<html><head><meta http-equiv="content-type" content="text/html; charset=utf-8"></head><body dir="auto"><div><span></span></div><div><div></div><div><br></div><div>On 13 Mar 2018, at 11:41, Karsten Loesing <<a href="mailto:karsten@torproject.org">karsten@torproject.org</a>> wrote:<br><br></div><blockquote type="cite"><div><span>Hi teor,</span><br><span></span><br><blockquote type="cite"><span>On 2018-03-13 09:00, teor wrote:</span><br></blockquote><blockquote type="cite"><blockquote type="cite"><blockquote type="cite"><span>2. What analysis can the metrics team do to help with PrivCount</span><br></blockquote></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><blockquote type="cite"><span>design/development? There's something in the notes about flags changing</span><br></blockquote></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><blockquote type="cite"><span>in 24 hour periods or possible partition of relays. Can you elaborate</span><br></blockquote></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><blockquote type="cite"><span>and make these questions a lot more concrete? Maybe this is something I</span><br></blockquote></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><blockquote type="cite"><span>can do in the next few days, with enough time for you to discuss more</span><br></blockquote></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><blockquote type="cite"><span>with irl while you're in Rome?</span><br></blockquote></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><span></span><br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><span>We want to partition the reporting relays into 3 groups at random.</span><br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><span>(Or maybe some other number: there is a tradeoff between the number of</span><br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><span>groups, which resists manipulation by a single relay, and the quality of the</span><br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><span>resulting statistic.)</span><br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><span></span><br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><span>If we select relays from the consensus at random, do we get a roughly</span><br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><span>even distribution of consensus weight, guard weight, middle weight, and</span><br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><span>exit weight?</span><br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><span></span><br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><span>What if we only have 5% of relays reporting statistics?</span><br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><span>Can we still get roughly even total partition weights at random?</span><br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><span>(Please choose relays on the latest tor versions, because they will be the</span><br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><span>first to deploy PrivCount.)</span><br></blockquote></blockquote><span></span><br><span>Here's a graph (with and without annotations):</span><br><span></span><br><span><a href="https://people.torproject.org/~karsten/volatile/partitions-2018-03-13.pdf">https://people.torproject.org/~karsten/volatile/partitions-2018-03-13.pdf</a></span><br><span></span><br><span><a href="https://people.torproject.org/~karsten/volatile/partitions-2018-03-13-annotated.pdf">https://people.torproject.org/~karsten/volatile/partitions-2018-03-13-annotated.pdf</a></span><br></div></blockquote><div><br></div><div>0.3.2 has the expected consensus weight distribution.</div><div>And it's 2 months since 0.3.2 became stable:</div><div><a href="https://trac.torproject.org/projects/tor/wiki/org/teams/NetworkTeam/CoreTorReleases">https://trac.torproject.org/projects/tor/wiki/org/teams/NetworkTeam/CoreTorReleases</a></div><div><br></div><div>I would be happy to wait 2 months after a stable release for good statistics.</div><br><blockquote type="cite"><div><span>Let me know if this makes sense, or which parameters I should tweak.</span></div></blockquote><div><br></div><div>Can we focus on 0.3.2, and all relays?</div><br><blockquote type="cite"><div><span>For</span><br><span>example:</span><br><span></span><br><span> - Different number of groups (currently 3).</span><br></div></blockquote><div><br></div><div>Can we try 3 and 5?</div><br><blockquote type="cite"><div><span> - Different number of simulations (currently 1000).</span><br></div></blockquote><div><br></div><div>That's fine.</div><br><blockquote type="cite"><div><span> - Different number of consensuses as input (currently 1).</span><br></div></blockquote><div><br></div><div>We'll be collecting over a day, so please use 24 consensuses.</div><br><blockquote type="cite"><div><blockquote type="cite"><blockquote type="cite"><span>If we can't get even partitions by choosing relays at random, we will need</span><br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><span>to choose partitions weighted by consensus weight. Let's decide if we</span><br></blockquote></blockquote><blockquote type="cite"><blockquote type="cite"><span>want to do that analysis after we see the initial results.</span><br></blockquote></blockquote><span></span><br><span>Let me know if you want me to try out a different algorithm. The current</span><br><span>algorithm simply assigns relays to groups at random.</span><br></div></blockquote><br></div><div>That seems to get us what we want, let's keep selecting at random.</div><div><br></div><div>T</div></body></html>