A Based Thesis

co-written with arixon. inspired by justin. forked from charlie and dan. thanks to barnabe, mike, and justin for reading drafts of this.

On Ethereum L1, all applications run atomically on a shared state machine. The rollup-centric roadmap sacrifices this core property in order to scale Ethereum.

The current rollup approach works well while applications remain local to the rollup. However there is a limit to the number of applications each of these rollups can support (because of inherent sequential bottlenecks), and they are not designed to talk to one another.

Today, regulatory pressure and the lack of native interoperability is driving rollups towards middleware blockchains (or rollup frameworks in the spirit of superchains/hyperchains) that allow for shared sequencing (and hence some degree of liquidity sharing and atomic composability between them).

A possible end-state here is a world in which each new L2 needs third-party middleware – a shared sequencer service – to efficiently communicate with the others.

An important – and underrated – tradeoff with this approach is that rollups no longer inherit the underlying liveness guarantees of the L1 (a big part of what makes Ethereum special) nor the full force of its credible neutrality (since rollups would rely on an alternative consensus mechanism outside of Ethereum).

Based rollups offer a different vision for a censorship-resistant future: one built around base layer neutrality and liveness as a first principle. This vision is inclusive, not competitive, towards existing rollups. Optimism and other platforms will be able to become based, without harming their business model.

What is a Based Rollup?

To recap, based (or L1-sequenced) rollups are a special subset of rollups. The sequencing of such rollups is maximally simple and inherits L1 liveness and decentralization. Moreover, based rollups are particularly economically aligned with their base L1.

A rollup is said to be based, or L1-sequenced, when its sequencing is driven by the base L1. More concretely, a based rollup is one where the next L1 proposer may, in collaboration with L1 searchers and builders, permissionlessly include the next rollup block as part of the next L1 block.

Based rollups are unique because they inherit the base layer’s liveness properties and can achieve interoperability without relying on a middleware blockchain (allowing them to meaningfully increase their credible neutrality without reducing their efficacy). These features are best explained in contrast to other rollup architectures.

Rollup architectures

Most rollups today use a centralized sequencer. The sequencer collects transactions from the mempool, batches them up, and posts them to the L1. The main advantage of this approach is that the sequencer provides users with fast preconfirmations. It also helps to mitigate risks for early-stage rollups without fraud/validity proofs, and to mitigate the risk of bugs in the proof system for those who have them. If the sequencer is operated by a trusted entity (e.g., the Optimism Foundation), the likelihood of an invalid state transition occurring is significantly reduced.

The main issue with centralized sequencers (apart from the potential for MEV abuse) is that they present a single point of failure from a liveness and censorship-resistance perspective. While current rollups provide exit hatches and forced inclusion to safeguard against sequencer downtime and censorship, realistically, this won’t benefit a significant percentage of L2 users, who can’t be expected to spend a substantial amount on L1 transactions. Another potential issue is that if users are forced to use exit hatches then the network effects of that rollup reset. It’s also relatively easy for a powerful government or regulator to impose KYC or sanctions requirements on the chain through the sequencer.

Shared sequencers aim to address many of the issues associated with centralized sequencers, such as enabling interoperability between rollup ecosystems and enhancing decentralization: Espresso Systems and Astria are teams working on this approach. A nice aspect of the shared sequencer design is that almost all current rollups can implement this architecture, no matter if optimistic or zk. The pitch is that rollups who adopt this design will possess the ability to atomically compose with one another while maintaining a higher level of decentralization compared to a centrally sequenced rollup.

One downside with the external shared sequencer model is that rollups do not inherit the base layer’s liveness properties (an underrated factor of censorship-resistance). Another downside is that it will likely require its own token at some point (or else need to engage in an opinionated form of mev-extraction to be profitable), which means that the rollups that rely on it will, in all likelihood, be less economically aligned with the base layer.

A based rollup directly leverages L1 proposers as shared sequencers without depending on the external consensus of a shared sequencer system like HotShot for Espresso (and the intermediary token and/or mev-policy that comes with it). As such, it inherits more of the base layer’s neutrality.

Interoperability

By leveraging the base layer’s builders and proposers, based rollups are able to preserve interoperability between rollups, whose batches are submitted in the same block, without the need for any additional middleware.

