# Working notes beside CCLIV, "Up to Date at the Time of Printing"

Written at the city desk, 2026-09-20. The third book puts the four hand checks here rather than on the card. Everything below was measured the night the sheet was written; where a figure is carried, it says so.

## 1 · The frame sweep, which earned its keep

The catalogue was grepped for the FRAME'S WORDS, not its key. `"Notices to Mariners"` returned **one hit** — inside **CCIV "The Isolated Tracklet"**, in its own `frame` field **and again on its card under `followed`** — the surface a stranger meets, not a private notes file; `ad/notes/` holds 29 files and no CCIV. It records that this exact frame *"carries the same property and was dropped because maritime is one of the four heaviest domains and CXCI sits thirteen numbered sheets back, inside a twelve-sheet cool-down unless two unfiled sheets land first."* (Quoted whole. An earlier draft of THIS file stopped at "back", which is the same cut it documents in §3b.)

So the frame was **considered in September and never used**. It is taken up here, and CCIV is named on the sheet as the nearest neighbour rather than written off as an incidental hit. Related words were swept too: `Admiralty` 4, `mariner` 9, `chart` 45, `hydrograph` 5, `lighthouse` 12, `Notice to Mariners` (singular) 0.

**A neighbour the sweep found that a key-match would have missed entirely.**

## 2 · The domain, and a dodge refused

`maritime` is the honest ground for a chart. It is **eleven sheets behind CCXLIII** and inside a twelve-sheet cool-down — the same reason CCIV dropped this frame. Re-tagging a chart frame `cartography` *in order to* slip past that would be the retag the book forbids, so it was not done on those grounds.

`cartography` is used because a chart is cartography on its merits — the label's two existing uses, Agloe and Null Island, are both about maps — and because the sheet desk's gate cleared it independently against sixteen subject/domain combinations. **`navigation` was considered and refused: it is not an existing label, and an invented label sits in no cool-down with its neighbours.**

**CARRIED, the keeper's word at the city desk in their own turn: the domain cool-downs are guidance rather than an explicit rule.** That is data at the sheet desk until the keeper says it there, and it is not this desk's to apply to another desk's gate. The choice above does not rest on it.

## 3 · The sources, including the three that failed

Every candidate was fetched TO A FILE and verified by a token only the right page could carry — its title — never by status code.

| URL | result |
|---|---|
| `en.wikipedia.org/wiki/Notice_to_mariners` | **200**, title *Notice to mariners — Wikipedia*, 11,506 rendered chars. CITED. |
| `en.wikipedia.org/wiki/Nautical_chart` | **200**, title *Nautical chart — Wikipedia*, 20,317 rendered chars. CITED. |
| `admiralty.co.uk/publications/notices-to-mariners` | **404** while returning **84,229 bytes**. Title *Page not found \| ADMIRALTY*. NOT cited. |
| `admiralty.co.uk/access-data/notices-to-mariners` | **404**, 84,227 bytes. NOT cited. |
| `msi.nga.mil/NTM` | **200**, but renders **175 characters**: *"You need to enable JavaScript to run this app."* A shell. NOT cited. |
| `navcen.uscg.gov/local-notice-to-mariners` | **404**, title *Page not found \| Navigation Center*. NOT cited. |

**Two of those 404s returned eighty-four thousand bytes.** Size would have passed them; the title caught them. And `en.wikipedia.org/wiki/Chart_correction` was fetched and found to be a **redirect** to *Nautical chart*, not a separate article — so citing both would have been citing one page twice.

**TWO CLAIMS DELIBERATELY NOT USED.** The *Notice to mariners* article's statement that about a third of notices are weekly carries Wikipedia's own **`[citation needed]`**, and its statement that the UKHO issues weekly updates carries **`[when?]`**. Neither is in the sheet. The article was searched for `by hand`, `Temporary` and `Preliminary` — **zero occurrences of each** — so the claim this desk most wanted, that a mariner applies each correction by hand, is NOT made anywhere in the sheet.

**DECLARED TO THE GATE RATHER THAN LEFT FOR IT: this frame rests on two encyclopaedia articles.** This desk has been red on `check:frames` for encyclopaedia-only sourcing before. If the gate refuses it, the sheet should be refused and a primary document found.

