[tor-bugs] #23101 [Core Tor/Tor]: Predict and build specific HS purpose circuits (rather than GENERAL)

[Tor Bug Tracker & Wiki]

#23101: Predict and build specific HS purpose circuits (rather than GENERAL)
-------------------------------------------------+-------------------------
 Reporter:  mikeperry                            |          Owner:
                                                 |  mikeperry
     Type:  enhancement                          |         Status:
                                                 |  needs_revision
 Priority:  Medium                               |      Milestone:  Tor:
                                                 |  0.3.3.x-final
Component:  Core Tor/Tor                         |        Version:
 Severity:  Normal                               |     Resolution:
 Keywords:  tor-hs, tor-guard, guard-discovery-  |  Actual Points:
  prop247-controller, review-group-27            |
Parent ID:  #13837                               |         Points:
 Reviewer:  asn                                  |        Sponsor:
-------------------------------------------------+-------------------------

Comment (by mikeperry):

 Replying to [comment:14 asn]:
 > Initial review performed here on oniongit. This one is a hard and
 complicated feature to do well and clean. Let's do it!
 >
 > I'm a bit anxious about the new purpose added here, and how well it's
 handled by the whole codebase. I'd really like a way to test this new
 purpose, either unittest or integration test (currently all the
 integration tests fail `make test-network-all` because of #13837 issues).
 After a few #13837/#23101 fixes get in I'll also give the `vanguard`
 project a try.

 Ok, I rebased this on top of the fixups from mikeperry/bug13837, and then
 wrote some more fixups for your code review. They are in
 mikeperry/bug23101-rebased, and each fixup is linked in a comment on your
 oniongit review.

 The main outstanding issue from both bugs is what to do about
 choose_good_middle_server(). Since we don't like various things about how
 it is changed in each bug, we should probably get together on IRC and talk
 about how we want to refactor it to satisfy both sets of concerns,
 otherwise it will be a git conflict nightmare.

 Wrt checks, the vanguard code is basically meant to self-verify all of
 this. It verifies all of the generated paths are as expected using control
 port events. If there is no scaffolding in the current network tests to do
 things like that verify paths are as expected, then I'm leaning towards
 making a separate ticket for that work for all of our circuit building
 stuff, so that it does not block this.

--
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/23101#comment:16>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online