Skip to content

docs(evidence): P-014 — record owner ruling (GO-P014-BOUNDED-DESIGN / NO-PRODUCTION) #420

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
PhysShell wants to merge 1 commit into
base: main
Choose a base branch
from
Open

docs(evidence): P-014 — record owner ruling (GO-P014-BOUNDED-DESIGN / NO-PRODUCTION) #420

Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
38 changes: 31 additions & 7 deletions docs/evidence/p014-reference-coverage-2026-10/REPORT.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3,9 +3,13 @@
**Pre-registered result: `GO-P014-BOUNDED-DESIGN`.** The GO has narrow content: the measured gain in Own.NET
diagnostics is small (§5).

**Owner decision on review: `ACCEPT-REFERENCE-COVERAGE-FINDING / HOLD-FORMAL-GO / NO-PRODUCTION`.** The formal GO is
held pending the verification in §8, which was done without a new experiment and without changing rules or
thresholds. §8 corrects one claim of this report, about package identity.
**Final owner ruling: `GO-P014-BOUNDED-DESIGN / NO-PRODUCTION`** (§8.7).
- K4 is accepted as `PASS-COMPILATION-CONFLICT`, with the limit `SITE-LEVEL-CONFLICT-NOT-REPRESENTED`.
- The earlier `HOLD-FORMAL-GO` is lifted.
- The GO allows a bounded design only. No production change and no scope extension.

§8 holds the post-review verification, done without a new experiment and without changing rules or thresholds. It
corrects one claim of this report, about package identity.

Rules: [`00-preregistration.md`](00-preregistration.md), commit `2e73767b` (04:43 +05; the "04:45" in its header is
my clock reading, and the commit time is the authoritative one). Own.NET `main` `c6e8a3c1`; extractor built offline
Expand Down Expand Up @@ -228,7 +232,8 @@ Against the pre-registered K4 ("recorded as ambiguous/conflict, never `RECOVERED
- **Never `RECOVERED_CORRECT`:** yes, because A == B and no recovery is claimed.
- **Not recorded at site level:** the per-site binding silently picks the first definition.

Whether compilation-level registration satisfies K4 is the owner's call. The facts above do not change with it.
The owner ruled on this as `PASS-COMPILATION-CONFLICT`, with the limit `SITE-LEVEL-CONFLICT-NOT-REPRESENTED`
(§8.7).
All seven control summaries are now in `data/controls/K*.summary.json`.

### 8.2 NuGet SHA-512 comparison
Expand All @@ -248,7 +253,11 @@ computed SHA-512 (base64) of the `.nupkg` file myself:
- The assets' `sha512` is NuGet's package content hash, the same value NuGet records in `.nupkg.metadata`. It is
not the hash of the `.nupkg` file, so my original comparison was wrong. All 231 present packages match their
assets entry, and **package identity is confirmed**, not approximate.
- **Not checked by me:** why the content hash and the file hash differ. Package signatures are the likely reason.
- **Why the two hashes differ** has a documented cause in NuGet's package-signing design: the file hash covers
the whole `.nupkg`, while the content hash is computed over the package content without the signature data.
- **Not done:** an independent byte-level recomputation of that normalized content hash. The evidence is the
agreement of the assets' values with the content hashes NuGet itself recorded in the cache (231/231). That
confirms the content hashes are consistent; it is not a recomputation.
- **Package sources** (`.nupkg.metadata` `source`): P4 and P5 are all, or all but one, from nuget.org. For P1–P3,
46, 2 and 2 packages came from another source. The source addresses are not published here (D13). Identical
content hashes make the content the same whatever the source.
Expand Down Expand Up @@ -297,8 +306,23 @@ reference set than R2's binding numbers, and the two sets of numbers are not one
- Criteria 1, 3 and 4 hold. Criterion 1 holds on the exact projects alone (P4, P5), and criterion 4 now holds
without the identity caveat.
- Criterion 2's controls hold, with K4's conflict recorded at compilation level only (§8.1).
- The pre-registered result stays `GO-P014-BOUNDED-DESIGN`. The owner's `HOLD-FORMAL-GO` applies until the K4
reading is decided.
- The pre-registered result stays `GO-P014-BOUNDED-DESIGN`. The owner's ruling on K4 is in §8.7.

### 8.7 Owner ruling (recorded after the merge of #419, `ad17d7b2`)

`GO-P014-BOUNDED-DESIGN / NO-PRODUCTION`.
- **K4 = `PASS-COMPILATION-CONFLICT`.** CS0101 satisfies the pre-registered conflict registration, and no false
`RECOVERED_CORRECT` was claimed.
- **Limit kept: `SITE-LEVEL-CONFLICT-NOT-REPRESENTED`.** The shared `CSharpCompilation` still binds `Dup.T` to its
first definition. This is not evidence of correct project isolation.
- **Scope of the GO.** It allows a bounded design only, given the limited diagnostic effect: 1 actionable TP, 2 minor
TPs, 6 FPs. No production changes and no automatic scope extension.
- **First design must cover:**
- an honest `binding coverage` metric;
- the bounded applicability of `project.assets.json`;
- mandatory control of false OWN001/event findings.

NuGet references go into the extractor only after that design.

## 9. Files

Expand Down
Loading