16:01:22 #startmeeting SponsorR 16:01:22 Meeting started Fri Jan 9 16:01:22 2015 UTC. The chair is asn. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:01:22 Useful Commands: #action #agreed #help #info #idea #link #topic. 16:01:27 that might be better. 16:01:39 #agreed 16:01:44 the tbb people know how to give a topic to the meeting 16:01:49 dont know how to use it :( 16:01:50 otherwise you need to dig through many IRC logs to find it 16:01:51 ok may i start with my progress update ? 16:01:58 ohmygodel: yes 16:02:00 syverson: welcome. 16:02:03 syverson: we just started. 16:02:04 hi syverson 16:02:10 syverson: ohmygodel will do his status thing now. 16:02:12 what i did in the last week was write up some statistics aggregation protocols 16:02:17 Great. Hi all. 16:02:24 i sent it to the tor-dev and hidserv-stats mailing lists 16:02:31 end progress report 16:02:36 ok me 16:02:39 start 16:02:42 thanks ohmygodel. i read your first mail. not your second. 16:02:44 robgjansen: go for it! 16:03:15 i had no progress during hte last week. will be working with dgoulet in person next week at the memex meeting to hopefully get shadow+lttng up and running for good 16:03:18 done 16:03:21 yay quick :) 16:03:31 all three of you NRL guys gonna be at the meeting? :) 16:03:37 yes 16:03:40 thanks robgjansen 16:03:41 at least on the first day 16:03:41 * dgoulet next? 16:03:44 dgoulet: go! 16:04:03 I'll be there all week, unless there's a surprise. 16:04:20 grand battle with robgjansen and shadow, running long 6 hours experiement for performance measurement, review tech report 16:04:21 done 16:04:32 thanks dgoulet 16:04:33 let me go next 16:04:39 i was sick for a few days. now i'm better. 16:04:46 since yesterday i've been helping around randomly 16:04:57 I looked a bit at karsten's work with the extrapolation and his analysis 16:05:02 very solid 16:05:13 i took care of the tech report a bit. 16:05:32 i read and replied on ohmygodel's stat aggregation scheme. i need to read his new reply. 16:05:53 i added one or two new tasks to the SponsorR tasklist 16:05:55 and that's that. 16:05:58 for the discussion part 16:06:07 actually we can do this later 16:06:10 karsten: you are next! 16:06:17 ok. pasting: 16:06:18 - Asked operators of fast relays to enable hidserv-stats on their relays. 16:06:18 - Briefly looked into broken Laplace unit test on Raspberry Debian (#14090). 16:06:19 - Continued analysis of reported statistics, made some graphs for next week's presentation. 16:06:24 done. 16:06:38 happy to explain more later. 16:06:47 (or now, of course) 16:07:12 oh #14090. 16:07:14 was not aware. 16:07:15 thx 16:07:19 yeah.. 16:07:26 I'm lost there. 16:07:27 will look. 16:07:30 so anyway 16:07:30 ok 16:07:33 done with the discussion part. 16:07:47 no sorry, let's move to the discussion part. 16:07:54 one of the things that karsten and me want to talk about 16:08:03 ok id like to discuss the meeting next week a bit 16:08:14 that's very good. 16:08:18 unless syverson had something to report? 16:08:26 oh we skipped him? i'm sorry! 16:08:27 syverson: ! 16:08:32 OK 16:09:06 Mostly did some planning for meeting, talked to Rob and Aaron, worked on UK PostNote. 16:09:18 Oh and started on monthly report. Done. 16:09:23 great. 16:09:26 whats uk postnote? 16:09:45 Pariliamentary Office of Science and Technology 16:09:54 explain more please ? 16:09:55 what's going on there? 16:11:06 They do reports on various topics in science. They're working on one on"online anonymity and the darknet" 16:11:33 ah, and you got intereviewed? 16:11:38 I've been clarifying a bunch of stuff, and pointing out how the currently published statistics are very preliminary and misleading. 16:11:49 awesome glad you are doing that 16:12:13 ok that's good. 16:12:14 Yes. I was interviewed. So was andrew I think. Steven also sent them extensive notes, but I think I'm the one spending the most time on it. 16:12:23 strength to you! 16:12:32 so, shall we move on to the next phase? 