On Sat, Sep 07, 2013 at 05:14:17PM +0200, Benedikt Gollatz wrote:

On 2013-09-07 16:39, David Carlson wrote:

...

As of the time-stamp on this e-mail it is extremely slow using TBB 2.4.17 rc to access such places as gmail.com. It took me about 10 minutes since I first fired it up to sent this email.

So far, 377 relays out of the 4429 in the consensus (~8.5%) are running 0.2.4.17-rc, which prefers 0.2.4.8-alpha and later over 0.2.3.x in circuit creation. It is quite likely that you are just unlucky with the relays Tor picks for your circuits.

You should count capacity (fraction of consensus weights), not number of relays, in all stats like this.

Otherwise you'll end up falling victim to the same sorts of problems many researchers fall into -- which ends with crazy statements like "20% of the Tor network runs Windows, and everybody in the world can break into Windows computers, therefore I can control 20% of the Tor traffic" (if you'll recall that talk a few years back).

In this case, it looks like these are the summed consensus weights, using the 2013-09-08 07:00:00 consensus:

Tor 0.2.2.x: 431869 (4%) Tor 0.2.3.x: 5020458 (45%) Tor 0.2.4.x: 5363503 (48%) Tor 0.2.5.x: 256831 (2%)

If I'm counting this right, Tor 0.2.4.17 by itself is 3681624 (33%).

--Roger