Hi,
as Roger mentioned in a side note [1], tor relays running Tor v0.2.2.x have been removed from the consensus (#11149) [2].
Even if relay operators with ContactInfo were contacted directly I think this information would be worthwhile to be mentioned on tpo's twitter feed and blog. This might increase the likelihood that people still running ancient tor versions get to know about it. (even if people running such old tor versions are unlikely subscribers of either news feed ;)
Question arisen from looking at the relays by version graph:
If you look at that graph you see that on 2014-04-08 the number of relays (in the consensus) running 0.2.2.x were about zero, and now (2014-04-21) we are back at about 170 v0.2.2.x relays.
https://metrics.torproject.org/network.html#versions
Why is that? Has the decision postponed/revisited due to the concurrent reduction of relays (heartbleed)?
One (fast) v0.2.2.x relay as an example: https://atlas.torproject.org/#details/F499B62325DA32EEAA6451ABF0764E096C8CC1...
[1] https://lists.torproject.org/pipermail/tor-relays/2014-April/004362.html [2] https://trac.torproject.org/projects/tor/ticket/11149 [3] https://twitter.com/torproject
On Mon, Apr 21, 2014 at 01:36:53PM +0000, Nusenu wrote:
Question arisen from looking at the relays by version graph:
If you look at that graph you see that on 2014-04-08 the number of relays (in the consensus) running 0.2.2.x were about zero, and now (2014-04-21) we are back at about 170 v0.2.2.x relays.
https://metrics.torproject.org/network.html#versions
Why is that? Has the decision postponed/revisited due to the concurrent reduction of relays (heartbleed)?
The changes are in https://trac.torproject.org/projects/tor/ticket/11149 which is merged to git master and will be out in Tor 0.2.5.4-alpha.
So only when a majority of directory authorities are running with that patch will the 0.2.2.x relays be cut out of the network.
But it's actually more subtle than that -- it's more accurately described as "only when 5 or more directory authorities are online and not running the patch will those relays be published".
https://consensus-health.torproject.org/ indicates that three authorities are running the patch, so six aren't. If two of those six went offline, that's enough for the remaining authorities to not put together a majority of Running votes for the 0.2.2.x relays. It happens that two of those six (turtles and dannenberg) *did* go offline while they were waiting to do the openssl bug cleanup.
--Roger
https://consensus-health.torproject.org/ indicates that three authorities are running the patch, so six aren't. If two of those six went offline, that's enough for the remaining authorities to not put together a majority of Running votes for the 0.2.2.x relays. It happens that two of those six (turtles and dannenberg) *did* go offline while they were waiting to do the openssl bug cleanup.
Roger, thank you for your clarifications.
While thinking about that process and the "Recommended Versions" voting I was wondering if it would make more sense to have dirauths vote on the oldest acceptable tor version on the network like they vote already for "Recommended Versions", so removing version x relays would be editing a config file on the dirauths instead of a code change.
These days when trying to reach affected relay operators (whether it is heartbleed or tor v0.2.2.x) I was wondering if you are also using the Tor weather DB to find ways to contact relay operators if they do not have ContactInfo set, but have an weather subscription associated with their relay FP.
(regardless of the fact that a subscriber is not necessarily the owner/operator of the relay)
tor-relays@lists.torproject.org
-
Nusenu -
Roger Dingledine