[tor-bugs] #4339 [Tor Relay]: Turn on the last part of proposal 110

[Tor Bug Tracker & Wiki]

#4339: Turn on the last part of proposal 110
-------------------------+--------------------------------------------------
 Reporter:  nickm        |          Owner:                    
     Type:  enhancement  |         Status:  needs_review      
 Priority:  major        |      Milestone:  Tor: 0.2.3.x-final
Component:  Tor Relay    |        Version:                    
 Keywords:               |         Parent:  #2664             
   Points:               |   Actualpoints:                    
-------------------------+--------------------------------------------------

Comment(by nickm):

 Hrm.  I probably should have had it count the number of non-dropped EXTEND
 cells too so we know what fraction of users would be affected by this.

 Okay, here are the possibilities to consider:

    * These are very old versions of Tor that some users are still running.
 We totally expect 0.2.0.x to break, for instance.
    * These are not-so-old (0.2.1.x and later) versions of Tor, but they're
 buggy here in a way we didn't expect.  We should double-check that 0.2.1.x
 and later do the right thing.
    * This is one of the Tor clients that we didn't make -- jtor or
 onioncoffee or silvertunnel or something.  We

 Time for testing, I guess.  From a quick look at the source, it seems that
 silvertunnel doesn't support RELAY_EARLY.   (At least, Cell.java has other
 cell types, but not RELAY_EARLY.  Not VERSIONS or NETINFO either, afaict.)
 And JTor has RELAY_EARLY in its list of cell types, but afaict it never
 actually uses it (if I'm looking at the right version of the source).

 So this might just be nonconformant clients.

 Oh, are they nonconformant?  To be fair, the spec says only:
 {{{
    When speaking v2 of the link protocol or later, clients MUST only send
    EXTEND cells inside RELAY_EARLY cells.
 }}}
 so that would seem to imply that if you don't upgrade to v2 protocol
 support, you don't have to use RELAY_EARLY cells.  But then the spec turns
 around and says:
 {{{
    [In a future version of Tor, relays will reject any EXTEND cell not
    received in a RELAY_EARLY cell.  See proposal 110.]
 }}}
 so it's not like anybody can say we didn't warn them we'd break
 eventually.

 So I think our next steps here are to contact jtor (hi Bruce!) and
 silvertunnel people (hi silvertunnel people!) and let them know that we're
 thinking of doing this migration on the 0.2.3.x timeframe, so they aren't
 blindsided.  Then we should investigate our own clients and make sure that
 nothing supported will break.  And I should tweak that log message to say
 how many cells _did_ go through.

 Thoughts?

-- 
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/4339#comment:3>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online