Fast preconfirmations (on the order of 100ms) are trivial with centralised sequencing, and achievable with an external PoS consensus. Fast preconfirmations with L1 sequencing can be achieved by leveraging EigenLayer, inclusion lists, SSLE, and mev-boost.

Simplicity

Based sequencing is maximally simple; significantly simpler than even centralised sequencing (although based preconfirmations do introduce some complexity). Based sequencing requires no sequencer signature verification, no escape hatch, and no external PoS consensus.

Based sequencing (without preconfs) is working on testnets today. The first based rollup Taiko, is preparing for mainnet, and expects to go live in Q1 2024.

The Based Rollup Pitch

One of Ethereum’s superpowers, and key differentiator compared to Solana or Cosmos BFT chains, is its ability to self-heal after stalling (a direct consequence of its liveness guarantees). This emphasis on dynamic availability allows the base layer to be extremely resilient and to thrive even in a highly adversarial environment – WWIII resistance is in fact an explicit design goal.

While the prevailing wisdom is that force-inclusion designs allow rollups to leverage the L1’s liveness, the reality is that non-based rollups suffer degraded liveness (even with escape hatches).

Compared to based rollups, non-based rollups have weaker settlement guarantees (transactions have to wait a timeout period before guaranteed settlement), are liable to toxic MEV (from short-term sequencer censorship during the timeout period), and often require users to incur a time and gas penalty to exit (because of suboptimal non-batched transaction data compression).

As a consequence, they run the risk of their network effects resetting in response to a mass exit triggered by a sequencer liveness failure – for example a 51% attack on a decentralised PoS sequencing mechanism.

The main idea behind based rollups is to use L1 proposer-builder separation to include L2 blobs (including any compression) natively rather than using a sequencer. From this perspective, they inherit whatever the L1 has to offer.

The initial Arbitrum implementation was a based rollup. The sequencer was only introduced later because of user demand for faster transactions. Based preconfirmations resolve this tension. Once EigenLayer, inclusion lists, and SSLE go live (longer proposer lookaheads), based rollups will be able to inherit the L1’s liveness and censorship-resistant properties without compromising on user experience.

This vision is inclusive and not competitive to existing rollups and their revenue models. In particular, based rollups retain the option for revenue from L2 congestion fees (e.g. L2 base fees in the style of EIP-1559) despite potentially sacrificing some MEV income.

Based rollups also retain the option for sovereignty despite delegating sequencing to the L1. A based rollup can have a governance token, can charge base fees, and can use proceeds of such base fees as it sees fit (for example to fund public goods in the spirit of Optimism).

Advantages of Based Rollups

1. Lower costs: Based sequencing enjoys zero gas overhead. There is no need to even verify signatures from centralised or decentralised sequencers. The simplicity of based sequencing reduces development costs, shrinking time to market, and collapsing the surface area for sequencing and escape hatch bugs.
2. Economic alignment: There is no need to use a middleware solution with its own token. Additionally, MEV originating from based rollups naturally flows to the base L1. These flows strengthen L1 economic security and, in the case of MEV burn, improve the economic scarcity of the L1 native token. This tight economic alignment with the L1 may help based rollups build legitimacy.
3. Better neutrality and liveness: Based sequencing inherits the decentralisation of the L1 and naturally reuses L1 searcher-builder-proposer infrastructure (which makes it more credibly neutral). Based sequencing enjoys the same liveness guarantees as the L1 (in contrast to non-based rollups who suffer degraded liveness).
4. Simplicity: Based sequencing is maximally simple; significantly simpler than even centralised sequencing. Based sequencing requires no sequencer signature verification, no escape hatch, no token, and no external PoS consensus.

