A candidate, not a verdict

What a row says

A row records a string that passes the spelling grammar, and the evidence for it: which of the 8projection.published_sets published word lists list it, which way it came in, whether Google Books or wordfreq holds it, and, for a row of the English OpenList, what decided it there. That is all a row claims, and each claim can be checked against a pinned source.

Why no verdicts

Word lists disagree because they answer different questions. A spell-checker's dictionary, a tournament's word list, and a corpus's vocabulary each draw the line somewhere else, and none of them is wrong by its own rule. The universe keeps every candidate its sources gave and records each source's answer beside it, so a reader can apply a definition of their own and state exactly what it is.

That is what Compose a list is for: a definition becomes a set of criteria, the criteria a count, and the count a list with the code that reproduces it.

What a row rests on

Every row falls in 1 of 4 provenance classes, by whether a published channel brought it in and whether a published list or a corpus holds it.

Even the strongest class is evidence, not a ruling: a row that every published list lists is still a candidate, and a row that only a corpus holds may still be one a reader wants.

"Valid," in the English OpenList's sense

The English OpenList calls the entries of its main list valid. The universe records that as a fact about the English OpenList at revision c1139698: its eol column is true on the 344,713rows.eol rows that list held, and it means nothing more. Elsewhere on this site, a string of the universe is never called valid.

What the universe does not record