Localized slice
One, two, and three chains: the commit taxonomy before HotStuff
How many consecutive certifications a protocol needs before saying “this will not move”, and how that answer prepares QC, highQC, and Fast-HotStuff.
Committing is not just voting nicely
In BFT a signature is a verifiable promise: the replica claims the proposal does not violate its lock or the known safe evidence. Commit needs a mechanical rule that survives bad leaders, delayed messages, and Byzantine replicas.
One-chain and two-chain: latency traded for evidence
A one-chain rule finalizes quickly but demands very strict locks and proposal restrictions. Two chains shorten the happy path and move the cost to the view change, where aggregated evidence such as an AggQC must be carried.
- One-chain: useful as intuition, delicate under equivocation.
- Two-chain: appears in Fast-HotStuff and Jolteon; guard the view change.
Three-chain: the order of basic HotStuff
Three consecutive certified blocks commit the ancestor under one uniform rule. Every view runs the same pattern, which lets Chained HotStuff overlap phases and keep a linear view change without extra proofs.
References
[1]
Yin et al., HotStuff: BFT Consensus in the Lens of Blockchain (2019)Defines the three-chain commit rule and its chained version.
[2]
Buchman, Kwon and Milosevic, The latest gossip on BFT consensus (2018)Tendermint: locks and validRound as a two-phase alternative.
[3]
Jalalzai et al., Fast-HotStuff (IACR ePrint 2020/088)Two-chain variant that pays for the view change with aggregated evidence.