OSDI Camera-Ready

Deliver the final paper and the conference obligations. Dates and budgets below are the OSDI '26 cycle (notification March 26; final printable PDF due June 9, 2026; conference July 13–15, Seattle — verified 2026-07-08); confirm the current Requirements for Authors page before trusting any of them for another year.

The June 9 deliverable

The final paper is not the submission with names added. Between March and June three transformations happen, in this order:

1. Shepherd-mandated changes (conditional accepts): completed and approved, or the acceptance lapses — this dominates the calendar (osdi-author-response).
2. Expansion into the final budget: up to 14 pages including figures and tables, plus unlimited pages for references and appendices — the appendix right the submission never had. Execute the expansion plan annotated in the source at submission time (osdi-supplementary): restore cut failure cases, attach the full proof, publish the complete result grids.
3. De-anonymization: real authors and institutions on the title page; the renamed system optionally reverts to its public name — decide deliberately, update every occurrence (body, figures, artifact, README), and keep paper and artifact names consistent forever after.

Plus the earned extra: papers that passed artifact evaluation may add a two-page Artifact Appendix in the standard format, condensed from the AE runbook (osdi-artifact-evaluation).

De-anonymization checklist

What USENIX publication means (and does not cost)

OSDI is a USENIX venue, which changes the publication economics authors expect from ACM conferences:

• Open access, immediately. The proceedings PDF is free to everyone from the day the event begins — no paywall, no ACM DL entitlements, no separate "open access option" to purchase. Your paper's canonical, citable home is a free usenix.org URL.
• No author publication charge. USENIX funds events through registration (with hardship discounts offered for OSDI '26); there is no APC line item to budget.
• Slides and video posted after the event are likewise open. Assume the talk is a permanent public artifact and prepare it to that standard.
• Registration rates and any presenter-registration requirement were not readable at verification time (待核实) — confirm both on the attend pages before booking.

Consequence worth planning for: the paper is public during the conference, so reproduction attempts, blog dissections, and follow-up emails start in July, not after some embargo. Have the artifact live and the README polished by the event.

Presentation and the event

• One author presents in Seattle (July 13–15 for the '26 cycle); check the Instructions for Presenters page for slide specs, session logistics, and A/V deadlines.
• The Jay Lepreau Best Paper Awards are announced from the accepted program — the opening-session tradition since OSDI '08.
• Prepare the talk against the same arc as the paper (pain → idea → evidence), not as a section tour; the audience holds the free PDF in their laps.

Between final paper and Seattle

The five weeks from June 9 to the July 13 opening are not idle:

• Talk first draft within a week of the final paper, while the argument is warm; systems talks fail from cramming the evaluation, so pick the two results that carry the idea and cut the rest to backup slides.
• Dry-run twice, once with systems people (they attack the design) and once with outsiders (they reveal which slide loses the room).
• Artifact final polish: the README's public debut is effectively July 13; run one last cold-start reproduction on a fresh machine or fresh container.
• Coordinate the announcement: with open-access publication at event start, the lab blog post, the repository going public under its real name, and any preprint update should all land the same week, pointing at the canonical usenix.org URL.
• At the venue: sessions are recorded and posted openly; treat Q&A answers as citable statements about the system's limits.

Final-week mechanics

final-week checklist (against the '26 Requirements for Authors)
[ ] shepherd approval recorded (conditional accepts) — BLOCKING
[ ] 14-page budget respected; appendices compile; refs unlimited but clean
[ ] artifact appendix (if badged) in standard format, <= 2 pages
[ ] de-anonymization table above executed; grep for the anonymized name
[ ] figures regenerate from the ledger at the final commit (no stale panels)
[ ] printable PDF validates: fonts embedded, no overfull cropping, metadata set
[ ] uploaded to the submissions system before June 9; confirmation archived
[ ] consent/copyright forms: check current cycle's requirements (待核实)

Discoverability after publication

Small June decisions determine how the paper is found for a decade: keep the title byte-identical everywhere (final PDF, HotCRP record, talk slides, repository README) so citation databases merge rather than fork the entries; put the artifact's DOI in the paper and the paper's usenix.org URL in the artifact so each permanently resolves the other; and once the dblp conf/osdi/ entry appears, adopt it as the canonical BibTeX rather than letting five hand-typed variants propagate from the lab's webpage.

Output format

[Deadline runway] days to final-paper due; shepherd status (n/a | pending | approved)
[Expansion] 14-page budget used how; appendix plan executed? artifact appendix in?
[De-anonymization] checklist rows outstanding: <list>
[Open-access readiness] artifact live + README public-grade before the event? yes/no
[Presenter logistics] registration/attendance items confirmed vs 待核实