[tor-bugs] #988 [Tor Relay]: Different TLS certs for incoming vs outgoing

[Tor Bug Tracker & Wiki]

#988: Different TLS certs for incoming vs outgoing
--------------------------+-------------------------------------------------
 Reporter:  arma          |         Type:  enhancement
   Status:  needs_review  |     Priority:  minor      
Milestone:  post 0.2.1.x  |    Component:  Tor Relay  
  Version:  0.2.0.34      |   Resolution:  None       
 Keywords:                |       Parent:             
--------------------------+-------------------------------------------------

Comment(by rransom):

 Replying to [comment:15 Sebastian]:
 > As far as review, I think public_server_mode() is an unfortunate
 function name.

 If anyone can suggest a less bad term for a non-bridge relay than ‘public
 server’, I will happily rename `public_server_mode`. I couldn't come up
 with one.

 > To me it immediately implied that we'd be dealing with the
 !PublishServerDescriptor options, instead of a simple bridge check.

 That test is done in functions whose names contain `publishable_server`.
 ‘Publishable server’ is probably a bad term for that, as the documentation
 states that a Tor controller could publish descriptors for a Tor server
 with !PublishServerDescriptor off, thus making the Tor server publishable.

 > I generally wonder if it is necessary to make this distinction at all,
 because I can't see how it would hurt to maintain separate client and
 server certs for public relays. Maybe I'm missing something here?

 See
 [https://gitweb.torproject.org/tor.git/blob/c18bcc8a55dfaef21637b9f6f38d05610b6ab50c:/doc/spec
 /tor-spec.txt#l555 Tor spec §5.3.1]. If an OR X opens a TLS connection to
 an OR Y, and a client asks Y to extend a circuit to X, that circuit should
 generally go over the existing TLS connection between X and Y, even though
 X was the client (and used its client certificates) for that connection.

 > The functions want better documentation wrt accepted parameters and
 return values, I think.

 Yes. Also, the `tor_tls_context_t` reference manipulation should be
 refactored out into utility functions, because I can no longer tell at a
 glance whether `tor_tls_context_t` references are referenced and freed
 properly.

-- 
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/988#comment:16>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online