16:12:37 go 16:12:43 ok. so one topic is the meeting next week. 16:13:04 another topic thati'ev been discussing with karsten is how good our HS traffic figure is. 16:13:27 You mean the extrapolation? 16:13:29 that is, how good accuracy does our 5% consensus weight give us? 16:13:30 yes. 16:13:41 because we only have 5% of the network wrt RP probabiltiies. 16:13:59 and only few of the relays taht report statistics are actaully fast relays that see shittons of RP requests. 16:14:10 and we were wondering if we can get a figure of how accurate it is. 16:14:26 (even if we had to ignore the noise that got added for now.) 16:14:31 so that's another topic. 16:14:32 right. I totally want to learn that. though I think talking about next week is more urgent. 16:14:36 who else wants to talk about things? 16:14:49 i wouldnt mind a quick discussion about the stats protocols i suggested. it might be faster than email to discuss it here and clarify a few things and for me to ask some questions of karsten and asn. 16:14:54 just saying, we can still talk about extrapolation in a week or two. 16:15:05 karsten: ack. 16:15:14 ohmygodel: sounds good. 16:15:25 so that's that? 16:15:33 oh well, if you have more ideas you can say them while talking. 16:15:37 so what should we tackle first? 16:15:50 ohmygodel: I hope it doesn't require having read the thread. because I didn't find time for that yet. really sorry. :( 16:16:01 np it doesnt karsten 16:16:02 wanna start with the stats protocol? the other discussions seem more lengthy. 16:16:17 fine with me 16:16:19 ok 16:16:24 wanna start, ohmygodel , pleasE? 16:16:28 tor people i had a question 16:16:50 that scheme relies on a set of “Statistics Authorities” (StatAuths) to collect anonymous statistics 16:17:18 i want to know the feasibility of getting 1 or 3 of those up and running 16:17:44 hm 16:17:45 2 doesnt make much sense, because the point of having multiple StatAuths would be to prevent a malicious StatAuth from screwing with the stats 16:17:58 (malicious/faulty) 16:18:04 i don't know who would run them. 16:18:12 it also depends on the threat model. 16:18:29 would they be a separate set than dir auths? 16:18:38 they could use the same servers 16:18:39 for example, i would really not want to run something that can learn unique things about the network. 16:18:49 i think it is useful to keep them as a distinct service though 16:18:52 the dirauth operators won't be super excited about this. 16:19:01 no StatAuth can learn anything private 16:19:02 but some of them could theoretically be persuaded. 16:19:09 i.e. all the data they receive can/should be published 16:19:26 all the individual data from each relay? 16:19:27 well presumable some server is currently running the metrics stuff 16:19:34 yes all the individual data from each relay 16:19:45 ok 16:19:46 its anonymized and obfuscated 16:19:52 yes now I remember the scheme :) 16:19:54 the blind one. 16:19:57 ok. 16:20:19 i can see a perfectly fine first step to just run a new service on whatever is running collector.torproject.org 16:20:41 sounds doable. 16:20:50 and worry about distributing the StatAuths later 16:20:55 and then move to 3 for redundancy? 16:21:02 karsten: yes 16:21:12 cool 16:21:15 and asn 16:21:21 yes, I can run the first server. 16:21:30 i suggest you rethink the AnonStats1 protocol 16:21:51 i probably read it wrong. 16:21:54 i think that if we dont worry about individual relays messing up our aggregates 16:22:09 then that protocol is simple and can be made quite private 16:22:19 handling easily the objection you brought up about AnonStats2 16:22:36 ok that is all i have to say about those 16:22:40 thx 16:22:57 ok. 16:23:20 so next topic? 16:23:27 or? 16:23:45 shall we talk about next week? 16:23:50 I guess it doesn't make sense to read AnonStats1 now. so 16:23:53 let's move to next topic. 16:23:54 yes 16:23:55 talk about next week. 16:24:12 what remains to be done? 16:24:17 ok next week 16:24:20 we sent some input for the presentation. 16:24:27 and maybe we'll send more? 16:24:30 karsten those graphs are great 16:24:35 and actually i think are probably enough 16:24:42 we only have 15 minutes to present everything everybody has done 16:24:53 right 16:24:54 all the stats stuff is getting put on one slide 16:25:04 it would be sad to not have something about dgoulet's stuff. 16:25:10 karsten: there is a slide for that 16:25:13 i think dgoulet has a slide 16:25:16 great! 16:25:17 yeah performance stats are a separate slide 16:25:20 he was making the graph for the optimistic data? 16:25:22 or something. 16:25:45 yeah I just finished the 0.2.6 graph showing 100% performance improvement from 0.2.5 16:25:46 for the performance improvement that is. 16:25:54 neat. 16:25:55 very neat. 16:25:57 100%. 16:26:00 100% is an amzing number. 16:26:15 because of what? 16:26:16 optimistic data? 16:26:19 the scheduler? 16:26:22 and probably scheduler 16:26:24 something else? 16:26:27 ok. 16:26:40 good. 16:26:45 so i want to mention 16:26:52 that most of the week consists of 16:27:05 “working groups” and “challenge problem hackathons" 16:27:31 it would be great to come up with ideas for how we might contribute to the latter 16:27:32 btw, would it help to be available via irc/email next week? 16:27:57 as in, being available to spend significant time on something? 16:28:36 "challenge problem hackathons" <-- what's the context of that, problem on anonimity, stats collection, ? 16:28:39 ohmygodel: what's a "challenge problem hackathon" again? 16:28:48 god only knows 16:28:49 oh good, don't have to ask myself.. 16:28:52 perf question: optimistic data vs scheduler: what scheduler do you speak of? 16:28:57 but you should have gotten the agenda 16:29:00 robgjansen: Andrea's work 16:29:11 robgjansen: that has been merged in 0.2.6 16:29:12 sent by gaby araujo on 1/6/15 16:29:36 on the memex list? 16:29:43 can you be more specfic? do you mean the global scheduler? #9262? 16:29:49 looks like I wasn't subscribed at that time. 16:29:55 robgjansen: yup that one 16:29:59 can anybody fwd? 16:30:05 don't have it either i think. 16:30:10 I can, give me a sec 16:30:11 oh ok. 16:30:12 ok, then to my understanding that should not have changed performance 16:30:12 There are three challenge problems: one of them is "Advance Understanding of Darkweb and HS" with a sponsor DoJ CEOS 16:30:33 Child Exploitation and Obscenities Section 16:30:48 robgjansen: maybe not but the performance diff between 0.2.5 and 0.2.6 is quite big, I have yet to pin point what changed that 16:30:53 based on #12889, we set the configs for the scheudler so that it defaulted to the same situation as wihtout the scheduler 16:31:31 of course it is possible that my simulations are not reflecting reality 16:31:53 but our configs were extreme so i would be surprised if that was the case 16:32:02 syverson: and 16:32:52 asn: and what? What are you asking? 16:33:03 syverson: what are the other challenge problems? 16:33:28 i guess no one really cares so n/m 16:33:31 asn: they are 16:33:34 robgjansen: could be something to work on next week also! :) 16:33:34 “Develop reusable tools for the discovery, collection, and extraction of Domain Specific Content” 16:33:35 and 16:33:44 “Advanced Image Processing" 16:33:52 more details in the agenda 16:34:14 dgoulet: i think it might be a good idea to isolate and simulate optimisitc data separate from the rest of 0.2.6 changes 16:34:14 Right, They're not really very related to us at this point. 16:35:40 robgjansen: yeah I tried actually this morning, but chutney wouldn't bootstrap on that commit!... so shadow+lttng, would be awesome for that ;) (we can discuss that after if you like) 16:35:44 so if anybody has ideas about how we might “advance understanding”, specifically related to CP, there will be many interested parties at this meeting 16:36:39 i dont have any suggestions, i just wanted to give you all a heads up about it 16:36:41 ohmygodel_: "define CP" ? 16:36:47 child pornography 16:36:56 ok thx 16:37:03 yes, i don't have good off-hand suggestions myself either . 16:37:11 I'm hoping to engage them on whatever they can tell us about their crawling activities, or things that won't easily show up to crawling detectors, like domain specific search. To help us with our heisenberg problems. 16:37:15 hello 16:37:26 hi 16:37:31 Is the obfsproxy developer around ? 16:37:43 that would be Yawning, I think 16:37:55 < dgoulet> robgjansen: yeah I tried actually this morning, but chutney wouldn't bootstrap on that commit!... 16:37:58 mikeperry: #10395 is now in needs_review :) 16:37:58 :/ 16:37:58 i want to add the ability to send traffic to an UDP port on the server 16:38:02 dgoulet: can you point me at a branch for opt data? i’ll send shadow after it 16:38:05 i wonder if we might ask what challenges theyve found during crawling 16:38:16 asn: wait its you :) 16:38:28 masterkorp: yes, but i'm wearing a totally different hat right now :( 16:38:35 like slowness, failures, low bandwidth, etc. that we would like to improve for them and for all HS users 16:38:38 masterkorp: yawning might indeed be a better person. 16:38:43 masterkorp: at laest for the next 20 minutes. 16:38:50 asn: so, whole is keeping obfsproxy ? 16:39:01 robgjansen: top commit that you want on master is: 09183dc3150000007ccc7b8fecd3b5b762a5b698 16:39:33 this will be an opportunity to talk with law enforcement about how they try and monitor CP on Tor 16:39:45 it could be a useful learning opportunity, at least 16:39:52 ohmygodel: right could be interesting, I had someone contacted me which I think was from that project (crawling) about improving torsocks and tor to support a lot of parallel circuit to tor 16:39:57 Yawning: hello 16:39:58 so clearly there are interests I guess 16:40:11 dgoulet: perfect 16:40:17 dgoulet: so opt data and scheduler changes are not part of memex, right? why are we planning on showing this 100% improvement at the meeting? 16:40:27 they are 16:40:28 robgjansen: optimistic yes 16:40:31 dgoulet: i mean what story are we going to tell that sounds good to  chris’s ears? 16:40:40 client-side performance improvements are part of memex, as we understand it. 16:40:41 not scheduler though, that was something done way back by athena but merged recently 16:40:49 robgjansen: also what asn says :) 16:40:55 oh! opt data is memex! ok 16:41:16 then yeah it would be nice to be able to show that opt data specifically results in improvements 16:41:21 so i think shadow can help there 16:41:26 robgjansen: both improvement are part of Memex the way we understand it in the form of "HS reachability, scalability and performance improvements" 16:42:10 like the parallel intro point being flagged unusable fix we did, we consider that part of memex 16:42:49 well i guess in any case we can claim the improvements are from memex improvements 16:42:54 for which the latter was a gift from the sky for the crawling team ;) 16:42:57 dgoulet: is that kind of stuff going into a slide or is it too detailed? 16:43:03 with a broad enough interpretation ;) 16:43:04 syverson: slide 9 16:43:13 syverson: I sent more info on all that yesterday to phil 16:43:23 and I think we are suppose to discuss it today by phone 16:43:29 i would still personally like to know where the 100% improvement came from though 16:43:36 robgjansen: so do I! 16:43:39 Yes. 2PM eastern. 16:43:51 b/c that is very useful information 16:43:54 robgjansen: could be a good goal for next week hackfest to figure out thus improve our measurement framework :) 16:44:17 also! ok good 16:44:38 ok next topic ? 16:44:51 hmm, wait, also for next week, 16:45:04 should we present a plan what we're going to do in the next three months? 16:45:23 i personally sent an email to Roger/David telling them what I would be interested in doing. 16:45:26 For the presentation Monday? 16:45:27 ok karsten good point 16:45:34 syverson: possibly? 16:45:45 we currently have one slide for future work 16:45:55 tl;dr I'm interested in doing performance/security stuff over next months. 16:46:11 Phil is listing open questions in a single slide for Darkcrawler, stats, everything. 16:46:45 it's unclear how future work is decided on this sponsor 16:46:48 It's all there's time for. 16:46:51 i don't really understand it. 16:47:21 but i'm hoping that this real life meeting will help us decide what we should be doing over the next 3 months. 16:47:33 We will look very closely at Chris when he speaks next week ;) 16:47:37 because roger/david will be able to gauge what interests the sponsor. 16:47:52 ok. so no need to decide on something beforehand? 16:47:53 we are planning to pin chris down and have a fuller future work discussion with him 16:48:04 I think that's the goal is to figure out the future at the meeting 16:48:07 so if you have something in mind 16:48:16 make sure somebody attending at the meeting knows about it 16:48:19 Yes, questions such as how much development time he wants us to put into peerflow. 16:48:20 ok. 16:48:27 masterkorp: ? 16:48:50 guess I'll send such a mail that asn sent to everyone involved. 16:49:05 i wouldnt mind having a bit of discussion about it here 16:49:15 because it doesnt potentially involve coordination among us 16:49:17 what did I do this time? 16:49:29 i realize we are running low on time at this point 16:49:56 ohmygodel: I don't have a good plan yet. your stats thing sounds interesting. more work on the tech report and turning that into a proposal and code sounds good. 16:50:53 things i want to potentially work on: 1. implementation of PeerFlow, 2. exploring the “single researcher” statistics collection model, 3. developing protocols for private statistics aggregation 16:51:02 i guess roger will have a good handle about what to do to keep chris happy by the end of the meeting 16:51:07 for tor anyway 16:51:08 ok karsten, sounds cool 16:51:10 yes. 16:51:17 i'm really hoping that we can find non-statistics stuff to do. 16:51:28 asn: what do you have in mind 16:51:31 asn: sure. performance stuff sounds interesting, too. 16:51:35 performance improvements. 16:51:38 encrypteds ervices. 16:51:47 social hidden services. petnaem systems. blah blah 16:51:53 scaling hidden services. 16:52:13 All cool stuff. Have to make sure we can show how it's memex related. 16:52:27 yeah memex is not interested in improving HS security 16:52:39 I'm really hoping that memex is not just 4 years of "More HS statistics". 16:52:42 so I need to be a subscriber to post to post to tor-dev ? 16:52:46 asn: i have an idea for you 16:52:57 i think we should explore 4-hop hidden services 16:52:58 well, its 2-3 years 16:53:01 Unless we can spin it as making stats and crawling more accurate and robust or some such. 16:53:06 that would be a big perf win 16:53:17 and fits under memex for more scalable crawling 16:53:18 robgjansen: 4-hop hidden service? you mean 5-hop as the proposal went? 16:53:19 and is not stats 16:53:27 masterkorp: yes 16:53:29 robgjansen: the one where the HS and client collectively pick the RP? 16:53:36 cool thanks 16:53:37 also not seeing the use case for udp 16:53:45 You mean like the HS stuff I did with Lasse with ephemeral contact points? 