[tor-relays] Exit Node Geographical Location

[Sebastian Hahn]

> On 09 Dec 2016, at 09:34, teor <teor2345 at gmail.com> wrote:
> 
> 
>> On 8 Dec. 2016, at 22:08, Sec INT <sec.int9 at gmail.com> wrote:
>> 
>> US just has alot of people trying to exit there - so its always busy
> 
> Tor clients choose exits at random, based on the ports the exit allows.
> They *do not* try to find an exit close to the site they are going to.
> 
>> - I find Tor follows the money mostly - high concentration in W.Europe and US but drops sharply anywhere else - 
> 
> All the tor bandwidth-measuring authorities are also located in either
> Western Europe or the US. Relays closer to a bandwidth authority
> (lower network latency) are measured faster than those further away.
> 
> This is a side-effect of measuring the delay in transmission inside
> the relay itself.
> 
>> On 9 Dec. 2016, at 06:23, Duncan Guthrie <dguthrie at posteo.net> wrote:
>> 
>> Thus, running relays in Africa and Asia should be a priority right now.
> 
> To make this work well, we would need bandwidth authorities in Africa
> and Asia. Otherwise, those relays won't be used much.
> 
> (We're working on it - I hope!)

Just adding bw auths in Africa won't do too much, because the relevant
factor is who is dominating the median. If we had a majority of bwauths
there, the european/us relays would get measured worse. Also, the more
diversity we have, the worse the latency gets anyway - this is not to
say that we shouldn't add more diversity, but there'll be clearly
noticable issues.

Maybe we could add something to the current system where we try to
estimate how much path length will make a measurement worse by default,
and compensate for that somehow. Otoh, the current state of bwauths is
so sad that I don't know if that'd be even remotely possible. Also such
a system must be resistant to tampering, of course.

Cheers
Sebastian