What's holding L2 based rollups back today?
Last updated
Last updated
Spire Labs has paused development on the Based Stack since releasing it. Two main factors limit L2 based rollups on Ethereum today, from our perspective:
Centralized sequencers are aligned with their users values because they are run by L2s that have a clear incentive to attract users with the best UX possible. Centralized sequencers choose not to frontrun or sandwich their users because they have determined that this would be bad for business.
In contrast, leads to incentive structures where validators choose to extract as much short term value as possible from users of the chain within the configurations and constraints of offchain PBS. This has led to an inherently unfriendly sequencing model today.
Don't hate the players, change the game. Ethereum researchers are exploring various mechanisms to make Ethereum sequencing more friendly. Preconfirmations are one such research direction under that umbrella, but none of it matters until it's live and starts changing things.
This is a blocker for Ethereum L2 based rollups because applying an unfriendly sequencing system to an L2 is a strictly worse product for developers and users when compared to a friendly centralized sequencer.
Two things are required for synchronous composability: coordinated sequencing and shared finality. Coordinated sequencing is achieved with the Based Stack today, but shared finality is not. Shared finality means that once an Ethereum block and a based rollup block are proposed, both must completely finalize together. This means that a based rollup that want to offer synchronous composability cannot use fraud proofs (which require a minimum finality of about 7 days) and must use validity proofs. Because these validity proofs are needed at the time of block proposal, proof generation latency is crucial.
Synchronous composability is such an important differentiator for Ethereum L2 based rollups that we consider the lack of it a major blocker.
At Spire, we call the kind of we need subsecond proving. Subsecond proving is not currently possible with popular zkVMs.
Subsecond and is possible with alternative (non-zk) models like TEEs, but not without tradeoffs. Additionally, these systems are not yet production ready.
