search

EntEthAlliance / enhanced-bft

a workspace for developing improvements to BFT consensus
Other
8 stars 3 forks source link

Optimal Byzantine Safety Threshold #7

Closed saltiniroberto closed 5 years ago

saltiniroberto commented 5 years ago

GPAv1 feedback GPAv1 states “Also note that in a related article here: https://pegasys.tech/another-day-another-consensusalgorithm-why-ibft-2-0/, it is suggested that to address a similar limitation described in Lemma 9, in proposed BFT Variant, 4 validators are required to reach agreement. Is that not a direct violation of optimal byzantine safety threshold? It seems that, in proposed BFT Variant, the threshold of honest nodes required is being increased up to ~80% as opposed to usual and optimal ~67%.”

Reply The Optimal Byzantine Safety Threshold states what the maximum number of Byzantine nodes is, not what the minimum number of honest node is. In more details, as proved in Dwork et al. (https://groups.csail.mit.edu/tds/papers/Lynch/jacm88.pdf), the maximum number of Byzantine nodes for a network of n nodes is floor((n-1)/3). So for a network of 5 validators, the maximum number of Byzantine validators is 1 which implies that we must assume that at least 4 validators are non-Byzantine.

kubasiemion commented 5 years ago

This should go into requirements

kubasiemion commented 5 years ago

close

kubasiemion commented 5 years ago

Close

kubasiemion commented 5 years ago

Closed as agreed on Aug 14th