Trusted timestamps for signatures (from the transparency log, or RFC 3161) issue #240 closed: completed andymccall opened this · 4 comments
andymccall set labels security ·
Design proposed in docs/transparency.md (2026-10-17, overnight): the log (C2SP tlog-tiles, signed checkpoints, witnesses) and, from it, integrated time for issue #240. Three decisions for Andy at the end of it; nothing built yet.
referenced by 2 changes on 17 Oct 2026
Andy, 2026-10-17: docs/transparency.md accepted as recommended: one log for everything, hubs witnessing each other plus an oxy witness on a third machine, warn for a release then refuse.
Decided by Andy (2026-10-18): the third witness is oxy witness serve on plex, as a small user service; Earth's and GCP's hubs witness each other once GCP serves. That unblocks A6 (integrated time from the earliest witness cosignature).
Update from Andy (2026-10-18): narvi (the RISC-V box) is a witness too, beside plex: two independent oxy witness serve instances, on plex and on narvi, plus the hubs witnessing each other once GCP serves.
closed by change 13b27cb2200c ·
Our timestamps (HLC) are the signer's claim. To answer 'was this signed before the key was revoked or expired?' a signature needs an independent time: its log inclusion (checkpoint time, with the transparency log) or an RFC 3161 timestamp token from a TSA the site trusts. Verification of delegations, revocations and expiry then uses that time, not the claimed one.