Leaderless Sovereign Roll-Up

Here is an idea on how a sovereign roll-up can have permissionless decentralized sequencers, without relying on a leader schedule.

The idea is to have a “fork-choice rule” where the network accepts the first valid roll-up block produced within a celestia slot, ordered by how much gas the roll-up block consumed, denominated in the roll-up’s gas token. The idea here is that the roll-up’s leader selection protocol strongly favors liveness, and inherits safety from Celestia.

A diagram below:

if multiple blocks with the same height arrive within the same celestia block, accept the one with the higher fee revenue. if one arrives a slot late, reject it.

Some attacker could produce blocks that censor, but it would be expensive to continue such an attack for prolonged durations, as he would have to burn a lot of gas to continue winning the fork-choice rule. I assume a EIP-1559 fee market for the roll-up, and maybe something like a flat inflationary reward for the roll-up block producers.

7 Likes

I think this works in a pessimistic or ZK rollup system, but with a fraud proof based system I suppose the problem is that if anyone can submit state commitments, then they can force many fraud proofs to be generated.

2 Likes

can’t this problem be addressed by adding a simple requirement that block producers must also lock up some amount of rollup token for longer than the fraud-proof window?

Only the canonical winning roll-up block needs fraud proofs. There’s no reason for watch dogs to send fraud proofs for the losing / invalid ones

to determine the canonical block, each block must be executed to meter gas- which could be expensive. Here’s how to solve it:

  1. Put the gas used in the block header
  2. fraud proofs for when proposers lie about gas used.
  3. repeat until a winning valid block is made known to the light clients.

(OR skip that and use validity proofs)

If a fraud proof for lying about gas requires full execution of the block, then bonding and slashing becomes a requirement for these fraud proofs to deter fraud and false alarms.

We can expect the Celestia block producers to always front-run these block proposals, the rational strategy. They’ll likely bully other block builders off the chain, monopolizing roll-up block production. So what if the Celestia block producers are bad at building roll-up blocks? One solution here is to adjust the inflationary roll-up block reward according to how much gas is used in the block, to encourage the validators to pack 'em tight- this may lead to an outsourced PBS or flashbots-style, for the roll-up. If it’s inevitable, might as well accelerate it.

I previously suggested slashing non-winning proposers, to accelerate the dominance of celestia BPs in roll-up block building, but that would be bad as it may discourage liveness in the event that Celestia BPs stop proposing blocks for the roll-up.

This scheme continues to incentivize the optimal blocks for gas usage, bullish for the roll-up’s health and discourages censorship.

to deal with front-running on celestia: sort by stake weight in the fork-choice rule.

When there are two valid proposed blocks with equal gas, choose the one produced by a higher-staked proposer, denominated in the roll-up token. the celestia validators could still drop them and front-run them though. so it’s possible we still need PBS, and this may unfortunately be the case for any leaderless roll-up scheme.

Fraud proofs for gas useage are costly… but could the fraud proofs be SNARK’d?

Validity proofs are costly to produce, but in this case, they only need to be produced in the worst-case scenario