nucleation

ADR-002: published: is the exact date of first appearance where a source gives one

2 min read

Properties3
StatusActive
Version2
Tagsrecord

Graph View

Context

The LIT template, ported from the anthology, said published: is “the arXiv posting month, from the id: 2205.11487 → 2022-05”, written as the first of that month. Every arXiv abstract page states the v1 submission day, and every court filing, dated report and blog post in this record states its day too. Recording 2025-10-01 for a paper posted on 23 October is not a coarser truth; it is a false date that looks exact. It surfaced when a close reading (NOTE-005) flagged the discrepancy and the record’s first answer was to defend the convention. The owner’s ruling: use exact dates where we have them.

Decision

published: is the exact date of first appearance:

  • arXiv: the v1 “Submitted on” date from the abstract page.
  • DOI without arXiv: the earliest full (year-month-day) date Crossref reports among posted, published-online, published-print and issued.
  • URL-only documents: the date the document itself carries (a filing date, a hearing date, a post date).
  • No day available: the first of the month, as before. This is the only case where -01 means “month precision”, and it should be rare.

Deposit or indexing dates (Crossref created) are not publication dates and are not used.

When unsure, the earlier date. Where sources disagree, or the rules above leave open which appearance came first, published: takes the earliest date any of them gives, and the history note names the alternatives. That covers a preprint posted before the journal version under its own DOI or title, a publisher’s online date earlier than every Crossref field, and an arXiv copy posted after the journal version. (Owner’s ruling, 2026-10-09.)

Alternatives

  • Keep the month convention — rejected: it records a date nobody meant and discards one every source gives.
  • Add a separate published_precision: field — rejected for now: every current LIT but three has a day, so a field to mark the exceptions would be a field almost every record shares (the anthology’s DP-002).

Consequences

All 29 LITs were backfilled on 2026-09-25 by script (exact_dates.py: arXiv abstract pages, then Crossref) and by hand from the documents for the six URL-only filings. Three keep month precision because no source gives a day: the ISW report, the Toxicon paper (Crossref gives only the print month) and the concussion review. The template comment now says “exact date”. The anthology carries the same decision.

Backlinks