Can Tezos Become EVM Compatible?

My name is Thomas Letan. I work 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 our work to bring EVM compatibility to the Tezos ecosystem.

What is Tezos?

Tezos is a Layer 1 blockchain that has been around for over five years. One of Tezos’ key differentiator —at least initially— is its on-chain governance system. The governance system is a voting procedure to propose, select, and automatically deploy core upgrades for the blockchain. This governance system enabled the Tezos network to seamlessly go through 14 upgrades since the beginning of our Mainnet. These upgrades allowed us to keep innovating at a strong pace, and today Tezos notably features its own smart contract language and runtime , a 2-block finality consensus algorithm called Tenderbake  since April 2022Tezos was initially deployed with a different consensus algorithm, which means we have hotswapped Tezos consensus while it was live. , and a first-class citizen framework to deploy and operate so-called Smart Rollups .

Smart Rollups are optimistic rollups enshrined in their Layer 1 blockchain (Tezos). We are today at EthCC[6] because we believe they can be a game-changer for Tezos, and we are excited to share this with you.

The starting point of Smart Rollups was the scaling roadmap of Tezos. It has been our main focus for the past months, leading us to demonstrate on a public testnet how 1,000 rollups can process 1,000,000 transactions per second together. But Smart Rollups as they came to be also enable us to explore brand new use cases. One of them is EVM compatibility, something Tezos itself does not provide to our users.

During the remainder of this talk, I will provide you all the keys to get a grasp on what Smart Rollups are, and how we intend to bring EVM Compatibility to our ecosystem through one Smart Rollup.

A Primer on Smart Rollups

The selling pitch for Smart Rollups is to present them as new ways to implement decentralized applications. They offer a trade-off that is slightly different than smart contracts already propose. In a nutshell, Smart Rollups get more computational power compared to a smart contract, but since these computations are offloaded from the Layer 1 blockchain, you need to bring your own infrastructure to carry it.

Optimistic Rollups Done Once

Under the hood, Smart Rollups are an optimistic rollup framework with three critical properties.

• They are permissionless by default, meaning that any Tez holderTez is the native token of Tezos. can create a rollup, and compute, publish and defend the state of any rollup created on Tezos. That is, we have permissionless interactive fraud proofs already!
• They are enshrined. A Smart Rollup on Tezos does not rely on a collection of smart contracts, but is instead of first-class citizen of the Tezos protocol itself with a dedicated address. Optimistic rollup maintenance operations are dedicated operations of Tezos, optimized for performances and gas efficiency. Besides, they benefit from Tezos being able to evolve and improve over upgradesIt’s more than a promise already. Smart Rollups have been enabled on Tezos Mainnet on March 2023 with the Mumbai upgrade. Since then, Nairobi has been activated and has improved the developer experience for every Smart Rollup developers. .
• They are programmable. The same way you create a smart contract by providing a script, Smart Rollup are provisioned with a so-called kernel (a WASM program) when they are created, and this program will dictate the semantics of the rollup. This is the key property that makes Smart Rollups a solid contendent for addressing a wild range of use cases, from generalized Layer 2 blockchains to specialized AppChains.

The Smart Rollup Runtime

But what is exactly a Smart Rollup, from a developer perspective? The best way to answer this question is to provide a high-level description of the runtime used to execute its kernel.

The first component a kernel can interact with is the shared inbox. The shared inbox contains messages coming from the Layer 1 blockchainI want to insist on that: the inbox is shared in the sense that all the smart rollups deployed on the Layer 1 blockchain have access to all messages sent at a given time. . These messages can be sent by wallet-owned accounts or smart contracts alike, but they they must compete for Layer 1 block space to do so. This is fine until it is not —in case of significant adoption for instance— so we equipped Smart Rollups with an additional way to import data from the outside world: the so-called reveal channel. At its core, the reveal channel is a primitive which allows a kernel to retrieve an arbitrary data, provided it knows the data’s hash. Although a smart rollup cannot guess a hash out of thin air, the competition for block space becomes a lot less fierce when the competitors only need 32 bytes per block each.

Similar to smart contracts’ storage, a kernel can read from and write to a “durable storage.” The key difference between the two is how a kernel does not pay for bytes allocated in the durable storage. The rationale is the same as for execution time: since everything happens offchain, the cost for allocating more Layer 2 state is higher requirement for the hardware infrastructure.

A kernel can send back messages to wallet-owned address and smart contracts of the Layer 1 blockchain through the outbox. Smart Rollups being optimistic rollups, the usual limitations apply for this communication channel. Messages of the outbox can only be triggered on the Layer 1 blockchain once the Layer 2 state advertising them has been published and cannot be contested with an interactive fraud proof anymore. For Smart Rollups, this means waiting for two weeks as of today.

