Skip to content

Errata in published sources

Implementing a standard clean-room means re-deriving every formula, constant and worked example from the source document rather than from anyone else’s code. Done across hundreds of documents, that process finds defects in the sources themselves: a worked example that contradicts its own normative clause, a constant with a digit dropped in typesetting, a cross-reference that points at the wrong equation.

This is the registry of those findings. It covers every kind of published source the library implements from: standards (ISO, IEC, EN), guidance documents and technical reports (EASA, ECAC, NRL), textbooks and journal papers. Non-normative sources are marked as such in their entry.

Each entry names the printed edition and the exact location, quotes what the document says, shows why it cannot be right, gives the independent evidence, and states which reading the library implements. Where that reading changes a number the library reports, the entry names the check or regression test that pins it; where the defect is a label, a cross-reference or a table the library never reads, the entry records that no change was required. A defect listed here is never a defect of the method: in every case the intended reading could be established from the document itself or from physics.

Read it alongside the conformance report, which shows the numbers the library computes; the registry explains the handful of places where the printed expected value is the thing that is wrong.

Each entry lives on the page of the document it is about, one page per document family: every part and edition of ISO 3743 on one page, a book on a page named after its authors. Every entry is listed below under its document. A link to an entry written as /reference/errata/#entry, from before the registry was split, still lands on it.

The registry is maintained in docs/ERRATA.md and its entries are transplanted into these pages at build time by make site-reports, so the two cannot disagree.

Status legend: unreported (recorded here only) / reported (submitted to the issuing body, with date and reference).

A claim that turns on the exact characters of a formula, constant, coefficient, symbol, inequality or table cell is verified against the page as printed, and its Evidence bullet cites that page by PDF page index and printed folio. Extracted text may locate a page; it is never quoted as “the print”, because PDF text layers delete glyphs silently (most of the sources cited here emit no √ at all, so f_T/√2 extracts as f_T/2). The page offset of each document is established empirically, because it differs per document and drifts between chapters of the same book. Entries that rest on something else, a recomputation or a comparison of two sentences, say so either in a leading notice or on the allowlist of scripts/check_errata_evidence.py, which is the check that enforces the rule; see CONTRIBUTING.md.

A Spanish edition of this registry, translated entry for entry, is maintained in ERRATA.es.md. The wording here is the authoritative one, and quoted print, mathematics and printed values are reproduced there untranslated; make site-reports holds the two editions to the same entries in the same order.

One page per series, which gathers its parts, its editions and its national adoptions.

DIN 4150 16 entries

One page per document, named after its authors or its issuer.

Properties of the sources that look like misprints and are not, recorded so that nobody corrects the library to match them.