Try a check

1. Read the draft that sets the rules: draft-wright-verdict-record, published at the IETF.
2. Download the test cases from github.com/diggertydog/verdict-record-test-vectors. Each test case is one JSON file holding the exact bytes of a record, the evidence a checker holds, the checker's policy, and the answer the draft text gives.
3. Run the reference checker from the same place on the first test case:

python3 vrcheck.py a01-baseline-released.json

It reports the state released, which is the answer the test case states.

4. Now run one built to fail:

python3 vrcheck.py d01-result-mismatch.json

It reports refused: the rule re-runs to pass, and the record states fail.

The reference checker was written by Sencor AI. The stronger test is a checker you write yourself from the draft, run on the same files. A check reaches one of four results: released, held, refused or no verdict.

Checkers written outside Sencor have run these test cases.

The test cases cover the draft's rules for reading a record and for the check itself. They do not cover signing. They were written for revision -01 of the draft.

If your checker gives a different answer from a test case, write to Darren Wright at [email protected]. To read what the standard is, see The standard.