16:53:47 sorry i never read that proposal 16:53:56 guess i should start there 16:54:05 Yawning: that was for me 16:54:27 as I understand it the sponsor was also interested in "More social uses for hidden services. How to make hidden services more Lawful Good." 16:54:35 Yawning | also not seeing the use case for udp 16:54:39 well you dragged me out of my fortress of solutide 16:54:40 was this for me ? 16:54:41 yes 16:54:52 Yawning: http://sourceforge.net/p/openvpn/mailman/message/33216641/ 16:54:55 the use case :) 16:55:10 asn: that doesn't sound right to me, can you say more? 16:55:30 yeah doesnt sound right to me either, asn 16:55:35 i am not talking about the transport layer between the obfsproxy 16:55:42 syverson: wait 16:55:43 openvpn supports tcp 16:55:52 https://lists.torproject.org/pipermail/tor-dev/2014-October/007652.html 16:56:02 tunneling tcp over udp over tcp will have absolutely terrible performance 16:56:05 see (6) 16:56:14 Yawning, masterkorp: can you move to #tor-project maybe just until we finish the meeting? 16:56:28 oh sorry 16:56:33 sure my bad 16:57:04 asn: can you send me link to the 5 hop hs proposal 16:57:18 robgjansen: yes 16:57:30 robgjansen: it's not official tpo proposal 16:57:35 hi 16:57:41 robgjansen: https://lists.torproject.org/pipermail/tor-dev/2014-February/006198.html 16:57:42 i cannot convince myself why 5 hops 16:57:43 thanks 16:57:58 robgjansen: i don't know if you were referring to that crazy idea though!! 16:58:04 maybe you were talking about something else! 16:58:16 yeah something else 16:58:27 so what is 4-hop hidden services? 16:58:30 i will think about this 16:58:48 and try to write something up 16:58:52 any channels/chat active about france's facts? ty 16:58:56 asn: i think that was arma's wishful thinking, but we can figure out how to show it's related maybe. 16:58:58 have more to say later 16:59:09 sound ok? 16:59:16 its 1 hour now 16:59:17 robgjansen: yes sure! 16:59:28 Gotta run to drive to NRL for more meetings. Bye. 16:59:31 so, any urgent deadlines? 16:59:36 bye syverson 16:59:38 syverson: bye! 16:59:41 anything to finish before the meeting? 16:59:47 or are you all set? 17:01:02 I think we covered evertything for next week (being two slides :) 17:01:17 :) 17:01:26 ok. 17:01:44 ok id like to wrap up soon if possible 17:01:53 ohmygodel: yes you are the only NRL remaining! 17:01:53 when's the next meeting? 17:01:59 ohmygodel: tell us! 17:02:02 ahh! 17:02:23 fridays are nice 17:02:31 lets stick with it 17:02:32 ohmygodel: sorry for that! 17:02:33 i like friday 17:02:34 oh 17:02:37 you actually like it? 17:02:40 so 11EST friday after next ? 17:02:42 i don't think it works very well for me. 17:02:53 yeah, fridays are not really great. 17:02:54 ok what would you prefer 17:02:55 we can try to do it, but I bet it will fail in a week or two. 17:03:04 i would suggest we roll back to good ol' tuesdays 17:03:06 so many things you want to do, and then it's weekend... 17:03:07 tue 16h00 UTC are nice 17:03:10 hi everyone 17:03:23 wed or thu would also work. 17:03:33 also guys, I dstill don't have my uni schedule for next smester. 17:03:34 karsten, asn: slides for next week just popped on the memex list 17:03:36 so i'm a bit unstable. 17:03:50 ok then how about next mtg 1600 UTC on 1/20/15 17:03:51 i think 16:00 UTC should be fine whatever the weekday, but not 100% sure. 17:03:57 asn: bah we'll reschedule if you have issue, not too much of a problem 17:04:02 ohmygodel: yes. 17:04:05 i will send the mail! 17:04:10 great 17:04:12 sounds good! 17:04:15 dgoulet see you sunday 17:04:17 ok 17:04:19 ok thanks all! dgoulet: see you in a couple of days 17:04:20 ohmygodel: yess! cheers 17:04:21 good luck next week :) 17:04:24 ohmygodel: and feel free to send mail 17:04:24 robgjansen: also! 17:04:25 if you want anything 17:04:26 adios 17:04:33 ok fair enough 17:05:06 karsten: wrt the extrapolation thing, we should still write that blog post, so we should truy to answer the question in the future :) 17:05:11 ;) 17:05:16 #endmeeting for now