Being a Stage 2 Rollup from Day 1: Etherlink’s Journey

My name is Thomas Letan. I’m working for a company called Nomadic Labs Nomadic Labs is one of the largest R&D centers working on the Tezos ecosystem. , and today I am here to talk about Etherlink . More specifically, I will talk about our journey to be a stage 2 rollup from day 1.

What is Etherlink?

Before going any further, it is probably worth giving a bit of an introduction for Etherlink, because you might not already be aware of what it is already. Etherlink is a newcomer in the Layer 2 landscape, and has the particularity of being part of the Tezos  ecosystem. That is, it is not an Ethereum rollup. Our Mainnet has been available in Beta since May 15.

Etherlink has been built with three core properties in mind: we wanted it to be fast (with a 500ms soft confirmation time delivered by a sequencer), fair (as we will discuss during this talk), and nearly free (your typical transaction costing one-tenth of a cent).

Our main focus during Etherlink’s beta is onboarding new partners. We have already onboarded several infrastructure partners, and our immediate focus now is on bringing dApps to the chain.

Etherlink is (nearly) a Stage 2 Rollup

Besides, what we think is interesting about Etherlink is that it is already (nearly) what L2beat  calls a stage 2 rollup , and it has been since its genesis block.

To support this claim, let's go back to the definition of what L2beat stages are, and what a stage 2 rollup is about.

• Etherlink naturally fully qualifies as a stage 0 rollup. Every transaction is posted on Layer 1 (Tezos in this case), along with L2 state root hashes. Besides, you can recompute the state of the chain locally using free and open source software (released as part of the Octez  suite).
• Etherlink fully qualifies as a stage 1 optimistic rollup as well. We have the full implementation of the optimistic rollup infrastructure, including a trustless bridge and a fraud-proof system.
• Because this fraud proof system is permissionless (and has actually been permissionless since the genesis of Etherlink), we also meet a crucial requirement of stage 2 rollups. We are missing a 30-day exit window for users wanting to leave their funds out of Etherlink to fully qualify as a stage 2 rollup. This is because we favor a shorter feedback loop with our early adopters during the course of our betaWhile the exit window is currently 24 hours, it is already fully permissionless. Users do not need the help or authorization of a third-party to remove funds from Etherlink. .

Etherlink is a Non-Custodial Rollup

Meeting the main requirements of being a stage 2 rollup has been a direct consequence of our primary objective for Etherlink: to be a non-custodial rollup. That is, it was always clear to us that no closed set of third-parties should ever be granted exclusive, irrevocable rights on Etherlink. In other words, we wanted Etherlink to adhere to the same standards as Layer 1 blockchains, like Ethereum or Tezos.

To that end, we built Etherlink on top of two cornerstones: a permissionless fraud proof system, and a decentralized on-chain governance. My goal for the remainder of this talk is to give you a peek into the kind of solutions we came up in these two areas.

Etherlink is Powered by a Permissionless Optimistic Rollup

I was at EthCC last year to introduce Tezos optimistic rollup stack . We won’t revisit the details of our solution in depth, but let’s use this opportunity to give you a refresher.

Optimistic Rollup in a Nutshell

We are in the rollup track, so we can skip the 101 course on optimistic rollup. To me, the three points you need to keep in mind from now on is that an optimistic rollup

1. offloads computations away from a layer 1 blockchain in a new layer
2. enables bi-directional communications between the two layers with a trustless bridge
3. relies on an interactive fraud proof system to ensure the safety of your assets while they are locked in this second layer.

You cannot remove one of these properties without making an optimistic rollup either unsafe or useless.

Smart Rollups are Optimistic Rollups as a Common Good

Etherlink is powered by Smart Rollups, a one-year-old feature of the Tezos protocol. You can think of Smart Rollups as a general-purpose framework for building optimistic rollups. I like to think that we have implemented the full optimistic rollup infrastructure once (and —to the best of our knowledge— correctly), and now every member of the Tezos community can benefit from this effort.

This is not the first time I’m tasked to give an introduction to Smart Rollups, and I like to put forward the following properties:

• You have understood it by now, but Smart Rollups are permissionless by default. By default, any XTZ holdersXTZ is the native token of Tezos, that is used to pay gas, validator fees, and participate in Tezos governance (more on that later). can post and defend L2 state root hashesIt is also possible to create permissioned Smart Rollups, for use cases that require it. This is not the case of Etherlink, obviously. .
• Smart Rollups are not implemented by means of smart contracts. They are directly enshrined in Tezos protocol, making them a first-class citizen of the Layer 1 blockchain. This was instrumental in enabling several optimizations to increase performances and reduce end-users costs.
• They are fully programmable. Smart Rollups are not about creating EVM-compatible Layer 2 blockchains, they are way more general than that. The same way you provide a script when creating a smart contract, Smart Rollups come with a (WASM) program that decides the semantics ruling the newly created Layer 2 rollup.

Etherlink is not the only use-case for Smart Rollups! To name a few examples,

• We have been able to demonstrate Tezos can support one million transactions per second on a public testnet using 1,000 smart rollups  last year.
• We are working on a Javascript runtime (called jstz ) for Layer 2 blockchains powered by a Smart Rollup.
• Engineers from Plenty , a DeFi platform deployed on Tezos, have been experimenting with Plenty Rollup  to exchange assets with low-latency and high-throughput.

Etherlink Decentralized Governance

Smart Rollups are the technological foundation of Etherlink, but they are not enough to make it a non-custodial rollup. The same way you can create a centralized smart contract on top of a decentralized blockchain, you can definitely build a custodial Smart Rollup by embedding a public key in its code.

This brings us to decentralized governance.

