Method and source policy
- Every claim carries a graded citation. The grade is rendered next to each reference so a reader can see whether a number came from a refereed paper, a NASA technical report, or a program author's own documentation.
- Code is a primary source for what a program does. Where a program is public, the note cites the file that computes the quantity.
- Formulas are quoted, not paraphrased. A note that says "the contact times are found iteratively" without the iteration is not finished.
- Negative findings are recorded as findings, with their scope. That a thing does not exist, or is not published, is a result a reader can use.
The grades, as applied here
| Grade | Used for | Example |
|---|---|---|
peer-reviewed |
Refereed papers, the Explanatory Supplement to the Astronomical Almanac, almanac-office publications | Stephenson, Morrison and Hohenkerk on ΔT; Lamy and colleagues on the eclipse solar radius |
primary |
NASA, USNO, IMCCE, JPL, IERS and PDS publications, technical reports and data products | The Five Millennium Canon technical publication; a LOLA digital elevation model release; a JPL DE ephemeris memo |
preprint |
arXiv and other unrefereed manuscripts | Ephemeris comparison preprints |
company |
A program author's documentation of their own program or product | Occult's help pages; Solar Eclipse Maestro's feature list; an SVS page describing its own method; a repository README |
trade |
Practitioner writing carrying data or method, outside a refereed venue | An IOTA member's mailing-list analysis of edge observations |
survey |
Aggregated or sampled data with a stated method | A comparison table of predictors compiled by a third party |
vendor |
A supplier describing a product it sells, where the claim is promotional | An app store listing's accuracy claim |
unsourced |
Forums, undocumented repositories, encyclopaedias | A Reddit thread; a GitHub repository with no stated method |
The rules
A program's documentation proves what the program does, never how accurate it is. Accuracy comes from comparison with observation, from a refereed paper, or from an explicit error budget.
A formula is quoted with its symbols defined and its constants named. Eclipse computation mixes units: Earth equatorial radii on the fundamental plane, arcseconds on the sky, kilometres on the ground, and three time scales. A note that leaves the unit implicit has left the error in.
An ephemeris, a ΔT model and a limb dataset are versioned, and the version is stated. "DE440", "LDEM_128", "Espenak and Meeus polynomial" are claims; "a JPL ephemeris" is not.
The algorithm and the implementation are distinguished. Chauvenet, the Explanatory Supplement and Meeus give algorithms. Espenak, Jubier, Herald and Wright chose constants and data. When two programs disagree, the note says which choice explains the difference.
Negative findings are findings. "No open-source implementation of a limb-corrected path edge exists" is quotable once its scope is stated.
The site states positions, not revisions. When a finding supersedes an earlier one, the page is rewritten.