Use this when assembling ICDE supplemental material. The supplement can support the paper, but the 12-page body must stand on its own — a reviewer must be able to accept the headline result without opening it.
Because ICDE weighs availability in the score, treat the artifact as first-class:
run_all.sh (full reproduction) and run_small.sh (a fast subset for a reviewer without a
cluster), both writing to a predictable output location.| Supplement item | Inspection likelihood | Practical implication |
|---|---|---|
| README + run_small.sh | High | Must build and run cleanly on a stock machine |
| Figure-to-script claims map | Medium-high | Every headline figure needs a traceable command |
| Extended sweeps and baselines | Medium | Reference each from the body or it goes unread |
| Full raw logs | Low | Include for completeness, never depend on them |
| Secondary proofs | Variable | Restate the claim before proving it |
A submission on a write-optimized index: the body keeps the two-tier mechanism, the lazy-merge
pseudocode, the throughput and latency-tail figures, and the deferral ablation; the supplement
holds the full append-to-scan sweep, an extra RocksDB-style baseline, the trace documentation,
and the artifact with run_small.sh reproducing the crossover on one machine. Nothing
decision-critical lives only in the artifact, because deep artifact inspection is at reviewer
discretion.
[Supplement status] Ready / Needs fixes / Not ready
[Body vs supplement] <what stays in 12 pages vs moves out>
[Artifact package] <README / run_all.sh / run_small.sh / generators / claims-map>
[Availability signal] <does the supplement demonstrate a runnable artifact? y/n>
[Hygiene checks] <secrets / caches / bloat removed>
[Main-paper dependency] <what breaks if the supplement is ignored>