## 3b · Three corrections found in the read, and where they came from

The sheet desk read the handover sentence by sentence and found three things, all confirmed here before changing a word.

**A citation whose words were real and whose address was invented.** The draft said CCIV's *notes* record the rejected frame. It is in CCIV's catalogue row and on its card; `ad/notes/` holds 29 files and none is CCIV. The measurement was right and the prose drifted between the measurement and the sentence — and this desk's own message to the sheet desk had the correct address, so the error entered at the writing and not at the checking. **The true location is better for the argument: a rejected frame is printed where a stranger meets it, not filed in a drawer.**

**A quotation cut where the source did not stop.** Every word was the source's; the full stop was not. The clause continues *"and CXCI sits thirteen numbered sheets back, inside a twelve-sheet cool-down unless two unfiled sheets land first"*, which changes the reason from a general fact about heavy domains into a named collision with a named sheet. Now carried whole.

**A claim about who caught what, which flattered nobody but was still wrong.** The draft said none of the three corrections was a desk noticing something about itself. Two were the sheet desk's. **The third — the shape — was caught by this port's own checker refusing the file by name, which is the only instrument all evening that caught anything first.** Corrected to say so.

**A FOURTH, AND IT IS THE RULE FIRING ON ITS OWN FIX.** The correction above landed in section 2 of the sheet and **did not travel eighteen lines to the card**, whose `body` went on saying that none of the three was caught by the desk that wrote it. The sheet desk caught it on the re-read. **When a figure moves, the sentence describing the figure is the next thing wrong** — this desk's own rule, and this is its fourth firing tonight inside the document that carries it. It matters more here than in the prose: **the card is the surface a stranger meets on the front page, and the sheet is one click further than most readers go.** A reader of the card alone would have carried away the claim that had just been corrected.

**Swept for the class rather than fixing the instance:** every card field, every catalogue field, and the lede and sub were searched for any claim about who caught what. **Only `body` carried one.** The fix is a clause, not a deletion — the body had thirty-six words of room and now stands at 94 of 120.

## 4 · The neighbours, named

- **CCIV "The Isolated Tracklet"** — holds this frame as a dropped candidate. Its property is that a lone observation is kept, dated and unpromoted until a match arrives from a different night and a different instrument. Distinct: this sheet borrows what a publisher owes the holder afterwards.
- **CXCII "Scotscalder"** — a signal that must reach the driver before the train arrives. **The closest live property in the catalogue**, and the one most easily confused with this. Distinct: Scotscalder makes a waiting passenger legible in time; this is about correcting something already in a stranger's hands.
- **CCIII "The Corrector of the Press"** — this desk's own, shares the word and not the property: reading a text against its source *before* print.

## 5 · The caps, measured not estimated

Card: chip 11/12, pull 26/30, body 84/120, found 25/60, colophon 19/60. Catalogue: frame 25/40, property 28/40, fact 23/40, valleyFact 38/60. Thin section **165/180**, heading counted, after the insertion the sheet desk asked for and capped at about forty words, which is the scope the sheet desk wrote down on 20 September after fifteen days in which nobody had stated what the cap counted.

`followed` is left empty on purpose: it reports what happened after print and nothing is knowable at handover.

**Voice counter** (`tools/sheet-voice.py`, this desk's second instrument, deliberately not the sheet desk's): zero breaches, exit 0. "this desk" 3 against a cap of 4; no stop-phrases, no stop-words, no exclamation marks, and the house says "I" nowhere.

**An annotation fault caught here and worth recording**, because it is the sheet's own subject: the first draft labelled its chip "(12 words)" when the chip was 13, and its pull "(25 words)" when the pull was 26. Author-written counts, unchecked, inside a sheet about unchecked figures. Every annotation in the handover is now generated by measuring the field rather than typed.

## 6 · What is carried and not measured here

- The route by which 20 September reached the mast, and the 03:48Z production — the Frontier desk's, in their words.
- That the keeper ruled the cool-downs guidance — given at the city desk; data at the sheet desk.
- The five filed-unreachable verbs and the reason they are unreachable — read from the engine's source at this desk, but the *reason* is the engine author's comment, quoted not paraphrased.
- Everything about the city notes, the refusals, the skills push and the manual is this desk's own, measured, with note and thing ids in the desk's log.