rewrite this title Checkpoint #4: Berlinterop | Ethereum Foundation Blog

Ethereum’s weekly All Core Developer calls are a lot to keep up with, so this “Checkpoint” series aims for high-level updates depending on what’s happening in core development. See the previous update here.

This is a special edition of the series!

Kicking off Berlin Blockchain Week, ethereum core devs and researchers got together for an interop hacking week to make progress both on long-term research directions and short-term implementation of the Fusaka upgrade and gas limit increases. Two of these days solicited feedback on longer-term research directions from L2 and zk teams.

The most recent in-person interop was in Bangkok prior to Devconnect but previous interops focused on Pectra & PeerDAS (Nyota), Shapella & Protodanksharding (Edelweiss), the Merge (Amphora), and Eth2 (Ontario)

Short-term implementation

Fusaka

Last week’s interop, Forschungsingenieurtagung (or more practically referred to as Berlinterop), focused on an all-week coworking session where devs launched fusaka-devnet-1 on day 1 and berlinterop-devnet-2 on day 5. Throughout this hacking week, devs found changes that would be useful but could not include them in a canonical “fusaka-devnet-2” without consulting the public ACD governance process on these decisions, which they’ll do this Thursday.

Following this progress, devs will launch a fusaka-devnet-2 and, in an optimistic scenario, don’t expect to need a devnet-3 before moving onto the Sepolia testnet around the end of the (boreal) summer.

This week’s All Core Devs Testing call covered the Fusaka devnet timeline here.

Gas limit testing

In order to make way for the network to safely handle ambitious goals in gas limit increases, devs got together to identify and remove hurdles for throughput increases.

The week included a stress-testing challenge with a leaderboard where devs were awarded points for breaking or hardening devnets. Shout out to Kamil and pk910 for their invaluable participation!

They did indeed come to consensus on a safe immediate higher throughput level and a plan for higher levels, which will be shared from the EthPandaOps twitter account and in the Eth R&D discord when client optimizations that ensure the safety of the 45M throughput level are released within the next week.

This Mondays’s All Core Devs Testing call covered Berlinterop gas limit testing.

Long-term research directions

More detailed summaries and chronological notes from sessions related to all the following sections (and more!) will be posted in the coming weeks in the ethereum/pm Github repo.

Slot restructuring

Devs & researchers discussed two possibilities of slot restructuring: shortening slots and rebalancing the sub-slot timings. They also covered the interplay of various proposals that touch slot structure or are affected by it: ePBS, Delayed Execution, FOCIL.

The session then covered the benefits of shorter slot times: better markets with less stale data, makes big blocks smaller, more competitive builder markets, faster + cheaper interop, more leaders per second, higher censorship resistance.

Two action items that came out of this were to address open questions in this PR to prepare to merge and to adjust the language in specs so that clients must attest as soon as a block is validated and wait until the 4 second mark.

History expiry

There was encouraging progress on history expiry! Expect a blog post here in the next couple of weeks on how validators will default to dropping pre-merge history on mainnet 🎉

There was good agreement on Era file standards, and further updates will be provided in the next two months on rolling history expiry and on the implementation of a distribution mechanism for dropped history. There will be a public community call this coming Friday to discuss the future of Portal.

These updates were covered on this Monday’s All Core Devs Testing call.

CL hardening

Devs met to evaluate areas they’d like to improve to make the consensus layer more robust against disruptive situations like the Holešky Pectra fork and came out with 26 areas for improvement. These areas should be addressed in the next year or so and range from simpler items such as being able to checkpoint sync from a nonfinalized state to more complex ones such as how to optimize client resource usage during nonfinality periods.

These changes will help maintain a healthy network even in the case of nonfinality. Keep up with this progress in the #consensus-dev channel of the Eth R&D Discord server.

L2 day

Representatives from Arbitrum, Base, Linea, OP Labs, Polygon, Scroll, Soneium, Starkware, World Chain, and ZKsync provided feedback about optimizing the L1 <> L2 relationship going forward and helped to identify three areas of focus:

Requests from L2s as users of the L1: more blobs and faster finalityL2s as stakeholders in EVM changes: because EVM equivalence means that changes affect them, they’d like to be considered and kept in the loop to have time to prepare for any adjustments. Some specific focuses were calldata pricing and finding more extensibility pointsL2s pointed out that they’ve accumulated a wealth of knowledge on running high-throughput networks and can be useful in collaborating on scaling designs on the L1

ZK day

Representatives from Brevis, Ethproofs, Irreducible, Kakarot, Linea, Lita, Matter Labs, OpenVM, powdr, RISC Zero, Scroll, Snarkify, Starkware, Succinct, Whirlaway, Zirkuit, Zisk, and ZKM collaborated on the path to a zkEVM future with various areas of focus:

Guest programs & primitives – Teams unanimously agreed that it was too early to enshrine any particular ISA. They favour sticking with a generic RISC-V target (riscv64gc-unknown-linux-elf). Nethermind wants a compilable execution layer guest and hopes to collaborate by EOY on the zkvm benchmarking framework they’ve been working on. Hash choice is still open-ended: No consensus on poseidon2 or a specific field.Standardisation + security – supported standardizing around syscalls and shared Rust libraries that invoke precompiles. There was consensus that 300KB proof is reasonable and that groth16 wrapper could be dropped.zk-stateless client roadmap – Year-end goal for a zk-verified stateless client (starting on Reth); biggest questions are censorship-resistant state sources and how to pay/align provers.

Summary

Interop weeks are enormously productive as they remove async-communication barriers and create contagious motivation among devs and researchers – a lot of things that have been taking months to make progress were given a big push this last week. Having in-person sessions with L2 and zk teams was also beneficial to orient ambitious research directions with wider stakeholder feedback.

As for current development… I’m beginning to think we really will see a 2025 Fusaka!