cause a problem, but only a temporary one since that stake is immediately destroyed. (An important improvement over just assuming majority honest nodes is that you can prove who violated any of the slashing conditions.) There's been a lot of work since that paper, much of it discussed at ethresear.ch. Casper FFG is the version currently heading for production; in the paper it's described as a finality layer adding security to proof of work. The original plan was to reduce PoW rewards by 80% once the FFG layer was added. The new design is a bit different and I think it relies less on the PoW layer, but I'm not sure how it all fits together. There's also Casper CBC, which is a more purist form of PoS they hope to implement later. That seems to be making good progress but it's still in the research phase.

DennisP 8y ago

The number isn't known until all m submit their signatures. I think n is a protocol parameter, so the community will decide by the software they agree to run. If you sign conflicting blocks you get penalized, but "conflicting" is tightly defined. E.g. signing two different blocks at the same block height is a conflict. That's not something you can do if you're on one side of a network split, and it's certainly something you can avoid doing.

keymone 8y ago
> signing two different blocks at the same block height is a conflict i don't need to sign at same height to be a participant in an attack.
- DennisP 8y ago
  That's true. There are several specific slashing conditions, and I forget the exact details, but they're all similarly well-defined and limited. There are proofs that the network converges properly as long as most stake obeys those conditions, and that's why only violations of those conditions are penalized.
  - keymone 8y ago
    > There are proofs that the network converges properly as long as people don't violate those conditions do you have a link to those proofs? i've seen proofs already where underlying assumptions were something like "majority of nodes are honest" or "network is reliable", etc.
    - DennisP 8y ago
      I'm not up to date but here's the Casper FFG paper, which says: > Accountable safety means that two conflicting checkpoints cannot both be finalized unless ≥1/3 of validators violate a slashing condition (meaning at least one third of the total deposit is lost). Plausible liveness means that, regardless of any previous events (e.g., slashing events, delayed blocks, censorship attacks, etc.), if ≥2/3 of validators follow the protocol, then it’s always possible to finalize a new checkpoint without any validator violating a slashing condition. https://arxiv.org/abs/1710.09437 https://arxiv.org/abs/1710.09437 So if more than a third of the stake violates a slashing condition, that does cause a problem, but only a temporary one since that stake is immediately destroyed. (An important improvement over just assuming majority honest nodes is that you can prove who violated any of the slashing conditions.) There's been a lot of work since that paper, much of it discussed at ethresear.ch. Casper FFG is the version currently heading for production; in the paper it's described as a finality layer adding security to proof of work. The original plan was to reduce PoW rewards by 80% once the FFG layer was added. The new design is a bit different and I think it relies less on the PoW layer, but I'm not sure how it all fits together. There's also Casper CBC, which is a more purist form of PoS they hope to implement later. That seems to be making good progress but it's still in the research phase.