[tor-scaling] Summary of 5/31 meeting; next steps

[teor]

Hi Rob,

> On 5 Jun 2019, at 09:12, Rob Jansen <rob.g.jansen at nrl.navy.mil> wrote:
> 
>> On Jun 4, 2019, at 7:38 AM, David Goulet <dgoulet at torproject.org> wrote:
>> 
>> I've attached the latest list of versions we have in the network, you'll see
>> what I meant by quite spread out.
> 
> Thanks for this, David!
> 
> If it is easy enough, it would be helpful to also list some notion of the selection probability associated with each version in addition to the raw number of relays. This could be as simple as the normalized consensus weight for each Tor version. That will give us a better sense of the fraction of Tor traffic running over each version.

I was going to ask for consensus weight as well!

> For example, if 2000 relays are running a super old version but those relays carry a negligible amount of traffic, that will help to set priorities. At the same time, we want to make sure they don't carry a negligible amount of traffic *because* they run an old version :/

Just to be clear: torflow, sbws, and tor do not change the consensus weight based on a relay's tor version. (The fallback selection script does, but that's a trivial amount of traffic.)

But later versions might get more traffic because:
* they have performance improvements, or
* they have extra features that are preferred by clients.

T