RESULTS.md

Third-party results

Every measurement in this file was produced by someone other than the author, on a repository the author does not control, and is reproducible by anyone with the same deposit and the same commit.

That is the point of the file. The reference table in the README is the author measuring other people's repositories from the outside. This table is other people measuring their own, from the inside, where the author cannot reach — including private repositories whose contents never leave their machine, because the tool is read-only, needs no network, and the JSON report contains no source code.

How to add a line

python3 closure_drift.py --json > result.json

Open a Report a measurement issue with it, or send it to lfnsilva.invest@gmail.com if a public issue is not appropriate. That is the whole ask. It takes one command and reveals nothing about your code: the report carries counts, labels, hashes and the tool's own stamp — never file contents, never file paths outside what you chose to declare as your closure.

What you get back, and it is not a favour — it is the trade:

• your line in this table, with attribution exactly as you want it: your name, your organisation's name, a handle, or anonymous. Your call, stated when you send it;
• citable provenance for your own claim. If you publish anything that says your releases are unambiguously addressed, this gives you a dated measurement, produced by a version-pinned instrument with its own DOI, that a reviewer can re-run. Saying it is cheap; having the artefact is not;
• a straight answer about your result. If your repository drifts, which of the two known mechanisms it is — a release tagged without bumping the version file, or label collision across tag families — and whether it matters for how you publish. If it is neither, that is more interesting than a clean run and it will be said so;
• a correction, publicly, if the tool is wrong about you. A result that shows the detector misreading a legitimate scheme is the most valuable thing this file can receive, and it goes in the table marked as such, not quietly discarded.

Results that contradict the tool are as welcome as results that confirm it, and are marked in the table rather than dropped. A reference table that only ever agrees with its instrument is not evidence of anything.

What is required for a line to count

• the version DOI of the deposit used, not the concept DOI — a measurement whose instrument version is unknown cannot be checked against the report that produced it;
• the stamp block as the tool emitted it (measured_at_head, working_tree_dirty, detector_closure);
• whether the run used the default (--at tags) or --at commits, and any --closure you passed;
• a repository anyone can reach, or a statement that it is private — private results are accepted and marked private, with counts but no repository name.

Counts over repository history are functions of HEAD. A line without its stamp is a number without the state it was measured in, and this project does not publish those.

The table

Date	Repository	Reported by	Points	Labels	Worst label	Verdict	Deposit
—	no third-party results yet	—	—	—	—	—	—

This table is empty, and it is published empty on purpose. As of this release nobody outside the author has run this tool and reported a result. Leaving the row out entirely would let a reader assume otherwise; an empty table states the fact. When it fills, the fact it states changes — and that change is the only thing that would let this project claim it has been reproduced by anyone else.