On 16 Dec 2018, at 17:01, starlight.2018q2@binnacle.cx wrote:

The cause is

https://gitweb.torproject.org/tor.git/commit/?id=78e177d622f5f3b24023d04458f...

https://trac.torproject.org/projects/tor/ticket/24803

Would be appreciated if the Tor project published outputs of UpdateFallbackDirs.py job runs used when rebuilding the list. Thus operators who have expended effort to keep their relays eligible will know why when dropped.

We usually attach the logs to the relevant ticket.

This time, I saved the logs, but accidentally overwrote them. And I didn't ask Colin to attach his logs.

We'll try to do better next time: I've added a note on the ticket for 2019.

On 17 Dec 2018, at 10:45, starlight.2018q2@binnacle.cx wrote:

Ran the script: output is attached to this message for anyone interested. Live-network test results will vary by time and by the location of tester. Attached run was made over Tor itself using 'torsocks'.

Thanks!

I was bit by having disabled the unencrypted DIR port for one day recently as an experiment.

We rely on onionoo's last changed field: https://metrics.torproject.org/onionoo.html#details_relay_last_changed_addre...

Changing or removing a published address or port resets the last changed date. Adding an IPv6 address does not reset the last changed date.

I realise that it's disappointing for relay operators to lose a flag. But we're not too worried if a fallback drops out of the list for a release or two: changing the fallback list regularly makes it harder to block. And that's good for users.

T