The five
How often the whole state is saved, how often holdings are marked, how long an unclear send waits before a warning, how often such sends are looked over, and how often a mark is sampled. Each has a figure standing behind it, used where the setting is not given.
What the reading insists on
Absent, or holding nothing, and that figure stands in. Holding anything that will not read as a number, and the start is refused, the refusal naming the variable, what it held, and the shape expected. Holding a figure that is not finite, or one at or below nought, and the start is refused the same way.
So an endless figure is turned away as firmly as a misspelt one, and a wait of no time at all with it.
The one held tighter
The sampling cadence is then refused unless it is a whole number of seconds, and handed back as a whole number. The other four take a fraction.
How a cap differs
A cap is read from the same environment by a reader of its own, and the two part twice. A cap whose setting holds nothing is answered with nothing, where a timing setting is answered with its figure. And a cap of the wrong shape refuses with a failure of another kind, so anything sorting failures by kind sees the two apart.
Related
Source evidence
---
row: P117-01
baseline: 49aa659
created: 2026-11-09 09:12 UTC
modified: 2026-11-09 09:12 UTC
evidence:
- claim: "A timing setting is read from the environment, and a default stands in where it is absent or holds nothing"
source: "`launcher.py` 227, then 228 into 229"
- claim: "A value that will not read as a number refuses the start, the refusal naming the variable, what it held and the shape expected"
source: "`launcher.py` 231-232 into 233, then 234-238"
- claim: "A value that is not finite, or that stands at or below nought, refuses the start the same way"
source: "`launcher.py` 240 into 241-246"
- claim: "Otherwise it is handed back as given"
source: "`launcher.py` 248"
- claim: "Five settings are read this way: how often the whole state is saved, how often holdings are marked, how long an unclear send waits before a warning, how often such sends are looked over, and how often a mark is sampled"
source: "`launcher.py` 4591-4594, and 4601-4604, and 4619-4622, and 4623-4626, and 254-256, the defaults at 207-211"
- claim: "The sampling cadence alone is then refused unless it is a whole number of seconds, and handed back as a whole number"
source: "`launcher.py` 258 into 259-263, then 265"
- claim: "A cap read from the same environment is answered with nothing where its setting holds nothing, where a timing setting is answered with its default"
source: "`launcher.py` 696 into 697, against 228 into 229"
- claim: "A cap of the wrong shape refuses with a failure of another kind"
source: "`launcher.py` 705-707, against 240-246"
---