Repository navigation
OwnTransit 0.6.1 — Recoverable setup and end-to-end readiness
OwnTransit 0.6.1 makes setup recoverable and distinguishes saved progress from a working end-to-end tunnel.
- Lost the code? New receiver setup keeps an owner-only local copy. Retrieve the same unused, unexpired code with the exact command printed by setup or VPS approval. The private code never goes to the relay or its network worker.
- Retry without starting over. Normal setup retains pending and paired identities. Creating fresh receiver identities now requires explicit
--replace; paired replacements still ask for confirmation. A new receiver ID comes with its exact new approval command. - Know what to run next.
pair nextprints commands for the selected tunnel's actual state. Missing approval, interrupted client requests, stopped services, expired codes and alarms have distinct recovery instructions. Commands appear on standalone copyable lines. - Ready means checked. Setup and resume run the real end-to-end carrier check before reporting TUNNEL READY. Relay setup/approval are only UNDER CONSTRUCTION steps.
pair checkrepeats the test without replacing the pairing. SSH login still uses your own keys and host verification. - Installers accept 0.6.0 upgrades on Linux amd64/arm64 and Apple-silicon Macs, preserving pairing and SSH settings. Existing release bytes remain immutable.
Existing lost codes: versions 0.6.0 and earlier did not retain the private code. An upgrade cannot reconstruct it from a hash. The new retrieval command explains the explicit one-time replacement and new-ID approval when needed; it never silently resets your pairing.
The wire protocol, encryption, endpoint authority and fixed SSH destination remain unchanged. The private cache is validated against authoritative pending state and cannot revive a consumed, expired or alarmed pairing. No independent security certification is claimed.
Verification includes both race/vet profiles, source/security checks, lost-code recovery followed by real SSH through the integration fixture, negative cache/identity/permission tests, real Linux CLI retry and explicit-replacement dispatch in an isolated container, and isolated Mac installer upgrade/reinstall/removal checks. No new live-deployment end-to-end result or pristine-host qualification is claimed.