In this Video we discuss:
- high level components the blockchains contain
- comparison of Celestia with Ethereum
- whether Solana is suitable for rollups
- speculation about Ethereum’s grand strategy
- profitability of some Ethereum rollups
- challenges for optimistic rollups due to Ethereum’s original design choices
- possible paths to multiple sequencers for rollups
- the role of data availability solution for rollups
- critique of sovereign rollups and based rollups
Video Notes:
Continuing the conversation about Solana vs Ethereum, and extending it to the rollups.
We separate the blockchains into 3 main components: CL (consensus layer), DAS (Data Availability Solution), and EL (Execution Layer). Execution Layer, in turn, can be viewed as having 3 main components: state, virtual machine, and state commitment scheme.
Comparisons with Celestia. Celestia provides CL+DAS. Tendermint BFT vs Ethereum CL. Celestia is ahead in terms of quantity of blobs, but this is not too hard to fix. Fixing CL is arguably harder.
Challenges of building rollups on Solana, both Solana as L1, and Solana as L2. How Solana needs to be modified to accommodate this. Two approach to work around this without changing Solana’s commitment scheme: extra blockchain with its own state commitment and blocks of Solana, or placing merkle tree inside Solana
accounts.
Ethereum’s grand strategy of becoming CL + DAS. Ethereum EL would need to be optimized for verification of proofs from potential zk
rollups. Currently, zkRollups are not as profitable compared to Optimistic rollups. Needs to investigate why zkRollups are not as profitable. Explain where the profit comes from: difference between collected fees on L2 and paid fees on L1. The big idea is that multiplicity of L2 rollups would compete with other systems for users. They pass a fraction of their revenue to L1 by paying for blobs, but a small fraction of a larger pie may be bigger than a small pie.
Ethereum L2s need to solve the problem of incomprehensible state. They will not be able to simply adopt what Ethereum is doing, and use it as a standard. They will need to create their own base technology. Current incarnation of EVM-based rollups need extra infrastructure (for example one provided by Erigon) to store history in order to compensate for incomprehensible state.
One issue worth keeping in mind is the number of sequencers in any L2 rollups. As far as I know, all running sequencers owned by themselves, and many, perhaps all, are promising to introduce multiple. How can this be done? To me, it is not as simple as going from1 to N. It is similar to going from a Web server to Bitcoin. Here having parent L1 chain can provide some help, but it’d still not without challenges. One approach is to issue staking token and bootstrap PoS network of sequencers. Technically hard. Another approach is to piggyback on L1 consensus layer. I believe this is sometimes called “sovereign rollups”. Sequencers collate batches of L2 transactions and submit them to L1 as blobs, and in whichever order L1 places them, becomes L2 order. However, two issue need to be resolved here. First, L2 sequencers have to take turns in making batches, otherwise they will end up all including all transactions into their batches. This probably already means staking, and we are getting closer to the PoS approach. Issue number two – even with taking turns, transactions in batches may overlap. I am skeptical of practicality of sovereign rollups. Therefore, we will either see endless promises of multiple sequencers, or effectively L2 rollups starting to replicate not only EL but also CL.
In the meantime, single sequencer rollups have a mitigation in place. It is the process of publishing the L2 blocks in L1 blobs, and this is where the whole rollup name comes from, I believe. I will not get into the discussion of whether this mitigation is good enough. Let us instead think of whether this mitigation will still be required if there is full CL+EL rollup implemented with multiple sequencers. I think not. Also, the whole point of being L2 (and not L1 becomes questionable). What is the point of
discussion? Perhaps multi sequencers’ setup will never happen, and what we have now is going to be a standard. Controlled sequencers and L1 blobs as a mitigation.
Someone pointed out to me the based rollups. Based rollups suggested by Justin Drake in 2023. CL needs to be changed to be able to distinguish the blobs as belong to rollups. It also needs to be changed to form blocks for each registered rollup, according to some parameters specific to that rollup. This removes the need for L2 sequencers, users of L2 pay transaction fees directly to L1. But it is not clear to me how this is to be done. Perhaps L2 users need to have tokens on L1 and sign them away when submitting L2 transaction (at least two signatures in a transaction). Or they will simultaneously make L2 to L1 withdrawal. In both cases, it almost defeats the point of havingL2 in the first place, except perhaps having a separate L2 state. It reminds me on sharing in NEAR.
Original video on the Akexey Akunov Monoblunt channel on Telegram: https://t.me/monoblunt/124
Original notes on the Akexey Akunov Monoblunt channel on Telegram: https://t.me/monoblunt/125
