AllCoreDevs - Execution (ACDE) #215 (July 3, 2025)

Action Items & Next Steps

Fusaka

Glamsterdam

History Expiry

  • All EL clients to enable pre-merge history pruning by default in their next stable release
  • Teams should review the EraE draft and comment in #history-expiry

Summary

1. devnet‑2

  • Finalization stalled (~56 % participation): fork‑ID mis‑calculation in Besu, 4‑byte IDs mishandled by Nethermind; EL/CL Nimbus unable to sync.
  • Agreed on a standardised Genesis‑config RPC so devnets catch divergent parameters before launch; no client objects, all to estimate effort by Monday’s testing call.
  • Devnet‑2 will remain up only as a bug‑hunt sandbox; Devnet‑3 becomes the “happy‑path” network.

2. EIP‑7907 (Meter contract size & raise limit)

  • New limit 48 kB (2× current), down from the early “10×” proposal; code‑size index optional, NOT in consensus.
  • Inclusion gate: three PRs (clarification, patch removing code‑size warming, large‑contract fix).
  • Stress‑test methodology: spammer tx + large‑block replay in Devnet‑3; pull the EIP if DOS issues surface.

3. Blobs‑per‑Transaction Debate

  • Status‑quo complexity (6 → 12 jump inside BPO schedule) deemed pointless.
  • Decision: hard cap = 6 in PeerDAS; no BPO knob. Simple, less brittle, matches current L2 behaviour (Base averages <1.5 blobs/tx).
  • Base‑fee update fraction remains in BPOs – valuable future lever; removal would create awkward special‑cases when tuning throughput.

4. Precompile Re‑Pricing

  • ModExp (EIP‑7883): triple cost; beats worst‑case at 100 M gas limit. Consensus that performance‑first outlook outweighs small fee hike for a handful of power users (Matter Labs, EigenDA, Coinbase wallets).
  • secp256r1 / r1 curve (EIP‑7951): 2× bump to keep parity with EC ADD/MUL. No opposition.
  • Clients still free to pursue GMP / gnark acceleration, but gas schedule will assume “slow” path.

5. History Expiry Roadmap

  • EraE → single EL‑centric archival format (headers, bodies, receipts in columnar chunks + optional Merkle proofs). Supersedes Era‑1, avoids consensus‑layer cruft.
  • Once Era‑E spec is frozen and files generated, clients may prune history up to Cancun; rolling expiry research continues in parallel.

6. Glamsterdam Scheduling

  • Headliner proposal window closed (20 Jun). EL headliner reviews on July 31 & August 14 ACDEs.
  • forkcast.org launched – comparative matrix of every headliner & stakeholder impact; authors invited to update their rows.

Notes were generated using o3, and edited by me.