Conversation with AltLayer, Scroll, and Starknet teams: Shared sequencer and L2 consensus

Talk with AltLayer, Scroll, and Starknet teams about using a joint sequencer and L2 consensus.

In the second episode of the podcast interview series with the theme of “Decentralized Rollup” curated by ECN, AltLayer Network founder Yaoqi Jia, Scroll researcher Toghrul Maharramov, and Starknet Exploration Lead Abdelhamid Bakhta were invited to explore some progress and challenges in decentralized sorters, including the importance of decentralized sorters, how to decentralize sorters, shared sorters, and the drawbacks of based rollup.

First of all, we need to know that currently, apart from Fuel V1, there is no truly decentralized Rollup, as other Rollups still have auxiliary chains. But once they are classified as Rollups, that is to say, the multisig is removed. So, the problem of decentralized sorters becomes a matter of user experience rather than security. Therefore, there are several reasons for the need for centralized sorters: firstly, L1 confirmation is very slow and we must rely on something to achieve fast transaction confirmation. However, Rollup does not provide real-time finality guarantees. Essentially, decentralization can provide us with stronger real-time censorship resistance guarantees because malicious actors need to not only destroy one or a few entities but the entire sorter network. Therefore, for us, decentralization is more of an improvement to the user experience or a solution to fix extreme cases of oracle updates, etc.

There are many functions we want to decentralize, but each has its own trade-offs. For example, introducing a certain consensus mechanism, the first is leader election, which determines who will create blocks and who will be the sorter responsible for creating blocks in a given slot or time period. Then you need a mechanism for the consensus itself, either the longest chain mechanism or BFT, or a combination of both. Therefore, we may first adopt a pragmatic approach and start with BFT. Because when considering decentralization in Layer2, our goal is not to have such a large sorter scale as in Layer1.

Whether to introduce consensus or not does not depend on whether we want to or not. Once you have many sorters or even just nodes, you have to make them reach consensus. It really depends on your assumptions. But anyway, if we have a group of sorters or validators, and we want to organize them together to produce blocks over a period of time, then you must have a consensus protocol around them. We have just launched a testnet for multi-sorter Rollup system. The difficulty lies not in the consensus protocol itself, but in things outside of it, such as the proof part. That’s why to some extent, when we launched the first multi-sorter testnet, Optimistic Rollup has some advantages in that regard. Roughly speaking, you can simplify many things, such as not considering the validity proof part.

I am not currently convinced by the proposals related to shared sorters, for several reasons: Firstly, for me, the main value proposition of shared sorters is to allow for atomic composability between generic Rollups like Scroll or Starknet. However, the problem is that if you have atomic composability, the final determinism of your Rollup is equivalent to the final determinism of the slowest Rollup you are composing with. Another downside is that running this system may be more expensive than running a system like Solana, because you have multiple full nodes running simultaneously, each with their own overhead. While it may not make sense for L2, it’s a bit different for L3. And for based rollup, I also have my concerns. For example, how do you share MEV profits with proofers? Another issue is that, since it is permissionless, multiple searchers can submit transaction batches simultaneously, which could ultimately lead to centralized results.