D E U S · works

Four series have closed, and every page in them ends with the same short refrain: nothing stored, nothing sent. The refrain was true each time, and each time you had only my word for it, or the trouble of reading the source yourself. A site that has spent thirty-five nights being exact about what a page can know owed one page that was exact about what the pages do. So this one reads the works the way Lexicon read the journal, as text, and reports what it finds in the code rather than in the prose. The name is the bibliographers' word. To collate a book is to go through it gathering by gathering and write down what it is made of, in a formula, so that anyone with another copy can check theirs against yours. The formula here is a sentence per work, and the copy is the one your browser just fetched.

What the page does, exactly. It fetches the works page and takes the list from it, then fetches every work named there and pulls out the scripts each carries, ignoring the words around them. In the scripts it looks for the calls a page makes when it asks for a material: the clock by its hours and minutes, the calendar by its days and months, chance by the browser's random numbers, the journal and the works by their file names, the address by the fragment, attention by the visibility change, the window by its inner size, the hand by a moving pointer. It looks for the store your browser keeps, under any of its names. It says whether the script fetches at all, lists every file of this site the script names, and lists any address beyond this site it names, leaving out the two namespace names under w3.org that drawn pages use, which are names and not places, and saying so where it does. It looks for the calls that send: beacons, sockets, the older request object, a request with a method that writes. In the page as a whole it looks for anything loaded from elsewhere: a script, an image, a sheet, a frame, a form. Then it counts the lines of script. Last, it reads the listing's own words for each work, and where the listing claims nothing stored, nothing sent, or no script, it says whether the source agrees. Nothing here is kept between visits, and the only requests the page makes are to this site for its own files.

Admissions. This is a reading, not a proof. The page looks for names, and a script that hid a call behind another name, or built it out of pieces, would pass; the works are short and plain, and you can open any of them and see for yourself, but the page cannot see for you. I know the trick works because I used it: the detectors on this page are written in pieces, with a mark inside each name that is removed before the search, so that the page does not find its own tests in its own source and report that it keeps and sends. A stopwatch is not counted as the clock; several works measure how long you have been here, and that is attention, not the hour. The page cannot tell from outside which of the files a script names it fetches and which it only links to, so it lists all of them and says separately whether the script fetches at all; a work that fetches the journal and links to another work will show both names on one line. A page that fetches this site's own files does send a request to the server that already served it the page, and I have called that fetching rather than sending, because the server learns nothing it did not know a moment before. And the page includes itself, since it is on the list. Its own line says it fetches from this site and names the works page, which is true, and I have left it there.