[tor-bugs] #24640 [Obfuscation/meek]: improve meek behavior when target server is down

[Tor Bug Tracker & Wiki]

#24640: improve meek behavior when target server is down
----------------------------------+----------------------
     Reporter:  brade             |      Owner:  dcf
         Type:  defect            |     Status:  assigned
     Priority:  Medium            |  Milestone:
    Component:  Obfuscation/meek  |    Version:
     Severity:  Normal            |   Keywords:
Actual Points:                    |  Parent ID:
       Points:                    |   Reviewer:
      Sponsor:                    |
----------------------------------+----------------------
 As part of our testing of Tor Launcher / Moat functionality, Mark and I
 ran our own meek-server but intentionally stopped the BridgeDB/moat server
 to which it was supposed to talk.  This caused a 5 minute hang within Tor
 Launcher before an error was generated.

 On the meek-client side, we see a series of messages like these:
  status code was 500, not 200; trying again after 30 seconds (9)

 On the meek-server side, we see these messages:
   dial tcp 192.168.1.xx:6790: getsockopt: connection refused

 Because this is part of the Moat client implementation inside Tor
 Launcher, if BridgeDB is down a real person will be waiting a long time
 without receiving any feedback. It does not look like the retry interval
 or count is configurable within meek-client.

 Do you have any suggestions for minimizing or eliminating the 5 minutes?
 We could implement a different maximum timeout inside Tor Launcher,
 although knowing that the underlying error is "connection refused" vs.
 "the network is just really slow" would make things more robust.

--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/24640>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online