https://bugs.kde.org/show_bug.cgi?id=520110

--- Comment #1 from Mice <[email protected]> ---
Additional local validation details:

I confirmed the failure was not caused by missing tailnet reachability. From
the desktop side, packets reached the Android device's tailnet IP and port
1716. From the Android side, the same connection was visible to the app process
as loopback. Once I changed the LAN backend to treat loopback as an
explicit-proxy path only when custom devices are configured, pairing completed
successfully and the desktop reported the Android device as paired and
reachable via the tailnet IP.

The reason I think this belongs in KDE Connect rather than only in the
VPN/tailnet implementation is that KDE Connect already has an explicit peer
concept ("Add devices by IP"). In userspace VPN/tailnet environments, the
transport may erase the original peer address before the Java socket layer can
compare it with the configured address. So the trust decision may need to
preserve the user's explicit opt-in even when the app-visible source is
loopback.

A possible safer design than simply accepting loopback would be:

- keep rejecting loopback by default;
- allow it only if the custom device list is non-empty, or behind a more
explicit setting;
- still require normal KDE Connect identity/TLS handling after the initial
address gate;
- optionally log that the connection was accepted as a userspace VPN/proxy
loopback path.

I have a local fork with this behavior and can turn it into a merge request if
this approach is acceptable.

-- 
You are receiving this mail because:
You are watching all bug changes.

Reply via email to