MegaETH

Proposers submit state roots as children of any (possibly unresolved) previous state root proposal, by calling the propose() function in the KailuaTreasury. A parent state root can have multiple conflicting children, composing a tournament. Each proposer requires to lock a bond, currently set to 0.00001 ETH, that can be slashed if any proposal made by them is proven incorrect via a fault proof or a conflicting validity proof. The bond can be withdrawn once the proposer has no more pending proposals that need to be resolved and was not eliminated.

Proposals consist of a state root and a reference to their parent and implicitly challenge any sibling proposals who have the same parent. A proposal asserts that the proposed state root constitutes a valid state transition from the parent’s state root. To offer efficient zk fault proofs, each proposal must include 3600 intermediate state commitments, each spanning 1 L2 blocks.

Proposals target sequential tournament epochs of currently 3600 * 1 L2 blocks. A tournament with a resolved parent tournament, a single child- and no conflicting sibling proposals can be resolved after 7d.

The Vanguard is a privileged actor who can always make the first child proposal on a parent state root. They can, in the worst case, delay each tournament for up to 36558901084y 8mo by not making this first proposal. Sibling proposals made after the Vanguard’s initial one or after the 36558901084y 8mo vanguardAdvantage in each tournament are permissionless.

Any conflicting sibling proposals within a tournament that are made within the 7d challenge period of a proposal they are challenging, delay resolving the tournament until sufficient ZK proofs are published to leave one single tournament survivor.

In the tree of proposed state roots, each parent node can have multiple children. These children are indirectly challenging each other in a tournament, which can only be resolved if but a single child survives. A state root can be resolved if it is the only remaining proposal due to any combination of the following elimination methods:

1. the proposal’s challenge period of 7d has ended before a conflicting proposal was made
2. the proposal is proven correct with a full validity proof (invalidates all conflicting proposals)
3. a conflicting sibling proposal is proven faulty

Proving any of the 3600 intermediate state commitments in a proposal faulty invalidates the entire proposal. Proving a proposal valid invalidates all conflicting siblings. Pruning of a tournament’s children happens strictly chronologically, which guarantees that the first faulty proposal of a given proposer is always pruned first. When pruned, an invalid proposal leads to the elimination of its proposer, which invalidates all their subsequent proposals, slashes their bond, and disallows future proposals by the same address. A slashed bond is transferred to an address chosen by the prover who caused the slashing.

A single remaining child in a tournament can be ‘resolved’ and will be finalized and usable for withdrawals after an execution delay of 3d 12h (time for the Guardian to manually blacklist malicious state roots).

Validity proofs and fault proofs both must be accompanied by a ZK proof that ensures that the new state was derived by correctly applying a series of valid user transactions to the previous state. These proofs are then verified on Ethereum by a smart contract.

The Kailua state validation system is primarily optimistically resolved, so no validity proofs are required in the happy case. But two different zk proofs on unresolved state roots are possible and permissionless: The proveValidity() function proves a state root proposal’s full validity, automatically invalidating all conflicting sibling proposals. proveOutputFault() allows any actor to eliminate a state root proposal for which they can prove that any of the 3600 intermediate state transitions in the proposal are not correct. Both are zk proofs of validity, although one is used as an efficient fault proof to invalidate a single conflicting state transition.