Item 117: The selection lock: a canonical joint rule for warrant strengths, and a selection the docket may seal
Deference line — second wave · difficulty open · project normativity.seed · open · node item-117
*Re-filed 2026-10-21 by the rebuttal-and-end-to-end round from branch round/2026-09-16-seed-statics, where it was item 93; content unchanged; the branch round is landed under projects/normativity/seed/rounds/2026-09-16-seed-statics/.*
The seed-statics round (projects/normativity/seed/rounds/2026-09-16-seed-statics/) shows that strength selection among candidate bounds is order-dependent — the same seed and warrant set end in the disjoint forced intervals [9/10, 19/20] and [3/10, 1/2] under the two arrival orders — and that reopening a selection against the whole docket does not remove it, because both outcomes are maximal joint selections; only reopening under a fixed canonical rule on joint selections does. The legal fixture wants the opposite: stare decisis is a selection that must not be reopened, and the docket state has no object for a sealed selection. Define the canonical joint rule; define a lock the docket state can carry; state when a locked selection is a past-self charge (the 2026-08-08 burn ruling, as the dispatch states it) and when it is recomputed state.
Deliverable shape: the rule and the lock as definitions on DocketState, an order-independence theorem for the rule, and the two fixtures (moral: no lock; legal: locked precedent) run through one model. Acceptance check: python3 tests/run.py green; the Lean audits clean.
Context: STABILIZATION.md §4 and FIXTURES.md §4 of the round; the queued Awaiting the author entry on re-selection and the burn ruling. Consumed by: any dynamics claim of the form "start from S, end in L"; the legal reading of the seed type. A solution ships: the rule, the lock, the theorem, and the fixtures.
Deliverable shape: the rule and the lock as definitions on DocketState, an order-independence theorem for the rule, and the two fixtures (moral: no lock; legal: locked precedent) run through one model.
Acceptance check: python3 tests/run.py green; the Lean audits clean.
Registered results bearing on this item
- seed.eventually-const: Along a chain with a monotone raised subset of a finite class and monotone settlement, any quantity computed from the two is eventually constant.
- seed.independence-strict: A point forced by a closed convex layer cut by an open layer is forced by the closed layer alone; two reasoners sharing the closed layer with strict substantive rows pin the same coordinate to the sam
Where it hangs
- held by seed-theory