[EONRaider]

Summary

Releases Tier 2 of #170 (VRRP, RIP, OSPF, PIM, BGP: #243, #245–#248) as 2.5.0. It adds new public API and removes nothing, so this is a minor version. Every comparative figure in the README was re-measured per Rule 4 of docs/CLAIMS.md.

One headline claim no longer holds and is replaced. "Within 15% of dpkt on decode" is no longer true on the 265-frame corpus: dpkt decodes it 1.27–1.30× faster over three runs.

I ran a file-by-file A/B against the 2.4.0 tree (0a33a5d) on the same machine to find out why:

• On the five new routing captures, dpkt is 1.46× faster. dpkt decodes OSPF, PIM and VRRP to the same layer as netprotocols.
• On the 140 frames 2.4.0 was measured on, the two are within 10% of each other on both trees. 2.4.0's own --compare, re-run here, still reads 1.03×.

So the gap comes from the new workload, not from a code regression. I didn't profile why dpkt is faster on those protocols; that's #250.

The README bullet now quotes 1.3× and says where the gap comes from, and CLAIMS 1.2 is renamed to the figure it now holds.

What's included

• Version: 2.4.0 → 2.5.0 in pyproject.toml, __version__ and uv.lock.

• CHANGELOG: [Unreleased] becomes [2.5.0] - 2026-10-25, with Development entries for the re-measurement and the benchmark baseline.

• README "Why":

  Claim	2.4.0	2.5.0
  Decode vs scapy	4.1× faster	3.4× faster
  Decode vs dpkt	within 15%	dpkt 1.3× faster
  Decoded deeper than dpkt	39 of 140 frames	79 of 265 frames
  Encode vs dpkt	4.2× faster	4.6× faster
  Import (netprotocols vs scapy.all)	55 ms vs 428 ms	70 ms vs 397 ms
  Wheel size	124.4 KiB, 21.7× smaller than scapy	176.5 KiB, 15.3× smaller

  The 40 frames newly decoded deeper are BGP's 19 and RIP's 21, which dpkt leaves as raw bytes.

• README roadmap: names 2.5.0 and links feat(layer3): decode VRRPv2 and VRRPv3, dispatched from IP protocol 112 #243 and feat(layer7): decode RIPv1 and RIPv2, dispatched from UDP port 520 #245–feat(layer7): decode BGP-4 messages dispatched from TCP port 179 #248.

• docs/CLAIMS.md:

  • a new "Re-measured for 2.5.0" section, with each run, the per-file table and the A/B;
  • dated lines under claims 1.1–1.6;
  • claim 1.2 renamed and 1.4 retitled;
  • a §6 addendum.

• benchmarks/baseline.json re-recorded (3fed671): from 2.2.1 on 97 frames (normalized 6.5677) to 2.5.0 on 265 frames (5.4344).

Why the benchmark baseline was re-recorded

The CI benchmark gate failed on this PR's first head (54afd75): -17.1%, then -16.5% on its own retry, against a 15% threshold. This PR changes no library code. The same tree minus the version bump (master at 03387af) passed that job ten minutes earlier at -14.1%.

The cause is the baseline. It was recorded at 2.2.1 on the 97-frame corpus, and the gate walks the whole corpus, now 265 frames. Tier 1 and Tier 2 added captures that cost more per frame, so the gate drifted to its own threshold and runner noise decided the result.

The reviewer's A/B shows no code regression across the span the old baseline covered. On the same CPython 3.12.3 with 30 repetitions and best of 9, over six alternating rounds:

Comparison	Corpus	f/s change
2.5.0 vs 2.2.1	2.2.1's 97 frames	+3.0%
2.5.0 vs 2.4.0	2.4.0's 140 frames	-2.1%
Workload, 97 → 265 frames	2.5.0 code throughout	-16.7%

The -16.7% workload drop matches CI's -17.1%.

Per the steward rule (.claude/skills/steward/SKILL.md), an independent agent made the fix and another independent agent reviewed and approved it. The fixer re-recorded with the script's own --update-baseline, using the settings the gate reuses (30 repetitions, best of 9). The recorded value sits at the median of the surrounding runs, not at an outlier. The threshold, the CI job and benchmark.py are unchanged.

Re-anchoring accepts the drift since 2.2.1. From here, the gate guards against regressions relative to 2.5.0 on this corpus.

Verification

• uv run ruff check and uv run ruff format --check are clean
• uv run mypy is clean (strict)
• uv run pytest passes locally: 4743 passed
• CHANGELOG.md has an entry for 2.5.0
• The CI benchmark command, run locally after the re-record, read between -3.4% and +8.4% over six runs, all within the threshold.
• Measured on CPython 3.12.3 (Intel Xeon @ 2.10 GHz, 4 vCPUs) against dpkt 1.9.8 and scapy 2.8.0, the versions the lock pins:
  • decode: benchmark.py --compare --repetitions 30 --trials 9, three runs;
  • encode: benchmark_encode.py;
  • import: benchmark_import.py, best of three;
  • depth: benchmark.py --depth.
• uv build --wheel on a git archive of the release commit produced netprotocols-2.5.0-py3-none-any.whl at 180,744 bytes, identical across repeated builds. The baseline commit changes no file that ships in the wheel.

Notes

• Tag order: v2.4.0 hasn't been tagged yet. It needs pushing at 0a33a5d before v2.5.0 is pushed at this PR's merge commit.
• Import time grew by 15 ms. The five new modules define 27 frozen dataclasses, which are generated at import.
• Follow-up: perf: profile and narrow the decode gap to dpkt on OSPF, BGP and PIM #250 covers profiling and narrowing the decode gap on OSPF, BGP and PIM.

🤖 Generated with Claude Code

https://claude.ai/code/session_01CEQCZGyuvpMyZwUbDucF3G

[coderabbitai]

Warning

Review limit reached

You've used all free OSS reviews for now. Wait for the free limit to reset to keep reviewing this public repository.

Next included review available in 22 minutes.

Check out review usage here.

View limit details

Limit details: You’ve used the included review currently available.

Learn how review limits work.

Review configuration:

Run configuration

• Source mode: Auto
• Repository: Unavailable
• Branch: Unavailable
• File: No configuration file from this repository applied
• Commit: Unavailable
• Layers: Organization UI
• Review profile: ASSERTIVE
• Plan: Advanced
• Run ID: 0f7333ff-1f4b-45c5-84d7-d1175cd124ad

Commits

Reviewing files that changed from the base of the PR and between 03387af and 3fed671.

Files ignored due to path filters: 1

• uv.lock is excluded by !**/*.lock

Files selected for processing: 6

• CHANGELOG.md
• README.md
• benchmarks/baseline.json
• docs/CLAIMS.md
• pyproject.toml
• src/netprotocols/__init__.py

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}