Method

Nine vitrines. Each one states what was measured, how it was measured, and what came back. This is a record of how the book was made and it is not a record of what is in it: no vitrine says what any measurement means about any story, and none of them ever will.

Every figure on these pages is measured at the build that produces them, from the project’s own records and from the two editions themselves. Nothing is transcribed and nothing is remembered: where a passage is quoted it is lifted from the file at build time, and if the file no longer carries it the build stops rather than printing a sentence that has come loose from its source.

I: The page that is not translated
One of the sixty stands in both editions in the same words. How the site finds it, and what the site is forbidden to do with it.
II: The same page, a taller wall
What carrying the refusal into Turkish costs, measured from the other side. Nothing repairs it, and it is not a defect.
III: What must not be corrected
Some finished features of the book look exactly like mistakes. How many were written down, and whether the file that defends them reached the hands it was written for.
IV: What the ratio is predicted by
The Turkish crosses the book at about seven tenths of the English’s words — and the compression is predicted by the English, not by the piece.
V: Forty misfires
The instruments were wrong forty times in one stage, and the files were right nearly every time. The register, its criterion, and three of the rows.
VI: The census that hid its own repair
The marker that certifies a piece as verified was measured by a limb that could not tell two kinds of bracket apart — and the gap it left swallowed the work written to close it.
VII: A count with no criterion
Three different totals were asserted about one page. They disagreed because nothing had ever published how the thing was being counted.
VIII: The seal
283 standing invariants over the Turkish edition, each of which must convict twice — and the whole stage in which not one of them was ever asked to.
IX: A door that does not open
The working record exists, it is very large, and it is not published. This vitrine measures it and then stops.

The census on this pavilion

The City Guide passes a census of resolving formulas at zero across its whole text. These pages are held to the same figure, and it is printed here rather than asserted, because a zero nobody can locate is decoration.

Formulas scanned for: 5, named in build/site_lib.py and deliberately not printed here — a page that published the five phrases would fail its own census on the line that publishes it, which is site_lib’s own reason for keeping the banned stylesheet property out of the stylesheet
Scope: the 18 vitrine pages of this pavilion, both editions, over 9,559 words of visible text
Hits: 0

The check reads visible text and cannot tell a record’s voice from the site’s, so the zero above cost something. Of the 102 rows in the misfire register, 1 carries a scanned formula somewhere in the cells and cannot be carried onto a page here whole: V-78. Vitrine I quotes that row’s truth column and names the row, so a reader can go and read the rest of it in the register. Widening the check to admit a quotation was the alternative, and it was rejected: a census with an exemption for the material that trips it is not a census.