Finally, Smart Rollups get free execution time for every block created by the Layer 1 blockchain. This is a very different model than gas-bounded computation of smart contracts, and enables new use cases. For instance, you can imagine having a kernel reorganizing its storage across several Tezos blocks in a transparent manner, without the need to inject any message to the shared inbox.

Smart Rollup’s Adopter Starter Kit

Implementing a kernel is only the beginning of the journey. You still need to execute it, which is more involved than deploying a smart contract.

If you work on Layer 2 solutions yourself, you may know that optimistic rollups can be tricky to operate, and it would not be reasonable for every Smart Rollups adopter to have to reimplement from scratch their own daemon. From this perspective, our pledge has always been to make that as easy as possible, and to that end, we have invested significant efforts to provide a software suite that lets you focus on your business logic. In particular, we provide

• A high-level, type safe Rust framework to implement your kernel. This makes a very interesting alternative of the low-level, C-like API exposed by the Smart Rollup (because that is how WASM works).
• A debugger to test, inspect, profile and benchmark the execution of your kernel.
• A general-purpose daemon responsible to track the state of your rollup, publish it on the Layer 1 blockchain, and defend it if need be.

A Generalized, EVM-Compatible Layer 2 Blockchain for Tezos

The main reason why we have been eager to comme to EthCC this year is because we are about to bring EVM compatibility to the Tezos ecosystem, thanks in large part to Smart Rollups.

Building an EVM-compatible Blockchain

When we started working on this a few months ago, we have quickly realized that “building an EVM-compatible rollup” does not mean much... or rather, it can be interpreted as many different things. When you build a rollup, you are confronted to a lot of design decisions. The first of such decisions that comes to mind is your requirements in terms of latency: do you need a sub-second block time, or are you fine with sticking to your Layer 1 blockchain latency? The answer to this question has a direct impact on the decentralization of your rollup. As of now, the de facto solution to sub-second latency is the delegation of block production to a single entity (the sequencer). Another interesting question is about data-availibility. Are you fine with being limited by the block space of your Layer 1 blockchain, or do you know your requirements will far exceed them? And what about gas price in your Layer 2 blockchain? Smart Rollups do not have a notion of gas, but if you are building a generalized rollup emulating a full blockchain, you will need gas and a gas price to constraint what your users can do according to your infrastructure capacity. On the other hand, if you are building an AppChain with only a handful of Solidity contract whose execution can only be triggered by yourself, you probably don’t need that.

What was clear to us is that we did not want to take too many decisions like these too soon in the development process, but rather built a versatile solution which can be adapted to end-users need. As a consequence, what we are working on today is a twofold projectWe are providing a layered answer to a complicated question. , with:

1. A customizable “template” for generalized rollup, compatible with the Ethereum ecosystem at large.
2. A community instance of this template, focusing on decentralization and cost-efficiency.

The key idea of this approach is to leave the door open to potential adopters whose requirements are not addressed by the community instance we are working on. It should be straightforward to deploy a dedicated instance that is tailored to their needs.

To achieve this, we have drawn inspiration from the block proposer/block validator paradigm. Our rollup program is divided into two “stages.” The stage 1 interprets messages coming from the shared inbox to reconstruct block proposals. Then, the stage 2 validates these block proposals, and publishes the Layer 2 history. From this perspective, most of the work required to support a sequencer setup is focused on the stage 1, while manipulating gas fees is done by modifying the stage 2 for instance.

Ensuring Compatibility with the Ethereum Ecosystem

This architecture has proven itself to implement the semantics of an EVM-compatible Layer 2 blockchain, but this was only the beginning of the journey. For this new chain to be usable, we also need to be compatible with the galaxy of tools and solutions that is the Ethereum ecosystem. This is achieved by implementing the JSON RPC API . To be clear, this was unknown territory for us at the beginning of this project. The nodes participating in the Tezos network uses a different approachWe have settled on a REST API. . As a consequence, we are developing a new node to fill this gap, and we are now capable of serving the API to tools like Blockscout and MetaMask, or dApps from partners. To give you a concrete example, we had an improvised workshop with Arianee today  to deploy their application on our brand new TestnetIt was deployed the day before the talk, and it is still used today . . As far as I know, it worked! And maybe your dApp is next ?

I mention we are building a community rollup, and I am happy to unveil today that its name will be Etherlink .

What’s coming next?

Our journey to scale the Tezos ecosystem and open it to new use cases and opportunities is far from over. In the meantime, if there are two things that you should remember from this talk, it is that:

1. EVM-compatibility is coming to Tezos this year, starting with EtherlinkThis promise was a tad optimist. The Mainnet Beta of Etherlink was eventually launched on May 15, 2024. .
2. You can already test to deploy your dApp on our brand new Testnet.

Thank you very much, and see you on Tezos!