Meeting Summary:
The meeting covered discussions on various technical aspects of upcoming network upgrades, including implementation details, performance issues, and timeline planning. The team debated and made decisions on several key topics, such as transaction limits, base fee updates, and code size limits, while also addressing concerns about client performance and potential pricing adjustments. Additionally, the group reviewed proposals for future forks and discussed the implementation of new features like partial history expiry and error file formats.
Click to expand detailed summary
Tim and Akash discussed the timing and posting of Twitter Stream links on the Github agenda, with Tim suggesting posting them 24 hours in advance for better visibility. Tim also mentioned that they could start the stream in a minute, and Akash confirmed they were live. Tim welcomed everyone to the meeting and mentioned having a few items on the agenda for discussion.
The team discussed issues with the Fussaka scope and Glamsterdam headliners, noting that finalizing headliners in two weeks was unrealistic due to pending Fussaka finalization. Barnabas reported that Defnet 2 was not progressing well, with only 56% completion and various peering and fork ID issues affecting multiple clients. The team agreed to investigate implementing EIP-7910 for EL clients, with teams given a few days to assess the implementation effort before making a decision on Monday. Regarding EIP-7907, Dragutin explained that three PRs need to be merged, including one for code size limit reduction, another to remove warm read cost for CODESIZE, and a third for handling large contracts, with all clients appearing to be in agreement.
The team discussed merging three PRs and agreed to include the code size index as an optional implementation detail, leaving it up to clients to decide how to handle it. They debated the merits of increasing the contract size limit from 24KB to 48KB, with Allan representing Arbitrum expressing a preference for the full 256KB limit but acknowledging the need to start slow. The group agreed to include the index in the spec for Fuzaka, with plans to monitor performance and remove it if issues arise, while Łukasz emphasized the need for benchmarks before making final decisions.
The team discussed the implementation of Devnet 3, agreeing to set the code size limit to 48KB and conduct stress testing after its successful deployment. They also addressed the max blobs per transaction limit, with FLCL noting that clients currently use the max value by default, which ranges from 6 to 12 in the first VPO. The team expressed a preference for keeping the limit at 6, but FLCL suggested creating a PR to formalize this decision.
The team discussed the definition and implementation of a video limit for transactions, focusing on whether it should be a text pool limit or part of the consensus mechanism. Francesco and others argued for including the limit in consensus to prevent large transactions, while FLCL suggested keeping it flexible for future adjustments. The group agreed to hardcode the limit at 6 or 9 blobs per transaction, with plans to review and potentially adjust it in future forks. They also considered the impact on rollups and concluded that the current limit of 9 blobs is sufficient for most use cases.
The team discussed two main topics: transaction limits and base fee update fractions. They decided to remove the max blobs per transaction from the BPO specification and add a cap of 6 in the peer-to-peer CIP. Regarding base fee update fractions, they chose to keep the existing feature in the BPO config, despite the potential for future complexity, as it provides a useful lever for adjusting throughput. The team agreed to revisit the removal of this feature if it becomes a significant pain point.
The team discussed performance issues with MoD exp at higher gas limits, considering two paths: client optimizations or increasing the price. Marcin presented data showing current client performance and the need to either improve performance or reprice, with a proposal to triple the price. The team debated the impact of this change on users and agreed to investigate further before making a final decision. They also discussed changes to Blob base cost and curve repricing in other EIPs, deciding to proceed with doubling the price for EIP-7951 without more data.
The team discussed the timeline for reviewing headliner proposals for the Glamsterdam fork, agreeing to focus on the EL side in two calls on July 31st and August 14th after finalizing Devnet 3 on July 17th. They decided to move non-headliner EIPs to the PFI list and agreed to wait for the headliner decision before reviewing smaller EIPs. Matt presented a new error file format (eraE) for the execution layer, which aims to improve syncing and allow for history expiry, and requested feedback on the specification. The team also discussed the implementation of partial history expiry, noting that some clients like Besu and Nethermind plan to enable it by default in their next releases.
Next Steps:
- Tim: Reach out to MatterLabs, Coinbase Smart Wallet, and EigenDA regarding the impact of 3x price increase for MOD_EXP.
- Client teams: Review and provide feedback on the era-e file format proposal by the end of July.
- Client teams: Confirm in the History Expiry channel their plans for dropping pre-merge history by default.
- Client teams: Start reviewing Amsterdam headliner proposals before the July 31st and August 14th calls.
- Francesco: Open a PR to change the Blob base cost in EIP-7918.
- Marius: Open a PR to 2x the price of the r1 curve precompile.
- Guillaume: Leave a comment on the PR regarding the removal of previously CFI’d EIPs.
- Tim: Share a summary of the call and action items in the chat later today.
Recording Access: