[tor-bugs] #3472 [Tor Bridge]: Implementing the pluggable-transport spec (managed proxies)

[Tor Bug Tracker & Wiki]

#3472: Implementing the pluggable-transport spec (managed proxies)
------------------------+---------------------------------------------------
 Reporter:  asn         |          Owner:  asn               
     Type:  task        |         Status:  new               
 Priority:  normal      |      Milestone:  Tor: 0.2.3.x-final
Component:  Tor Bridge  |        Version:                    
 Keywords:              |         Parent:  #3591             
   Points:              |   Actualpoints:                    
------------------------+---------------------------------------------------

Comment(by asn):

 Replying to [comment:4 nickm]:
 > The filenames "pluggable_transports.[ch]" are pretty long.  Can we just
 have that be "transports.[ch]"
 >

 Done. I kept the prefix of public functions to pt_. I'm not sure if this
 is nice.

 > The argument-parsing code in parse_client_transport_line is O(N^2).
 That's ugly.  Also, instead of having it launch the transports, it should
 probably just populate a data structure; there should be a separate
 function to launch the transports.  (Rationale: it's easier to test the
 parsing.)
 >
 > Also, what is the point of "x = tor_strdup(foo); tor_free(foo);" ?  Why
 not just "x = foo;" ?
 >

 Alright, I stopped the smartlist_del_keeporder() business, and the code
 looks nicer than I expected.

 As far as testing is concerned, I didn't do it with a data structure -
 that's how I originally wanted to do it - because I noticed that no other
 config.c parsing function returns a data structure. I also thought that
 `validate_only` turns it into a testable function (even though at the
 moment proxy_argv is prepared only when !validate_only; this is easy to
 change.).

 > In  810a7a5fa09734, the ST_prefix is too overloaded.  Also, can the
 other code that uses tor_spawn_background be made to use your read-a-line
 function?  It's best if we can avoid duplicated code there.

 Done. I tested log_from_pipe() using obfsproxy and it seems to function as
 intended.

 >
 > I'll look more at the actual code later.

-- 
Ticket URL: <https://trac.torproject.org/projects/tor/ticket/3472#comment:5>
Tor Bug Tracker & Wiki <https://trac.torproject.org/>
The Tor Project: anonymity online