brycewang-stanford NSDI Camera-Ready
Acceptance starts a production pipeline with several deliverables owned by different people — the final PDF, the public artifact, the presenter designation, and the trip to the symposium. Facts below reflect the NSDI '26/'27 pages rendered on 2026-07-08; the '27 final-paper dates were not yet posted (待核实), so anchor every date to the notification email and the authors' instructions it links.
Deliverables and owners
| Deliverable | Owner to assign | Hard constraint |
|---|---|---|
| Final PDF | writing lead | same 12-page cap as submission |
| Public artifact release | artifact lead | before final-papers date for award eligibility |
| Production/copyright forms | corresponding author | per the authors' instructions |
| Presenter designation | presenting co-author | via the co-chairs' Presenter Information form |
| Travel, visa, registration | presenter | May 11-13, 2027, Providence, RI |
Assign these on notification day. Cohort timing matters here: a spring acceptance (notified July) has months of runway before the May symposium, while a fall acceptance (notified December) runs its production pipeline much closer to the event — same deliverables, tighter interleaving.
The final PDF
NSDI's cap is symmetric: final papers, like submissions, are limited to 12 pages including footnotes, figures, and tables, references and appendices beyond. There is no accepted-paper page bonus to spend, so camera-ready editing is substitution, not addition:
- De-anonymize: real names, affiliations, acknowledgments, funding — all inside the same 12 pages the anonymous version used, which usually costs a few lines; reclaim them from prose, not from evidence.
- Restore what blinding distorted: research-track papers that neutralized deployment context ("a large provider") may now name it — verify employer/legal sign-off on every restored detail, since this step is where confidentiality incidents happen.
- Real links replace anonymized ones: repository, DOI'd archive, dataset pages —
consistent with what the artifact ships (
nsdi-artifact-evaluation). - Honor the revision record: for papers accepted via one-shot revision, confirm every change promised in the change memo survived into the final text; shipping a final that quietly reverts a required fix betrays the reviewers who accepted on those terms.
- Then re-verify the page count — de-anonymization and shepherd-requested edits both nudge layout.
One deadline, two prizes
The final-papers deadline carries double weight at NSDI: it is the production cutoff, and it is the Community Award eligibility line — the award goes to the best paper whose code and/or dataset is public by that date. Sequence accordingly:
- Public release of the artifact repository/archive (with license and DOI) before the final-papers date.
- Final PDF cites those live, permanent URLs.
- Artifact-badge evaluation proceeds on its own schedule; badge outcomes do not retroactively fix a missed release date.
Presenter and symposium duties
- Accepted papers designate one co-author as presenter via the Presenter Information form the program co-chairs distribute. Decide early — visa timelines for a US venue (Providence, RI for May 11-13, 2027) run months, and the presenter should ideally be the person who can answer testbed questions.
- Registration requirements, speaker discounts, and remote-presentation options for '27 are 待核实 — read the instructions-for-presenters page of the current edition rather than assuming either flexibility or rigidity.
- Spring-cohort papers surface on the public accepted-papers list well before the event; coordinate any press/blog timing with your institution against the venue's publication schedule, not against the symposium date alone.
Open access, and what it changes
USENIX proceedings are open access: the final PDF becomes a permanently free
download on usenix.org, joining a record where every prior NSDI paper is equally
free. Practical consequences:
- No paywalled-version hedging — the
usenix.orgPDF is the canonical citation target; sync any arXiv/preprint copy to the final text or clearly mark it stale. - The paper will be read far outside the reviewing community (operators, OSS users, hiring committees); the abstract and figure captions carry more weight than at paywalled venues.
- Slides and talk materials commonly join the presentation page; treat them as part of the paper's permanent public footprint.
Production checklist
[ ] Authors/affiliations restored; ORCID/name spellings confirmed by each coauthor
[ ] Acknowledgments + funding numbers complete (grant offices care)
[ ] 12-page cap re-verified after de-anonymization
[ ] Employer/legal sign-off on restored deployment details
[ ] Revision-memo promises intact in final text
[ ] Artifact public with license + DOI before final-papers date (award clock)
[ ] All URLs in the PDF resolve and are permanent
[ ] Copyright/production forms submitted per authors' instructions
[ ] Presenter designated via the form; travel/visa started
[ ] Metadata (title/abstract) identical across HotCRP, PDF, and artifact README
After the symposium
The paper's public life is just beginning: keep the artifact's pinned dependencies buildable for at least the horizon on which one-shot-revision reviewers and follow-up work will consult it; route errata through the venue's published correction channel rather than silently editing preprint copies; and archive the talk recording/slides link alongside the repository README so the three public faces of the work — PDF, artifact, talk — reference each other.
Output format
[Deadline map] final-papers date + form dates (from notification email)
[PDF status] de-anonymized / cap verified / links live / memo honored
[Award clock] artifact public before final-papers date? y/n + date
[Presenter] designated? visa/travel risk level
[Open-access sync] preprint copies reconciled with canonical PDF
[Open items] owner -> item -> due date