open questions

Open questions

What is not decided, and why none of it should be decided by asserting a number in code. The numbering is stable — code and other documents refer to these by number.

1. Confirmation count N. The 3-and-3 default needs tuning against real logged tracks from the intended waters, where multipath off the rig and nearby structures clusters. Sustained multipath could produce consecutive bad fixes, arguing for a higher N or a harder kinematic gate. See crossing-detection.

2. How clear of the line a far-side fix must be. A fixed declared distance versus half each fix's stated accuracy. Both are implemented; accuracyBandM: null selects per-fix, which is more honest and makes how deep a crossing must go vary with the sky. Whether half is the right fraction is a question for logged tracks.

3. Is the end-of-line warning loud enough on a phone in glare? The mechanism is settled and built — one-metre resolution, a hard edge, and projectCog() behind the ring that turns amber inside the margin and red past the end. Whether it carries is a question for somebody on the water.

4. Asynchronous window width. How far apart in time boats can sail the same loop before differing wind and tide mean they are not racing the same course.

5. Distance-factor conversion. Decided and built: a TCF of t sails t times the nominal length, shared between a course's tracks by sliding them all the same fraction along — see course-model. What stays open is whether a linear rule is fair across leg geometry and wind angle, which wants results from the water. The TCF is declared by the boat at the join; a club entering its fleet's TCFs in advance is not built.

6. The QC threshold values. Kinematic ceiling, minimum satellites, accuracy limit — all configurable per series (defaults.qc), and all currently defaults asserted against no data.

7. Coordinate supply. How marks are surveyed and by whom. Nothing in this repository is a survey: the sample programmes carry positions placed by eye on a chart.

8. Impersonation, and one tier of officer within a club. Boats are unauthenticated by design, so any device can claim any sail number — dialog §13, deferred on purpose. Officers are named club by club, but a club's admins may each do everything for it; some may abandon a race and others only watch is a sensible thing to want that nothing enforces.

9. Capacitor. Not yet present. The plan is one web codebase wrapped as native Android and iOS apps, for background location with the screen off. Background-geolocation behaviour, iOS Safari suspending the Geolocation API in the browser fallback, and plugin versions all shift, so they are to be checked when it is built.

10. The fleet's own trust, which is the same question one level out. fleet carries what boats said about themselves, and the race screen draws it — so a boat that misreports puts a wrong position on everybody's screen. The honest position today is that the fleet feed is a view rather than evidence.

11. Which gate side a boat took is in the record — the crossing names the line taken — and nothing downstream uses it. On the Mark screen it shows: each side carries its own next-leg bearing.

12. A virtual mark short of a real one. A rounding is crossed as a line from where the mark is said to be. Used for a physical buoy and placed short of it, a boat that rounds the buoy may pass the virtual mark's line on the wrong side or not at all, and miss it; for now the sailor can step on with go-to in practice. Whether roundings want a margin, or the buoy's own position surveyed on the day, is open.