Summary

Order-fairness protocols add a safety property to BFT consensus that constrains how the total order is chosen, closing the gap that plain BFT leaves open to adversarial reordering (the root of much MEV). The dominant practical definition is γ-batch-order-fairness (batch-OF) (Kelkar et al., Themis): if a γ-fraction of replicas locally receive tx before tx', all honest replicas must deliver tx no later than tx'. This page anchors the two 2026 fair-ordering papers in the vault and contrasts fair ordering with the encryption- and auction-based MEV mitigations elsewhere in the wiki.

Key Points

  • Why it matters: BFT guarantees total order but places no constraint on the order relative to arrival; leaders can reorder/delay for profit. Order-fairness is an alternative to Encrypted Mempools (hide content) and FOCIL: Fork-Choice Enforced Inclusion Lists (EIP-7805) / MEV Auction Design: Open vs. Sealed, Timeboost, and Kairos (force inclusion / price ordering).
  • Leader-based (Themis): retrofits batch-OF onto HotStuff with standard liveness, but the single leader must collect all local orderings and compute the global fair order — a throughput bottleneck.
  • DAG-based (FairDAG-RL, DoD): multi-proposer DAGs remove the single-leader bottleneck, but both serialize the fairness layer (DoD constructs the order graph pre-consensus, an O(b²) op on the DAG critical path; FairDAG-RL constructs it post-consensus but processes one subdag at a time), surrendering the DAG’s throughput.
  • Herring (Paper: Herring — Parallel Batch-Order-Fairness on DAG-based Blockchain Consensus) parallelizes the fairness layer’s graph construction across committed subdags via post-consensus construction + explicit missing-edge resolution piggybacked on Narwhal’s reliable broadcast — ~90% higher saturation throughput than FairDAG-RL, ~100% over DoD-W; also patches new liveness bugs it found in both.
  • The revenue↔fairness tension (Paper: The Blockchain Execution Dilemma — Optimizing Revenue XOR Fair Ordering): enforcing strict fair ordering nullifies the gains of advanced sequencing — on Sui/Ethereum data, fair-ordering constraints cut validator revenue 50–60% during congestion. Validator-discretion sequencing (genetic optimizer) instead raised profit ~15% and cleared congestion up to 58% faster. Fair ordering and revenue optimization are framed as opposite ends of a spectrum.

Tradeoffs & Controversies

  • Order-fairness protects against reordering MEV but does nothing about CEX-DEX arbitrage or off-chain information asymmetry (those persist regardless of on-chain order).
  • “Fair” arrival-time ordering can be gamed by spam/latency (cf Paper: Timing Games — Probabilistic Backrunning and Spam (Flashbots/Offchain Labs), Paper: Blockspace Under Pressure — Spam MEV on High-Throughput Blockchains) — γ-batch-OF only constrains pairs many honest replicas observed in the same relative order; the most contested pairs (Condorcet cycles) are exactly the ones it leaves to a tie-break.
  • The execution-dilemma result is a strong empirical argument against rigid fair ordering for revenue-maximizing chains, and partly explains why Ethereum pursues inclusion-list / encryption routes rather than arrival-order fairness.

See Also