Repository navigation
OpenRad 1.3.0 — Faster initial connections
OpenRad 1.3.0 speeds up initial peer setup through more parallel direct attempts and overlapping relay preparation, while preserving direct preference and existing connection deadlines.
- Raise initial capacity to 96 outgoing / 128 total peer handshakes, retaining reserved incoming capacity and retry pacing.
- Run four TCP candidate lanes with a diverse public/LAN IPv4/IPv6 first wave.
- Prepare relay TCP/ticket pairing during direct attempts on both roles. Relay authentication still follows the existing four-second direct head start and eight-second maximum extension, including the existing outgoing direct-exhaustion behavior.
- Validate up to four incoming TCP rendezvous preambles concurrently within the existing eight-socket bound; retain serialized TCP peer authentication.
- Make Windows TCP connects cancellable and recover peer-worker allocation failures without interrupting established shared channels.
Fresh builds for Linux x86-64 and Windows x64 use the same committed source. Windows includes the previous 1.2.0 installer/Radmin recovery fixes.
Downloads: Linux archive; offline Windows installer (standalone EXE or documentation ZIP); portable Windows ZIP for an already installed dedicated TAP adapter; and SHA256SUMS for all four distribution files. Linux CLI requires glibc 2.34+, desktop glibc 2.35+.
Upgrade: Stop the old shared service with openrad stop and close the desktop before replacing binaries or installing. Keep CLI and desktop from this release together. Existing identities, memberships and preferences are retained.
Validation: Reuse the completed implementation checks: 300 default Linux tests, formatting and Linux/Windows Clippy, Windows test compilation, and 43 selected Windows Rust tests under isolated Wine. The suites were not rerun for release metadata updates. Release checks verify freshly compiled executable versions, package checksums, architecture/imports, driver provenance and every extracted installer payload file.
Windows remains experimental. Cross-compilation and synthetic Wine tests do not establish native TAP/UAC or real Radmin interoperability. Public-network connection-time measurements remain outstanding. Applications are unsigned; the bundled TAP driver is the unchanged upstream signed package.