GateTrue

What each generator documents about marking its output

Counts and patterns the register can support

A finding here is a count or a pattern that holds across the register and can be checked row by row. Each page names the rows it rests on, so a reader who disagrees can find the one that breaks it. As of 2026-09-22.

What a count on this register is a count ofEvery total here counts published statements rather than products. Nine grids that omit a marking line are nine grids, and turning that into nine marked free tiers would convert a claim about publishing practice into a claim about files that no page supports.A claim about pagesA claim about filesWhat is countedSentences, omissions, silencesBehaviour of outputEvidence neededThe vendor's own pageRunning files through somethingProduced hereEvery finding on this siteNothing on this siteIf wrongA row can be re-readWould need a test benchWhy the counts are phrased the way they are
Fig. 1 The distinction decides what a finding is allowed to say, and every page here states which of the two it is claiming.

The questions pages on this site answer a reader's question in a sentence. These pages do something narrower: they state a property of the register itself, of the form so many rows do this and here they are. That is a weaker kind of claim than a vendor statement and a more useful one for anyone deciding how much of this subject is actually documented.

Every count here is a count of published statements, never of products. Nine grids that omit a line are nine grids, not nine marked free tiers, and a page that says so is making a claim about publishing practice. Where the distinction changes the answer, the page says which of the two it is claiming.

Counts move when the register does. Each page carries the date the count was taken and names its rows individually rather than asserting a total, so adding a generator next month makes the page visibly incomplete instead of quietly wrong.

1The findings

  • Answers in a sentence — Four vendors write their free-tier behaviour out; the rest leave it to be read off a grid.
  • Crossing points with a price — Seven entries name the plan where the mark stops and print a figure beside it.
  • Crossing points with no number — Three entries place the end of the mark on the paid side and leave the figure unquotable.
  • Standards that get named — Two standards are named across the register, by the entries whose documents are technical.
  • Who can read the signal — Some detection routes are open to whoever holds the file; one is described as the vendor's own.
  • Reasons printed beside a mark — Three entries publish why the mark exists or why it may be switched off, and the reasons disagree.
  • Marks that get described — Three entries publish something about what the mark looks like; twelve name it and stop.
  • Paid plans still marked — On two entries a paid subscription produces marked output, so an invoice settles nothing.
  • Files already generated — One entry says what an upgrade does to files already made; the rest do not raise it.
  • Three kinds of empty cell — An empty free-tier cell means one of three different things, and a single dash would hide which.
  • Which columns a row fills — Rows come in two shapes: a crossing point with an empty file column, or the exact reverse.
  • Two prices for one tier — Four entries print two figures for the crossing point, and the gap reaches a third of the price.
  • Licences and marks — On one entry the commercial licence starts a plan above the marking line, which leaves a clean unpublishable file.
  • Products with two doors — Where the same models are reachable two ways, the route can decide the file and usually nobody says so.
  • Durability caveats — One vendor publishes the limit of its own mechanism; one warns publishers in general terms.
  • The spread of prices — Ten dollars to thirty for the same behaviour, which measures packaging rather than marks.
  • Readings that expire — Four values rest on a line that is absent from a grid, so a redesign destroys the evidence.

Grouped by what is being counted: evidence classes first, then prices, then documents, then the shape of the silences.

Elsewhere in the register: Generators, Fields, Readings, Side by side, Implementers, Questions, Source documents, Terms, Learn, Data. Field definitions live on the protocol page.