A Simple Bivalency Proof that t-Resilient Consensus Requires t+1 Rounds

Abstract

We use a straightforward bivalency argument borrowed from [FLP85] to show that in a synchronous system with up to t crash failures solving consensus requires at least t+1 rounds. The proof is simpler and more intuitive than the traditional one: It uses an easy forward induction rather than a more complex backward induction which needs the induction hypothesis several times.