[tor-bugs] #13000 [Tor]: Should relays wait with publishing a descriptor until they've run a bw self-test?

[Tor Bug Tracker & Wiki]

#13000: Should relays wait with publishing a descriptor until they've run a bw
self-test?
---------------------------+-----------------
     Reporter:  Sebastian  |      Owner:
         Type:  defect     |     Status:  new
     Priority:  normal     |  Milestone:
    Component:  Tor        |    Version:
   Resolution:             |   Keywords:
Actual Points:             |  Parent ID:
       Points:             |
---------------------------+-----------------

Comment (by arma):

 Replying to [comment:1 arma]:
 > I don't think the current behavior is too bad. Seeing the first
 descriptor starts the directory authorities doing their reachability
 tests. And usually the relay publishes a second descriptor pretty soon
 after.

 This part turns out to have been wrong too:
 {{{
 /** How frequently will we republish our descriptor because of large
 (factor
  * of 2) shifts in estimated bandwidth? */
 #define MAX_BANDWIDTH_CHANGE_FREQ (20*60)
 }}}

 So it'll wait 20 minutes to publish the second descriptor. We might want
 to speed that up.

 One hack would be to special-case "my current descriptor says bw 0 but now
 I'm going to say non-zero" and make an exception to the
 MAX_BANDWIDTH_CHANGE_FREQ check.

 Another fix might be to not actually publish the first descriptor if it's
 going to say 0, and then have some point in the code where we decide the
 bandwidth tests have both sent and received their cells (I think this is
 not easy to learn currently), and publish then.

 Also, since relays write their bandwidth into their state file, and re-use
 it on restart, this case actually only matters for relays that just
 started out.

--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/13000#comment:4>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online