Skip to content

paper: endorsement polish (claims audit, margin columns, forest plot, evidence map) - #57

Merged
divyanshusingh2903 merged 1 commit into
mainfrom
paper/endorsement-polish
Oct 23, 2026
Merged

divyanshusingh2903 merged 1 commit into
mainfrom
paper/endorsement-polish

Conversation

@divyanshusingh2903

Copy link
Copy Markdown
Owner

Final polish of the paper before it goes to arXiv endorsers. Paper-only changes (paper/); no code, config or result files touched.

Claims and wording

  • Abstract rewritten shorter and in prose: two headline numbers, what helped and what did not.
  • The aging claim is scoped to where it was tested (production flow, Huawei); Azure is stated as untested.
  • Sidecar comparison softened: different baseline and hosts, not a tested comparison.
  • Window-averaged and whole-run mean latency are defined and labelled.
  • Oracle numbers as paired means, lookup cost 2.7 us, misconfigured-map cost given, RabbitMQ 3.12 / 4.3 qualification.

Presentation

  • Margin columns on the gate tables, a forest plot for the trace results, per-seed dots on the production-flow figure, an evidence map at the top of Results.
  • Tables 5 and 8 pinned in the body instead of landing after the references.

References

  • Huawei 2025 release cited to Joosen et al. (EuroSys 2025), 2023 release to the SoCC 2023 paper; duplicate DOI, Dean issue, Cobham journal, River authors, missing URLs fixed.

Built from a clean copy with the Docker TeX Live image: 39 pages, no undefined references or overfull boxes. All numbers regenerate from gen_numbers.py.

The release tag v0.2.0-preprint (named in the paper) is already pushed at the head of this branch; merge with a merge commit, not a squash, so the tagged commit stays in main's history.

🤖 Generated with Claude Code

…olumns, forest plot, evidence map)

- Abstract rewritten shorter and in prose: two headline numbers, what helped and what did not, no tail claim.
- Scope the aging claim to where it was tested (production flow, Huawei); Azure stated as untested.
- Soften the sidecar comparison (different baseline and hosts, not a tested comparison).
- Define window-averaged against whole-run mean latency and label both.
- Add margin columns to the gate tables, a forest plot for the trace results, per-seed dots on the
  production-flow figure, and an evidence map at the top of Results.
- Oracle numbers as paired means; lookup cost 2.7 us; misconfigured-map cost given; RabbitMQ 3.12/4.3 qualification.
- Huawei 2025 release cited to Joosen et al. (EuroSys 2025), 2023 release to the SoCC 2023 paper.
- Bibliography: DOI duplicate, Dean issue, Cobham journal, River authors, URLs; no Kafka mirror link.
- Pin Tables 5 and 8 in the body; link Opencode; release tag now v0.2.0-preprint.
@divyanshusingh2903
divyanshusingh2903 merged commit 06adde5 into main Oct 23, 2026
6 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant