minimmit-primary-sources-2026-05-14

Minimmit primary sources

Reviewed: 2026-05-14 UTC

Formal paper

• https://arxiv.org/abs/2508.10862
  • Best canonical source for the protocol as a distinct consensus design rather than just a Commonware blog post.
  • Used to confirm the main framing: Minimmit is a Byzantine-fault-tolerant SMR protocol that separates view progression from transaction finality through different quorum thresholds under the 5f + 1 <= n assumption.
  • Also used to capture the paper’s headline empirical claim: in simulated globally distributed 50-processor settings, the design reduced view latency by 23.1% and transaction latency by 10.7% versus the compared state of the art.

Commonware explanatory post

• https://commonware.xyz/blogs/minimmit
  • Best source for why the protocol exists and what tradeoff its authors think matters.
  • Used to confirm that Minimmit is explicitly presented as optimizing for block/view latency rather than for stronger fallback robustness at higher fault assumptions.
  • Also useful because it states concrete target latencies in Alto-like and regionally biased deployments and openly notes that the protocol was not yet fully implemented at release.

Repository specification

• https://github.com/commonwarexyz/monorepo/tree/main/pipeline/minimmit
• https://raw.githubusercontent.com/commonwarexyz/monorepo/main/pipeline/minimmit/minimmit.md
  • Best source for concrete protocol mechanics.
  • Used to confirm the M = 2f + 1 versus L = n - f threshold split, the message types (propose, notarize, nullify, notarization, nullification), parent-selection rules, and the nullify by contradiction path.
  • Also useful because it exposes the extension surface directly: threshold signatures, erasure-coded dissemination, fast-skipping offline leaders, rebroadcast logic, and equivocation slashing.

Main takeaways from this pass

• Minimmit is worth keeping separate from Commonware because the protocol itself is a reusable comparison point for low-latency BFT design, not just an example buried inside a broader library ecosystem.
• The most important reusable mechanism is the split between small progress proofs and larger finalization quorums. That is the real design move to carry into comparisons with HotStuff-2, HotShot, Kudzu, Alpenglow, and ChonkyBFT.
• The main caveat is maturity: this is best treated as a paper/spec mechanism page, not as evidence of a production network or battle-tested deployed consensus stack.