D-0007 · Where the dates of the later books come from

Sefaria's recorded composition range for the book, converted to Hebrew years, with Sefaria named as the source. A book with no recorded range is marked undated — never guessed.

The options on the table

• Sefaria's composition ranges, converted to Hebrew years; undated where none is recorded (chosen)
• Sefaria's range, cross-checked against the author's lifetime
• The 50 most-cited books dated by hand first, Sefaria for the rest
• The old catalog's era labels

Why

For the books written after Chazal the traditional and historians' counts agree, so D-0002 does not reach them; only the source of the date matters. Sefaria records a composition range for most of its books — 38 of a random 40 on the day of this ruling — and every date taken from it names its source, so it can be checked and replaced one book at a time. A missing date stays missing: "undated" is an answer, a guessed date is not.

What was measured on the day

• The library SeferGraph serves held no dates at all; the old catalog held only an era label (one of seven), wrong in at least 229 places — commentaries dated by the text they comment on (Mizrachi as "tannaim"), and the Shulchan Aruch dated before its own author's earlier Beit Yosef.
• Sefaria's ranges are ranges, not points: Rashi 1065–1115 (≈ 4825–4875), Kessef Mishneh 1545–1575 (≈ 5305–5335), the Bavli 450–550 (≈ 4210–4310).

The measurement: experiments/2026-10-01-dates (the catalog's era labels, Sefaria's ranges, the random sample of 40).

What rests on this

• D-0008 · When one book counts as before another
• D-0009 · The Zohar's date

---

Generated from the register (docs/decisions/0007-book-date-source.md). A changed value is recorded as a new entry; the old one is never overwritten.