Gnosis Co-founder discusses "L2 Scalability and Technical Limitations," building an "Ethereum Universe" with ZK cross-chain bridges.

Ethereum infrastructure co-founder of Gnosis, Martin Köppelmann, discussed the limitations of Layer 2 solutions at the ETHDenver conference, emphasizing that L2 cannot effectively solve Ethereum's scalability issues from multiple perspectives. Additionally, he introduced the concept of the "Ethereum Universe" by referencing the design of Cosmos IBC, aiming to create a more decentralized and efficient Ethereum through the technology of ZK cross-chain bridges.

L2 Practicality, Limitations, and Trade-offs

At the beginning of his speech, Martin stated that due to Ethereum's scalability issues, the current mainstream solution relies on Roll up technology to create the thriving ecosystem of L2 today.

However, Martin believes that with new cross-chain technologies reducing risks like ZK cross-chain bridges, L2's attempt to solve problems has new solutions. This presentation focuses on exploring the limitations faced by L2 and why more L1 block space is necessary, making the blockchain more secure and decentralized.

Initial Concept of L2: A Temporary Transaction Space

Martin then mentioned the origin of L2 development, where initially L2 was used to execute batch transactions deposited from L1 and synchronize the results back to L1 for settlement. This makes L2 like a transitional space, where there is not much lingering, and the whole process is also the origin of the term "Roll up."

However, this transaction processing model is only suitable for applications where states do not expand, such as contract exchanges, as they do not require transaction history, only transaction results.

Current Transaction Capability of Ethereum

Although L2 does increase the transaction limit, Martin stated that L2's scalability for Ethereum is still constrained by L1. With Ethereum only able to process about 1.25 million transactions per day, L2 will eventually reach its scaling limit.

So, with L2 in place, what is the current transaction processing capacity of Ethereum? Martin gave the following two examples:

1. ENS: Assuming 10% of the world's population, about 800 million people, register Ethereum domain names together, it would take about two years to process these transactions.

2. Global Stock Market: If Ethereum were to be used as the settlement layer for the global stock market, with 45,000 publicly listed stocks, less than 30 transactions could be processed daily.

Cost Issues with L2 Transactions

In addition to scalability limitations, the cost of L2 transactions is also a problem that needs to be addressed. Martin stated that L2's gas fees are still proportional to L1's fees, and when costs soar above $1, it becomes unacceptable for many use cases.

Although after the implementation of EIP-4844, gas fees will be reduced by over 90%, demand will also rise. For use cases requiring fees below 1 cent, this is still unfeasible.

Furthermore, if the cost of leaving L2 is higher than the user's assets, the user's assets may be stuck in L2 and unable to be withdrawn. Even if everyone can afford the cost of leaving L2, L2's bandwidth limitations may prevent users from withdrawing all assets at once.

Limitations of Roll up Technology

Even if transaction costs and scalability issues are resolved, L2's Roll up technology cannot be used for all applications.

Using CirclesUBI and POAP as examples, Martin pointed out that these applications have large states, making it unfeasible to transfer them from L1 because the data cannot be compressed and settled on L1.

What if we never leave L2?

After discussing the current issues faced by L2, Martin posed the question: "What if we never leave L2?"

He raised two issues with this idea:

1. The current L2 sequencer is too centralized, where transactions could be excluded, transaction fees increased, or certain protocols treated differently.

Using Coinbase's L2 Base as an example, since Coinbase controls the centralized sequencer of Base, they can prioritize their transactions if they wish.

2. Centralized sequencers may face scrutiny or require their users to pass KYC in order to transact.

In addition to the above issues, Martin also mentioned that if the goal is just to issue native assets on the chain, is it really that important to issue them on L2? Although native L2 assets solve cross-chain risk issues due to not requiring cross-chain transactions, the security of L2 comes from validators on L1. Without the security of L1, L2 cannot function properly.

Furthermore, as Ethereum itself is still evolving, L2 mechanisms also need to keep up with the times.

Using the on-chain voting protocol Snapshot as an example, its trustless voting mechanism is carried out on L2 and synchronized back to L1, using Merkle Proof throughout the process. However, according to Ethereum's roadmap, they intend to switch Ethereum clients from Merkle Trees to Verkle Trees, rendering Snapshot unusable at that time.

In response, Martin stated that L2 must have corresponding upgrade mechanisms, but this would contradict its trustless goal.

Given L2's many limitations, what are the alternative solutions?

Martin proposed an idea to run an execution layer and consensus layer network identical to Ethereum's architecture and connect the two chains using ZK cross-chain bridge technology, with an example called "Gnosis." Gnosis will strive to remain consistent with Ethereum as much as possible, and will be compatible with future EIPs, making running Gnosis similar to Ethereum.

On Gnosis, Ethereum's light nodes will run and verify consensus and blocks using EVM to ensure they are signed by the majority of validators, maintaining a consistent state with Ethereum.

The concept of Gnosis' ZK cross-chain bridge is similar to Cosmos' IBC cross-chain communication model, allowing Ethereum and other EVM-based chains to interconnect without trust, ultimately forming an "Ethereum Universe" composed of multiple chains, Ethereumverse.