# Global Ring of Trust — Evidence and Conformance Checklist This checklist is a proposal for future implementers. It does not certify this repository or any OS developer. ## A. Claim clarity - [ ] Every claim has a named subject, issuer, type, scope, issue time, expiry, and policy version. - [ ] Project identity, maintainer identity, source revision, artifact digest, build provenance, and independent assessment are represented separately. - [ ] Claims state evidence, assurance level, limitations, and the method used to verify identity. - [ ] “Trusted” is never used without explaining what was checked and when. ## B. Release integrity - [ ] Release digests are computed with a current, explicitly named cryptographic hash. - [ ] A signature is bound to the exact manifest bytes and a declared signing-key identifier. - [ ] Key lifecycle documentation covers custody, access control, rotation, revocation, and compromise response. - [ ] Build provenance identifies source revision, builder identity, toolchain, build parameters, and output digest. - [ ] Reproducibility results and exceptions are published with their exact scope. ## C. Governance and review - [ ] Root and policy governance is plural, transparent, and has documented quorum and change history. - [ ] Reviewers disclose conflicts and publish assessment scope and limitations. - [ ] Membership suspension, removal, correction, and appeal have documented procedures. - [ ] Revocations are timestamped, publicly discoverable, and monitored for propagation. - [ ] No member or trust-root operator can grant OS-level privilege through membership. ## D. Privacy and accessibility - [ ] Personal data collection is minimized and retention is specified. - [ ] Public evidence is redacted where disclosure is unnecessary for verification. - [ ] Users can understand the difference between identity, provenance, assessment, and runtime permission. - [ ] Verification failures, unknown status, and service outages are displayed distinctly. ## E. OS-level-support implementation gate Before adding any trust-statement code to the C/C++ library: - [ ] Choose a published, interoperable data format and supported signature profiles; do not invent cryptography. - [ ] Define canonicalization and exact signed bytes; reject ambiguous encodings and duplicate fields. - [ ] Add size, depth, and resource limits for untrusted input. - [ ] Separate parsing, cryptographic signature checking, identity binding, revocation lookup, and policy evaluation into distinct result states. - [ ] Make network access, if ever required, explicit and caller-controlled. - [ ] Do not download/install trust roots automatically. - [ ] Do not execute code, load drivers, install hooks, change system settings, or elevate privileges based on a validated statement. - [ ] Test malformed inputs, truncation, expiry, revocation, key rotation, wrong scope, replay, and platform-independent behavior. - [ ] Obtain independent security review before using the result to influence software distribution or execution decisions. ## Current status All items above are proposed requirements, not completed implementation or certification. The Global Ring of Trust remains a concept pending independent governance and standards selection.