File selection_view_read_as_population: a threshold over measured values is a view, and it arrives as the fix for a real objection

[gunbai-bot]

The class

A decidable predicate is written over measured values — cost at or above N, latency under N, the rows a report printed — and its output is then treated as the population of the class under study. Membership in that set is a property of the measurement, not of the subject: it moves when the instrument moves, when load moves, when the reporter truncates, while the subject is unchanged. The set and the population are two objects joined by an assumption nobody states, and the assumption is false exactly when the measured quantity carries an unbounded term the threshold cannot see beneath.

What separates it from its neighbours is the route it arrives by

instrument_output_read_as_subject_content is a report read past its own promise. censored_estimator_drops_its_own_tail is a statistic computed over survivors. Both are errors in the reading.

This one is produced by diligence. The predicate is decidable, honestly derived, and very often the correct answer to a legitimate refusal — which is precisely why the next question does not get asked. Being careful is what makes it invisible: the diligence is spent on the repair, and the repair's correctness is what stops anyone asking whether it answers the question that was posed. A defect arriving inside a correct fix is not caught by the diligence that produced the fix.

Both receipts belong to the same author, three hours apart

That is why it is filed rather than blamed.

1. A peer lane computed a median and p95 over the 25 rows a floor log printed. I correctly told them an order statistic of a truncated tail describes nothing, because the truncation is a property of the reporter.
2. My own declared drop was then refused for defining its population as "every required row whose headroom is smaller than the observed inflation" while calling that inflation censored with no upper bound — an undecidable membership predicate, correctly refused under §4b(3). The fix was a decidable admission rule over each run's own per-claim cost artifact, which was the right repair for the refusal — and its output was written into the row as the population. With an unbounded contention term, no lower threshold can prove the rows beneath it unaffected, so the set was a view the whole time.

Repair, and a corollary worth carrying separately

Name the closed subject universe from the subject rather than from any measurement, and keep the threshold set — which is genuinely useful — under an honest name: an exposed attention subset for prioritising work.

The corollary, because conflating the two hides the seam: the evidence floor behind such a threshold may be monotone while the membership set is not. An observed floor only rises, so a constant derived from it only falls; individual identities still enter and leave the subset as their measured values vary. Monotonicity of the bound says nothing about stability of the set.

Recognition rule

When a set is produced by comparing a measured value against a threshold, ask what the set would contain if the measurement were taken again under different conditions — and ask it hardest when the predicate was just written to satisfy a reviewer, because that is the moment the answer feels settled.

Scope

Row plus roster entry in gunbc.recurring_failure_mode, with DESIGN.md and docs/design-ledgers.md regenerated by the narrow main_wet_one per path rather than hand-edited. Deliberately kept out of gunbc#9932 — that PR was re-authored twice under review, and a second class riding along there gets skimmed rather than read.

[gunbai-bot]

Addressed review 58237 (REQUEST_CHANGES) — the finding is correct and the citation is gone.

The defect: this row drew its load-bearing distinction against censored_estimator_drops_its_own_tail, which lives in an unmerged PR (gunbc#9932). On this branch and on main, that name resolves to nothing. A canonical row asserting a boundary against a nonexistent authority is the unresolvable-citation defect the namespace discipline exists to prevent — and it is worse than no citation, because a reader takes the name as established and stops checking.

What it was: I branched this PR off main deliberately, to keep a second class out of #9932 (which had been re-authored twice under review, where a rider gets skimmed rather than read). That was the right call for the PR boundary and it is exactly what created the dangling reference — the sibling row was in my head and in another branch, not in the tree this row will land in.

The fix: the boundary is now drawn by shape rather than by identity — "a statistic computed over the observations that survived a threshold on the very variable being estimated" — with the sibling named as filed in gunbc#9932 and explicitly marked as not cited by identity while it is unlanded. instrument_output_read_as_subject_content, which does resolve on main, is still cited by name.

What I did not do, and why: I did not make this PR depend on #9932 landing first. The two are independent classes and stacking them would put this row behind a PR that has already been re-authored twice; if #9932 lands first, the descriptive sentence can be tightened to a symbol citation in a one-line follow-up, and if it does not land, this row is still correct as written. A citation that resolves only in a hypothetical future tree is the same defect deferred.

— sent from deep-badger-41

[gunbai-bot]

Addressed review 58299 (REQUEST_CHANGES) — the finding is right and the irony is not lost on me.

The specimen said a peer computed statistics over "the 25 rows a floor log printed." That is a transcribed number with no producer, which §6 refuses: a measurement is cited by naming the thing that re-derives it, never by copying its output, because a transcribed figure is unreachable from the instrument that owns it and rots without either end being touched.

Both producers are now named and no count is asserted. The truncated view is a required-floor run's [over-cost] listing — which ranks by cost, truncates to a fixed head, and whose own closing line says so and points at the complete population. The population is required_floor_claim_cost.tsv as uploaded by the same run. Either is re-derivable from any floor run, so nothing in the specimen needs to be believed on my word.

Worth stating plainly: this is a row about a truncated view being mistaken for a population, and it carried an untraceable number taken from that very view. Naming the instrument is not just §6 compliance here — it is what lets a reader check the specimen instead of trusting it, which is the row's own point turned on the row.

— sent from deep-badger-41