[tor-bugs] #33029 [Core Tor/Tor]: dir-auth: Dir auths should resume sending 503's but never to relays or other dir auths

[Tor Bug Tracker & Wiki]

#33029: dir-auth: Dir auths should resume sending 503's but never to relays or
other dir auths
-------------------------------------+------------------------------------
 Reporter:  dgoulet                  |          Owner:  dgoulet
     Type:  defect                   |         Status:  needs_revision
 Priority:  Medium                   |      Milestone:  Tor: 0.4.3.x-final
Component:  Core Tor/Tor             |        Version:
 Severity:  Normal                   |     Resolution:
 Keywords:  tor-dirauth 043-should?  |  Actual Points:
Parent ID:  #33018                   |         Points:  0.4
 Reviewer:  nickm, teor, armadev     |        Sponsor:
-------------------------------------+------------------------------------

Comment (by arma):

 Replying to [comment:19 teor]:
 > What about the bridge authority?
 > Does it need to allow bridges to post descriptors

 None of these functions are about how to respond to
 directory_handle_command_post(), so posting descriptors should be fine and
 unaffected, both from bridge relays and from other kinds of relays.

 > and get directory documents?

 I don't think the bridge authority is any different from any other dir
 cache, with respect to publishing the consensus. The fact that it doesn't
 have the Guard flag will mean it is mostly ignored by clients.

 In an alternate universe where we'd finished building
 UpdateBridgesFromAuthority, then we'd want to make sure that it is willing
 to serve bridge descriptors to people who know the right fingerprint for
 them. But we don't live in that universe so I don't think it's an issue.
 And also that design is about responding to clients -- who might be
 routing their requests via Tor, in which case they'll get prioritized
 because the requests come from relays, and if not, they would be random
 looking clients anyway.

 Though! This does make me realize: our changed behavior here applies to
 begindir requests over the ORPort too, right? So we are now willing to
 give 503 responses to clients when we're under load. Maybe that's a
 feature, since "under load" means we'd be answering them poorly anyway. It
 definitely makes me more interested in the "several tiers of responses
 (like 'orport or dirport', 'micro flavored consensus or vanilla flavored
 consensus', 'requests compression or doesn't') where we reserve some of
 our bandwidth and don't spend it on the lowest tier" sort of design. But
 that's something we can save for #33018.

--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/33029#comment:20>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online