Meeting Summary:
This meeting focused on updates and discussions for the Ethereum community, covering multiple technical topics across DevNet, Glamsterdam, and Higota initiatives. The team discussed issues with DevNet 5, including peering bugs in Prism and fork choice bugs triggered by malicious clients, with Potuz explaining the PTC attestation problems and implementing a lazy fix to prevent sending invalid attestations. The group debated Lighthouse’s proposal to make QUIC the default transport protocol, with concerns raised by Lucas from Teku about the potential impact on nodes not fully supporting QUIC yet. James presented the Glass Proposer Preference APIs PR for the Beacon API, which was approved for merging. Raúl presented EIP-8243 on pre-aggregating attestations at source, proposing a mechanism to allow operators with co-scheduled validators to send aggregated messages instead of individual attestations, potentially reducing network traffic by 2x. The team discussed concerns about this approach, including the need for additional validation rules and the potential for spam vectors if aggregators were expanded too widely.
Click to expand detailed summary
The team discussed issues with DevNet 5, where Potuz explained that a bug in their system was sending invalid PTC attestation messages due to incorrect determination of validator duties based on head state. The bug occurred when validators were asked to send attestation for different shuffling branches where they weren’t in the PTC committee. Barnabas reported that while the Prism peering bug might be fixed, the Grendin fork choice bug remains unfixed, though a new image is being built for deployment within about 30 minutes.
The team discussed issues with PTC (Proof of Time Commitment) duties and fork handling in their blockchain system. Potuz explained that problems arise when there are non-finalized situations and long forks, leading to isolated nodes on different branches. The group debated whether clients should send PTC votes for non-canonical branches, with Potuz arguing this is necessary to avoid liveness risks, while others expressed concerns about implementation complexity and potential security issues. The team confirmed that a fix is being deployed to DevNet 5, though there are some ongoing synchronization issues with range payloads and blocks.
The team discussed issues with DevStar blocks and agreed that once forkjet bugs are fixed, the system should properly handle orphaning DevStar blocks. James presented a proposal to deprecate the prepare beacon proposer endpoint in favor of Beacon Committee subscriptions, which was approved to move forward. The main discussion centered on making QUIC the default transport protocol, with Lighthouse planning to enable this by default in their next release by the end of the month. However, concerns were raised by other team members about the risks of removing Mplex support too quickly, with Teku specifically noting they need more time for testing as their QUIC support is still experimental.
The team discussed the potential removal of MPlex transport protocol, with Kingy expressing concern about maintaining an unnecessary surface while others raised concerns about network stability and topology changes. The group agreed to make QUIC the default transport with MPlex as a fallback, and decided to schedule the complete removal of MPlex support for after GLOASS 2, potentially around 2 months later to avoid coinciding with major updates. Lucas confirmed that Nimbus has QUIC implementation ready but not yet enabled by default, and the team discussed the need to backport QUIC support to DevNet branches for testing.
The meeting covered several key topics related to Ethereum’s beacon chain and execution layer upgrades. Leo presented a proposal to rename validator deposit states to better reflect current workflow changes, which would require a version bump to avoid breaking existing tooling. Barnabas proposed adding peer scoring and disconnect reasons to the beacon API’s node peers endpoint for debugging and dashboard visualization purposes. The team discussed a significant proposal from Cayman and Nico regarding builder execution requests, which would create a separate builder deposit contract to avoid DOS vulnerabilities, though concerns were raised about the potential delay to the fork timeline and the complexity of caching implementations across different clients. Raúl presented EIP-8243 on batching attestations at source, proposing a mechanism to allow validator operators to aggregate multiple attester votes into a single network message, which could potentially reduce network traffic by up to 2x. The discussion included concerns about implementation complexity, latency impacts, and the effectiveness of the proposed solution compared to existing aggregation methods.
Next Steps:
- Barnabas: Roll out the new Prism image fix for the peering bug on DevNet 5 and monitor recovery toward finalization.
- Potuz: Investigate and assess the complexity of implementing proper PTC attestation sending for non-canonical branches (beyond the current lazy fix).
- Nico: Continue running the Death Star malicious client on DevNet 5 to test that fork choice bugs are properly handled once fixes are deployed.
- Enrico: Provide a Teku image with parameters for building late beacon blocks and publishing late payloads for additional fork choice testing on the devnet.
- Lucas: Backport QUIC support to the Teku DevNet 5 branch so it can be tested on the devnet alongside other clients.
- Barnabas: Check which clients are missing QUIC support as default on DevNet 5 after Teku backports it.
- James: Follow up with beacon API maintainers to get the Proposer Preferences PR (POST/GET endpoints + SSE + deprecation of prepare_beacon_proposer and register_validator) merged.
- cayman: Update the mandatory QUIC as primary transport PR wording as needed based on feedback, and continue coordinating with client teams on timeline for removing MPlex support (targeting before GLOASS, not at the same time).
- kingy_sigp: Coordinate with other client teams (off the public call) on the timeline for making QUIC default and putting MPlex behind a flag, aiming to do it well before the GLOASS fork.
- leobago: Engage client teams and the community on the beacon API PR for renaming validator deposit states (pending_initialized → new name, pending_queued → new name), and gather feedback on whether to do a breaking change or maintain old labels temporarily.
- Barnabas: Open or update the PR for extending the Peer API with optional peer scoring fields, and solicit client teams to review and attempt to expose downscore reasons.
- Bharath: Merge the staked builder API PR (with Barnabas’s approval) and iterate based on client implementation feedback.
- Client teams (Prism, Teku, Lighthouse, Nimbus, etc.): Review and engage on the builder execution requests EIP (EIP-8282 / separate builder deposit contract) and provide estimates of the effort required to implement proper signature caching as an alternative, ahead of the Monday ACDT call.
- Parithosh: Add the builder execution requests / separate deposit contract discussion (EIP-8282) to the ACDT agenda for Monday, and ensure Justin is present to quantify the severity of the identified attack scenarios.
- FLCL: Resolve the open questions on the engine API sessioning spec (provided by Marios) and proceed to merge the Nevermind PR once answers are confirmed.
- Raúl: Publish the explainer microsite with simulation stats for EIP-8243 (pre-aggregating attestations at source), and run the numbers on the probability of large operators having an aggregator among their co-scheduled validators to compare with simply increasing the aggregator count.
- All client teams: Vote on the ETH Magicians poll for the Glamsterdam mascot emoji.
Recording Access: