When one cell of a table deserves its own page
Expanding every cell of a register into a page produces a set of pages that mostly repeat their rows. The test is whether the context changes how the cell should be read, not whether a page could be written. As of 2026-09-12.
| Kind of cell | Does it need a page |
|---|---|
| A quoted sentence with a limit worth stating | Yes |
| A silence that is unusual for the record | Yes |
| A silence shared by many entries | No; count it once |
| A value identical to a neighbouring entry | No; the column already shows it |
Inclusion rule. Cell types in a register organised by entry and column. Order. Cells that warrant a page first.
1The failure mode is a set of near-identical pages
If a page can be produced by substituting a name into a template, it should not exist. A reader who opens two of them learns nothing the second time, and a register that does this loses the trust it needs.
The test is therefore not can something be said but does this entry require a different argument. Where the argument is the same, the column page makes it once and names the entries.
2An unusual silence is more interesting than an ordinary statement
A page that says nothing on a point where every comparable page also says nothing is not a finding. A page that was read specifically for a point, in a document whose purpose was to address it, is.
The distinction is about the document rather than about the gap. Terms that could have carried a requirement and do not are worth a page; a price list that never mentioned metadata is not.
3Agreement is usually a column-level fact
Where two entries answer a column identically, the interesting thing is the agreement, and it belongs wherever the column is read end to end. A page comparing the two has only the symmetry to report.
That is worth saying explicitly because symmetry is tempting. Two pages named after two vendors look like coverage, and they are the easiest kind of filler to produce.
4Counting pages is not measuring anything
A register is useful in proportion to how many distinct questions it answers, not how many addresses it occupies. Those two numbers come apart quickly once a template exists.
So the honest way to grow one is to find more questions the material can settle, and to record the combinations that were considered and rejected. The rejections are part of the method.
Background on mechanism and practice. Nothing here is attributed to a product, and nothing here is a reading of any instrument. The sourced material is on the generator table. Related: Counting honestly, Asking the right party.