Sigcomm supplementary
Skill brycewang-stanford/Awesome-Journal-Skills/SIGCOMM-Skills/skills/sigcomm-supplementary
Use when organizing the material around an ACM SIGCOMM paper — deciding what lives in the 12-page body versus references versus optional appendices, keeping the main paper self-contained, anticipating the shepherd's appendix-necessity approval, and placing extended measurements and proofs outside the cap without hiding load-bearing content.From its SKILL.md
npx -y skills add brycewang-stanford/Awesome-Journal-Skills --skill sigcomm-supplementaryAssembled from the repository path, not quoted from the project. Check it against their README if it does not work.
SKILL.md
3.7 KB, 731 tokens by cl100k_base, as published. Nobody here has run it
SIGCOMM Supplementary
Use this when splitting a SIGCOMM paper across body, references, and appendices. The controlling rule: the 12 single-spaced pages before references must carry the argument on their own; appendices and references sit outside the cap but do not rescue an incomplete body. Reopen the current CFP to confirm this edition's appendix handling.
What goes where
- Body (≤ 12 single-spaced pages, figures and tables included): the measured problem, the design principle, the mechanism, and the decision-critical evaluation. A reviewer must understand and believe the contribution without turning to an appendix.
- References (outside the cap): as many as needed; never trimmed for space or anonymity.
- Appendices (optional, after references, outside the cap): extended measurement sweeps, additional topologies, proofs or derivations, detailed configuration, and workload specifics — support, not substitute.
The appendix-necessity rule changes the calculus
Unlike venues where appendices are a free overflow bin, SIGCOMM's shepherd approves whether each appendix is necessary at camera-ready. Write appendices you can defend as load-bearing, and never park a step the argument depends on where a shepherd might cut it. A safe test: if removing an appendix would make a body claim unverifiable, that content probably belongs in the body or the artifact, not in a droppable appendix.
Placement table
| Content | Home | Why |
|---|---|---|
| The core mechanism and its principle | Body | The contribution reviewers evaluate |
| The one or two decision-critical figures | Body | The evidence the accept/reject turns on |
| Extra topology / parameter sweeps | Appendix | Depth without crowding the 12 pages |
| Full configuration and workload spec | Appendix or artifact | Reproducibility detail, referenced from the body |
| Proof or derivation for an analytical claim | Appendix | Body keeps the statement and a sketch |
| Raw logs and per-run tables | Artifact | Never the paper's job to hold these |
Keep the body self-contained
- Every appendix and artifact item must be cross-referenced from the body at least once; orphaned supplementary material is invisible and, at camera-ready, deletable.
- Order appendix sections to mirror the body's flow so a reader navigates by section, not by hunting.
- State results in the body and defer only the support — a reader who never opens an appendix should still be able to reconstruct why the contribution holds.
Vignette: splitting a congestion-control paper
A submission proposes a new congestion signal with a testbed evaluation and a small stability argument. The body keeps the measured motivation, the signal and its principle, the two decision-critical tail-latency figures, and a one-paragraph stability sketch. The appendix holds the full stability derivation, a switch-buffer sweep, and the fairness-across-flows tables; the artifact carries the driver and logs. Nothing the accept decision depends on lives only in a droppable place.
Output format
[Body self-contained] yes / no (what breaks without an appendix)
[Body budget] within 12 single-spaced pre-reference pages? yes/no
[Placement map] <content -> body / appendix / artifact>
[Appendix necessity] each appendix defensible to a shepherd? yes/no
[Cross-reference gaps] <orphaned supplementary items>
What ships with it
Read from the repository
Just SKILL.md. No reference files, no scripts.