Disadvantages of Based Rollups

1. Loss of MEV revenue: Based rollups forgo MEV to the L1, limiting their revenue to base fees.
   • Counterargument 1: MEV is a small fraction of rollup revenue today compared to congestion fees. It’s reasonable to imagine that this continues to hold true going forward (as apps become more mev-aware and mitigation techniques like threshold encryption become more widespread). For the same reasons, it’s also possible that MEV revenue decreases going forward.
   • Counterargument 2: On the whole, becoming based may in fact increase overall income for rollups. The rollup landscape is plausibly winner-take-most (due to the strong network effects of synchronous composability) and the winning rollup may leverage the improved security, decentralisation, simplicity, and alignment of based rollups to achieve dominance and ultimately maximise revenue.
2. More difficult to share costs between rollups: External shared sequencing gives you “for free” the ability to cost-share data posting e.g., buy a single blob to house the data from two rollups, which reduces costs vs buying two separate blobs.
   • Counterargument: It’s possible that the L1 proposer could cost-share between all the based rollups it sequences for. Taking this a step further, it’s also possible for the L1 proposer to cost-share with other services including shared sequencers.
3. Throughput still limited by the L1: All rollups share a single and throughput limited Ethereum L1. Not all of them can achieve their tps concurrently, since the data is onchain.
   • Counterargument 1: This is true for all rollups. If based rollups gain significant traction, then the L1 will, to a certain extent, evolve to match the requirements of based rollups.
   • Counterargument 2: 4844 (expected Q1 2024) will decouple DA pricing from execution layer competition. Ethereum with Danksharding can scale its DA throughput as high as internet bandwidth permits.
4. Shared state machine still better for atomicity: You still don’t quite get the same atomicity guarantees as transacting on a shared state machine (e.g., entirely on Ethereum L1) since the proposer is not the one executing the transactions.
   • Counterargument 1: Atomic async composability is overrated.
   • Counterargument 2: If atomicity does prove desirable in a async context then @EspressoSystems/SharedSequencing#Cryptographically-guaranteed-cross-rollup-atomic-bundles">atomic bundles + @EspressoSystems/SharedSequencing#Cryptoeconomically-assured-cross-rollup-atomicity">cryptoeconomic assurances are probably good enough for the majority of usecases.
5. Based preconfs bring additional trust assumptions: Since non-preconf transactions (from within the rollup) are queued until the next preconf slot is commited, unless 100% of validators are also engaged as preconfirmers the liveness guarantees of a based rollup with based preconfs are strictly worse than that of a based rollup that doesn’t use preconfs (and therefore stricly worse than the base layer).
   • Counterargument 1: The difference should, in practice, be negligble. Based preconfs only begin to work if you have 20-30% of the validator set engaged as preconfirmers since sufficient L1 validators must be preconfers to have at least one preconfer in the lookahead with high probability. Today, the beacon chain has at least 32 proposers in the lookahead. This means that if 20% of validators are preconfers there will be a preconfer with probability at least 1 - (1 - 20%)^32 ≈ 99.92%. If 30% of validators are preconfers then this increases to 1 - (1 - 30%)^32 ≈ 99.999%. If you’re concerned with re-orgs or preconfers randomly dropping offline, you can do the math on there being at least 2 or 3 (or n) preconfers in the lookahead for any given % of validators engaging as preconfers. There is a %, much lower than 100, for which the difference in liveness guarantees is negligible (though that precise number might differ depending on who you talk to). This is a world apart from relying on an external consensus for liveness.
   • Counterargument 2: SSLE (Single Secret Leader Elections) will allow for dramatically increasing the lookahead (e.g. to 1024 slots), which effectively removes this concern. Note that under SSLE preconfers can advertise (offchain and onchain) zero-knowledge proofs they are preconfers at their respective slots without revealing further information about their validator pubkey.
   • Counterargument 3: While SSLE fixes this, we don’t even need to wait for SSLE, since we can actually increase proposer lookaheads independently. And it is much easier to do so.

In Summary

Rollup protocol design is nebulous. There is no “correct” level of decentralization or security. Qualities like censorship-resistance cannot be exhaustively defined.

Today, rollups are pushed towards adopting blockchain middleware with external consensus in order to decentralize their sequencing and improve interoperability across domains. Based rollups offer a simpler, more neutral, and more economically-aligned alternative.

Based rollups with fast preconfirmations test the hypothesis that application developers (and their users) care about fully leveraging Ethereum’s liveness and credible neutrality superpowers if they can do so in a way that doesn’t require them to sacrifice efficacy (in this case confirmation speed).

With based preconfs, the user-experience tradeoffs dissolve.

Disclaimer：

1. This article is reprinted from [@sacha/based-rollup-thesis">hackmd]. All copyrights belong to the original author [sacha ]. If there are objections to this reprint, please contact the Gate Learn team, and they will handle it promptly.
2. Liability Disclaimer: The views and opinions expressed in this article are solely those of the author and do not constitute any investment advice.
3. Translations of the article into other languages are done by the Gate Learn team. Unless mentioned, copying, distributing, or plagiarizing the translated articles is prohibited.