Latency cost of censorship resistance

This post covers our new lower bound on the latency cost of censorship resistance. In a traditional BFT protocol, the leader has two roles: (1) It constructs (and therefore holds) the input; and (2) It proposes the input. Many BFT protocols optimize for the good case when the leader is honest. For partial synchrony, with $n \leq 5f-2$ parties, the good-case latency is 3 rounds. The leader’s monopoly over both... By Ittai Abraham, Yuval Efron, and Ling Ren