[draft] Postconfirmation protocol

[apenzk]

Summary

• MIPs: MIP-37.
• Categories: protocol-units

Introduces the role of acceptor in the MCR contract, see MIP-37.

The postconfirmer (prevs. acceptor)

• should be selected based on L1 randomness (not part of this iteration of the protocol)
• serves for a certain period
• receives rewards for service
• is replaceable by volunteer attester
• is checked for liveness

The volunteer-postconfirmer

• can postconfirm at any time
• gets only rewarded after a privilege window for the postconfirmer is over

Motiviation

Currently rewarding based on time of commitment and the complexity that arises from the last committer having higher commit cost and different rewards may cause race conditions, co-location and centralization.

This can be resolved by separating the tasks into attesters and postconfirmer, where postconfirmer is a new role that converts provided attester commitments into a postconfirmation and update the state (i.e., rollover to new epochs). The sole role of the attester becomes to commit.

As a simple first PoC the current acceptor could be selected based on "slow" randomness from L1. E.g. using hashes from L1-blocks. However, this would require that the protocol is modified to use L1 blocks instead of timestamps for epochs. This has been ommitteed for now, but as a consequence it becomes predictable who is the postconfirmer.

Changelog

• distinguishes superblocks from L1blocks: Postconfirmation: Distinguish between L1Blocks and superBlocks in the solidity contracts #1045
• epoch names [postconfirmation] epoch names #1052
• update stake related variable names StakeForAcceptingEpoch and CustodianStake #1060
• adds acceptor role [draft] update MCR to include an acceptor role #713
• Rewards for the Attesters as long as they attested before the postconfirmation [postconfirmation] attester rewards #1074
• Required delay added between first honest attestation and postconfirmation [postconfirmation] add minimum delay between postconfirmation and honest commit #1076
• Rename Acceptor to Postconfirmer, Postconfirmer rewards [postconfirmation] postconfirmer rewards #1077

Testing

forge test ok
cargo check ok
cargo test fail

Outstanding issues

• using time for acceptor term instead of blocks requires to use pre-calculatable randomness, while this enables rotation it does not make it unpredictable. (this will be fine for now but should be revisited later)