[tor-dev] Running a Separate Tor Network

[Nick Mathewson]

18:21 < nickm> tjr: are you still around?
18:26 < tjr> nickm: Yup
18:27 < nickm> So, speaking generally as a reaction: Yeah!
Bootstrapping Tor from zero should work better and be easier.  If you
want to push us that way, we'll get there.  If not, there are other
ways you can help us get there.
18:28 < nickm> My first suggestion would be: make a master ticket on
trac, then open a bunch of child tickets.
18:28 < nickm> (Or open a bunch of tickets with the same keyword)
18:28 < nickm> and then let's solve this foolishness and bring about
the new golden age^W^W^Wrestorable tor network
18:29 < nickm> Also: Cool!  Thanks for doing these tests!
18:30 < mrphs> GeKo: yeah but eh, brade's comment :/ I should probbaly
reply there
18:30 < tjr> haha awesome.  I will go the master ticket route, and
attach some easy tickets with suggestions/patches and harder tickets
that may have patches eventually
18:30 < nickm> wrt a dirauth accepted by some but not all dirauths:
This is explicitly not handled by the dirauth design.  But it would be
cool if our response were better in that case.
18:31 < tjr> nickm: Ah okay, that probably explains some stuff
18:32 < tjr> When you add/subtract one, do all the DirAuths have a flag day?
18:32 < nickm> basically yeah.
18:32 < nickm> It would be nice to make that a more tolerant flag day...
18:32 < tjr> The voting interval on Prod is an hour, so if you time it
right, the Running flag issue won't arise, but otherwise it seems
risky
18:32 < nickm> but the problem of how to have everybody who has
participated in a vote agree on the outcome of the vote when they
can't agree on who the voters are...  is not a solved problem today
18:34 < nickm> Joining two tor networks is a cool idea.
18:34 < nickm> I don't think it's supported though
18:34 < nickm> Please though, spam my inbox with a huge pile of trac emails!
18:34 < tjr> RE: agreeing on consensus but not voters. Yea, definetly
I'm pretty nervous about mucking around in that area - I'm going to
have to think about it quite a bit and do a lot of simulations
18:35 < tjr> I'm sure it's not supported, I'm just not entirely sure
how much logic inside an OR would get deathly confused by having to
support it.
18:35 < tjr> Mostly I'm wondering about parameters being on for one
and off for the other
18:35 < nickm> Think also about the basic results in byzantine fault
tolerance.  With >1/3 parties corrupt, no consensus can be reached
with any protocol.
18:36 < nickm> btw, okay with you if I copy-and-paste this
conversation to tor-dev in response to your email? :)
18:36 < tjr> Of course