Protocols
Platform conformance protocol, and an assessment of this repository's simulated body against it
Snapshot of this document in the GEMs repository. The repository itself is the current version. Bản chụp tài liệu này trong repo GEMs. Bản đang chạy nằm ở chính repo.
How a body demonstrates that it satisfies the platform contract.
| Document | Version | Status |
|---|---|---|
| Platform conformance protocol | 0.1 | draft |
assess.py |
— | runs the protocol against this repository's simulated body |
| Conformance record | — | generated; the body does not conform |
python protocol/assess.py # print the assessment
python protocol/assess.py --record # regenerate conformance-record.md
What this is for
Chapter 08 of the specification maps the requirements an external processing loop places on a body onto the places this body supplies them. That map is enough to design against and not enough to certify with.
The conformance protocol turns it into a test: sixteen requirements, each with the evidence that demonstrates it, and procedures for the ones a declaration cannot settle.
The shape of it
Conformance is binary. The envelope is declared.
A body either satisfies the contract or it does not — there is no partial conformance, because a loop whose preconditions are half-met does not half-run.
Separately, every body states its own figures: mass, endurance, protection, link, sensing configuration. Those are declared, not graded. The specification sets ranges and the operator picks the point, so there is nothing there to pass or fail.
Run against itself
A protocol that has never been run on anything is the kind of unchecked assertion this repository keeps removing elsewhere. So it has been run — on the only body available, the simulated one that §4 explicitly permits a claim for.
| State | Count |
|---|---|
| MET | 4 |
| MET (simulated) | 3 |
| not implemented | 2 |
| not demonstrable without hardware | 7 |
Nine of sixteen unmet: the body does not conform, and the record says so in its second paragraph rather than its last.
The seven executable requirements are not asserted — assess.py runs the procedures and reports what happened, including the pipeline-position half of §7.1 that a conclusion-only check would pass by mistake. The nine others carry a stated reason instead, and the difference is visible in every row.
The value is in what the gaps turn out to be. They cluster into four groups, each one purchase away from being testable, which converts "unfunded" from a sentence in the README into a shopping list with a priority order.
What it deliberately does not test
Whether the body's conduct was acceptable. The protocol establishes that physical amplitude at human contact is recorded; it never establishes that it was appropriate. That judgement belongs to a third party — the deploying or certifying party — and the protocol says so rather than leaving the gap to be discovered.