If you were aware of Tezos before this talk, it is probably because of its on-chain governance system. If you don’t know, Tezos is a Layer 1 blockchain officially launched in 2018, with a track record of 16 forkless upgrades. This is possible because the governance of Tezos is embedded in its protocol. We definitely wanted to leverage this when we started building Etherlink.

Etherlink (Absence of) Governance Token

It was clear from the get-go that Etherlink would “inherit” from Tezos strong and vivid governance. One consequence of this choice was to XTZ as its governance token, instead of introducing a new token. Besides, Tezos governance is larger than XTZ itself, because it is built on top of a pioneer liquid proof of stake mechanism. For Tezos upgrades, only the validators vote, but their ballot’s weight is determined not only by their personal stake but also (and sometimes mostly) by the balances of individual holders who have “delegated” their XTZ to themIt is clearly out of the scope of this talk, but for your information, Tezos liquid proof of stake has been significantly amended with the Paris upgrade to include a new staking mechanism  which complements the existing delegation approach. .

This should give Etherlink a head start, because the ecosystem of Tezos has a clear and well-established maturity when it comes to on-chain governance.

Handing over Etherlink to Tezos Validators

Similarly to Tezos, Etherlink governance happens on-chain, and once the vote is finished, there is nothing we can do to prevent its consequences from becoming a reality. Everything is embedded in Etherlink program.

The voting process is a lightweight revisit of Tezos’ one. The governance of Etherlink is carried through the Layer 1 blockchain, because this is the place where the voting power of the governance body (Tezos validators) is determined.

The vote proceeds as follows.

Firstly, any Tezos validator can propose a new upgrade during the proposal period. I believe this is a significant rule to highlight. Proposing a candidate is not a right limited to —let’s say— Etherlink core developers. On that, we follow the footsteps of Tezos, and what happens on Tezos during the proposal period is often quite interesting. We have seen time and time again the community using this opportunity to engage very actively with the governance process. For the Ithaca proposal for instance, a validator had submitted a counter-proposal disabling a feature they didn’t like. For the Oxford proposal, validators proposed a variant of the proposal tweaking some parameters of the key feature introduced at the time. This is “governance at work,” and the governance body can then select the proposal they find the best fit for the Tezos network.

Secondly, the promotion period starts with the proposal that has received the most upvotes. Obviously, winning an election among many participants does not ensure to the victor a majority of supportFor context, I gave this talk in the midst of France almost seeing a far-right government taking power, and the left “winning” a snap election with too few parliament members to apply their program. . The question asked to validators boils down to: “Do you approve this proposal?” The goal is to verify that the selected candidate can gather a supermajority of support. This additional step was designed to alleviate the risk to see unhappy validators fork Tezos.

Finally, the cooldown period gives an opportunity to (1) the ecosystem to adapt to the features and changes brought by the new software, and to (2) dissatisfied users to exit Etherlink.

The current parameters for the governance of Etherlink are as follows.

I will not spend much time discussing each number individually, and instead focus on the three kind of votes.

Firstly, and this is probably what you had in mind when I started to mention decentralized governance, is voting on the next version of the software powering the chain. We distinguish between “regular” upgrades (the default case), and “security” upgrade. The latter can be used exceptionally if a bug is found in Etherlink that can endanger the assets of Etherlink users. The security upgrade has shorter periods of vote, but requires a higher quorumDid you know? Etherlink has gone through a security upgrade already . The patch deployed fixed a critical issue in the withdrawal precompiled contract, which was found by Spearbit  during their audit. . It is worth mentioning that we do not have a security committee, as it is often the case for other Layer 2 blockchains.

Secondly, and I think it is a unique thing in our industry, Etherlink’s sequencer is elected by Tezos validators. This brings a clear incentivization for the sequencer to behave correctly. Otherwise, Tezos validators can simply vote them out of Etherlink. It does without saying —so it’s better when reaffirmed— that the sequencer power is limited to proposing blocks, the validation of said blocks remains permissionless. In particular, in the context of Etherlink being a Smart Rollup, only valid transactions that are actually posted on Tezos can lead to funds transiting from Etherlink to Tezos through the permissionless bridge.

Another interesting fact about Etherlink and its relationship with the sequencer is how Tezos consensus algorithm and low block time enables Etherlink’s blocks to become final under less than a minute. Tenderbake (a variant of Tendermint) features a 2-block finality property that is incredibly rollup-friendly. Since the sequencer typically posts its block proposals directly after their creation, it is not unusual for an Etherlink block to become final in 30 seconds or so.

For partners and institutional entities that cannot rely on soft confirmation from a sequencer, we believe this can be a game changer.

Why Non-Custodial?

This talk is slowly coming to its end, and one way to conclude it is to come back to this idea that we wanted Etherlink to be non-custodial from the get-go. This should raise an interesting question: why? Why was it so important for us to make it a priority that early in the development process?

This is particularly relevant when we take a step back and have a look at our ecosystem. To this day, the leaders of our industry are still powered by stage 1 (Arbitrum, Optimism) and stage 0 (Base) rollups.

The origin of our decision lies in our long term goal for Etherlink. Etherlink as a Layer 2 blockchain is important, and we have invested a significant amount of effort and time to make it as successful as we can, and it is also a component of a larger scalability roadmap we coined Tezos X . Tezos X aims to bring the full Tezos ecosystem into a scalable, decentralized optimistic rollup compatible not only with Michelson (the home-grown smart contract language of Tezos) and EVM, but mainstream programming languages as well (as suggested by our efforts on jstz).

In the meantime, I can only encourage you to take interest in Etherlink today. As mentioned, its Mainnet is already there and will soon move out of the beta phase. Our fast, fair and (nearly) free Layer 2 blockchain awaits you.