[tor-talk] TorBirdy seems to connect to the same exit node again and again

[l.m]

Hi teor,

You could run TorBirdy through its own instance of the tor client
software, with a separate socks port.

This  would avoid many of the issues you're trying to work around in
b) and  c), as TorBirdy could happily send NEWNYM to its own client
instance all  it liked. There is a slightly increased network load
involved in  running two instances, and there could be security
implications of  running separate tor clients - but mainly if their
connections are  distinguishable.

teor

Good point. Then again you can do that with any application and tor.
The main motivator is the use case for shared tor process. Tor itself
encourages this use case by supporting multiple socks ports and
isolation flags. Is it reasonable to expect everyone to run multiple
tor processes to isolate the NEWNYM signal? It also raises the
question of *how* they would issue the NEWNYM signal. A patch would
involve adding a simple controller to TorBirdy. In some use cases it
probably isn't even a concern to share NEWNYM. That is sometimes just
a NEWNYM is fine, ignoring the problem of changing exit. So if a patch
were created it should support both use cases: issue a NEWNYM or
emulate it for shared use-cases?

I think it might be too much to ask a tor process to issue NEWNYM to a
specific isolation context. But given the shared-process use case is
encouraged--is this a preferable solution?

--leeroy