Meeting Summary:
The meeting focused on updates and progress regarding DevNet 4 readiness across various client teams. Thomas provided an overview of recent events, including a “beast mode” day focused on execution and ZK VMs, and a “fault mode” event dedicated to consensus layer and post-quantum discussions. Client teams reported their progress on implementing DevNet 4 specifications, including key separation, multi-sig updates, and testing improvements. Key discussions centered around stabilizing dev nets due to increasing client participation, with emphasis on improving test coverage and creating better tools for interop testing. The team also discussed post-quantum features, including recursive signature aggregation, and ongoing research into optimizations and integrations.
Click to expand detailed summary
Thomas updated the team on a recent event focused on Ethereum’s execution layer and consensus layer, including discussions on DevNets, post-quantum signatures, and aggregation. He highlighted challenges with DevNet stability due to a high number of client teams joining and emphasized the need for more extensive testing and test suites to ensure consistency. Thomas mentioned that recursive signature aggregation would help improve stability for post-quantum features and promised updates on potential solutions for DevNet issues in the coming weeks.
The team discussed their progress on DevNet 4 readiness, with Gajinder reporting that they are mostly ready except for pending changes regarding bounding max attestations. Shariq mentioned that they have implemented key separation and are working on the new DevNet 4 branch in Dean Multisig, with plans to test interop with other clients in the coming days. Both teams are focusing on stability and scaling validators, with additional testing being conducted on Hive for multiple clients including Dream and Zeeam.
The team discussed organizing Hive with LeanSpec as the backend orchestrator, replacing the current setup with Kath. Kolby confirmed they could implement this approach and mentioned plans to write smoke tests, with LeanSpec potentially serving as the orchestrator next week. The team agreed to arrange a presentation once sufficient progress is made on the implementation.
Pablo reported progress on Net 4 development, including fixes for key validation, attestation, and block processing issues. The team received a patch from the Gene team and migrated their code to a custom library called Lib SSL. Ruslan asked about the support for fast key generation in the DevNet 4 branch, noting that new keys were being rejected.
The team discussed issues with Lean Quickstart and key generation, with Ruslan investigating a bug found in Kulin on DevNet 3 that will soon migrate to Hive. Mihir provided updates on DevNet 3 improvements, including memory fixes for Lantern and support for pruning fork choice states, and mentioned ongoing work on DevNet 4 and Lean spec test vectors. Gajinder noted they are making Zeam shadow compatible for simulations.
Shaaibu reported that their team has been working on stabilizing Gene on DevNet 3, including updates to the networking stack, gossip layer, and backfield sink for missing blocks. They also began reviewing DevNet 4 specifications to accelerate development after stabilizing DevNet 3. Mercynaps mentioned merging a PR for splitting block production and fixing bulk KAI in the spec, with plans to submit another PR by week’s end pending no changes to the spec.
Thomas provided updates on the specification and research sides. The team merged several key changes including switching to Poseidon 1 for security reasons, implementing ZK-friendly encoding for XMSS signatures, and enabling recursive signature aggregation. Thomas is now focusing on improving test coverage, including both unit tests and test vectors for client teams, and plans to work on simplifying the specification by consolidating duplicate aggregate methods. On the research side, the team is nearly complete with integrating their polynomial commitment scheme into Plonky 3, with only one PR under review before merging.
Parthasarathy provided an update on dev net activities, explaining that Katya has been running multiple dev nets including a larger one involving all clients and a smaller one focused on identifying edge cases. Parthasarathy discussed ongoing conversations with Panda DevOps about optimal configurations for scaling to a large number of validators, noting that while 5,000 validators can run on a single machine in some cases, their current configuration struggles with running two clients simultaneously. He plans to run experiments to determine the ideal configuration and start with a smaller 100-client dev net before scaling up.
Next Steps:
- Anshal: Put up a spec PR for bounding the max attestations so it can be merged by client teams for DevNet 4 readiness.
- Partha: Put up a new PR for Lean Quick Start to make it DevNet 4 compatible.
- Partha: Check with Ruslan to identify and fix the issue with Lean Quick Start rejecting new keys on the DevNet 4 branch.
- Kolby/Shariq (Dream team): Work toward having Lean Spec as the orchestrator in Hive by next week.
- Kolby/Shariq (Dream team): Arrange a presentation on Hive testing progress for all client teams once sufficient progress is made.
- Kolby: Coordinate with Felix to get Hive fork changes merged into the Ethereum Hive organization repository.
- Islambda (Pablo’s team): Finalize polishing, testing, and reviewing before publishing the first DevNet 4 Docker tag.
- Mihir (Lantern): Complete DevNet 4 branch work (PR496) and port memory/P2P improvements from DevNet 3 to DevNet 4 branch, targeting readiness this week.
- Mihir (Lantern): Submit additional PRs to Lean Spec for finalization test vectors after verifying sufficient justification test vector coverage.
- Granny/Mercynaps: Put up the DevNet 4 PR by end of the week (contingent on no further spec changes).
- Thomas: Open additional issues to guide contributors on missing test vectors to increase spec test coverage.
- Thomas: Work on simplifying the aggregate method duplication in the codebase when time permits.
- Thomas: Discuss with Will and internal EF coordinators about tooling for a visual recap dashboard of all clients and tests passed before interop.
- Partha: Run experiments today/tomorrow to determine ideal infrastructure configuration for DevNet 4, starting with a smaller ~100-client devnet before scaling up.
- T. (Emil): Continue experiments on integrating pull parameters and tweaks efficiently with pre-compiles in LeanVM.
- Will: Process and publish videos, audio recordings, and other materials from the CAN event.
- Thomas/EF team: Come up with potential solutions for DevNet stability issues (large number of client teams, instability) to present in coming weeks.
Recording Access:
DEVNET ROADMAP PLANNING: WEDNESDAY APRIL 22 @ 14:00 UTC.