Consensus: Parlia PoSA + BLS Fast Finality

New L1’s consensus is derived from Parlia, the Proof-of-Staked-Authority protocol securing BNB Smart Chain. Parlia has run in production for years, through several rounds of block-time reduction, and already has the properties New L1’s latency targets need. New L1 takes it with tuned parameters and a handful of deliberate divergences, stripped of BSC’s accumulated hardfork gating: the chain runs one configuration from genesis, with no activation heights and no feature flags.

Parameters

Parameter	Value
Block interval	200 ms
Finality	sub-second, via cross-block BLS votes
Proposer turn length	16 blocks
Validator set	up to 64 validators, PoSA

Block Production

Validators take turns proposing, rotating every 16 blocks in sorted order. If the in-turn proposer misses its slot, out-of-turn validators step in after a randomized backoff, with the delay schedule seeded deterministically so every node computes the same order. In-turn blocks carry more weight than out-of-turn blocks, so the chain converges back to the schedule as soon as the in-turn proposer returns.

Fast Finality

Finality comes from a BLS vote overlay, the same construction BSC uses, with one property that matters more here than anywhere else: validators sign a block’s ordering, not its execution result. That is what makes order-execution decoupling possible, and consensus never waits for the EVM.

The flow:

1. Each validator signs a vote for the head block on every canonical update.
2. The next proposer aggregates votes into a single BLS attestation once ⌈2N/3⌉ validators have signed, and embeds it in its block header. Finality proofs travel with the chain itself, so a node that has the headers has the proofs.
3. Justified and finalized heights are derived from the attestations using Casper-FFG rules: an attestation whose source and target are consecutive finalizes the source. There is no imperative “mark finalized” step anywhere: every node reads the same finality state out of the same chain.

In the common case a block is justified one block later and finalized two blocks later, roughly half a second at the 200 ms interval.

Finality also backstops asynchronous execution: a validator abstains from voting on any block whose published ExecutionCommitment its own local execution disproves, so an incorrect execution result cannot gather the ⌈2N/3⌉ votes it needs to finalize.

Validator Set

Validators stake BNB through the on-chain staking system (StakeHub, see Governance). Elections run once per day at a designated block, ranking candidates by stake; epoch-boundary headers bind the elected set, and every node validates against the set the chain itself recorded.

Double-sign evidence is detected automatically and submitted on-chain for slashing; each validator also keeps a local journal that makes it refuse to sign conflicting votes or blocks in the first place.

For Developers

• Validator admission and rotation are on-chain and queryable (see newl1_getValidatorSchedule).
• Block interval, turn length, and epoch length are chain parameters, not assumptions to hardcode into client code.
• For a bridge release or another action that depends on successful execution, wait until a later block records the result and finalizes, then verify the transaction receipt. Finality of transaction order alone does not confirm that execution succeeded. See Transaction Pre-confirmation for the faster, best-effort signal available before finality.

What’s Next

Parlia is the current default rather than a permanent commitment. The properties New L1 actually depends on are a fixed 200 ms cadence, ordering-only votes, and sub-second finality. Any consensus that preserves those can slot in behind the same interfaces. Next-generation designs are under evaluation, in particular Minimmit, a minimal BFT protocol that reaches finality in fewer communication rounds, as a candidate for Consensus 2.0. The consensus layer sits behind narrow seams in the node architecture precisely so that this migration, if and when it comes, is a component swap rather than a rewrite.