17:01:26 #startmeeting network team meeting, 28 Aug 2017 17:01:26 Meeting started Mon Aug 28 17:01:26 2017 UTC. The chair is nickm. Information about MeetBot at http://wiki.debian.org/MeetBot. 17:01:26 Useful Commands: #action #agreed #help #info #idea #link #topic. 17:01:28 hi hi 17:01:33 https://pad.riseup.net/p/hpZoy3YQ92uZ is our pad 17:01:39 hello 17:02:14 hi dgoulet , asn, isabela , isis, catalyst! Also ahf & komlo & teor [absent] 17:02:27 also pastly haxxpop and everything else! 17:02:28 yo 17:02:34 ay 17:02:36 o/ 17:02:43 reading updates 17:02:50 * isabela writing updates :P 17:02:50 getting more tea, then reading more updates 17:03:34 \o/ 17:06:05 rehi 17:06:43 one note from me: I'm going to be mostly out Wednesday through Monday: I turn 40 soon, and a long weekend seems appropriate :) 17:06:52 !!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!!1 17:06:56 :) 17:07:12 * nickm <---- cave-aged, vintage nickm. 17:07:23 catalyst: out of curiosity, why do we not care about dropping 21 bits of range in the mantissa? 17:07:32 nickm: !!!!!!! 17:07:39 nickm: happy birthday!!! 17:07:55 nickm: daaaaamn!!! burn it down!!! 17:07:59 thanks! 17:08:10 asn: I'm more likely to sleep in and go on nature walks tbh 17:08:30 eh. potato potato 17:08:43 isis: i meant dropping precision that we lose by 53 or fewer bits of mantissa; think about what happens if there are leading zero bits in the uint64 17:08:46 nickm: SHEIT! great stuff 17:09:00 nickm: well that's cool too :) :) 17:09:19 awww. thanks, everybody. 17:10:05 wrt montreal arrangements -- the team meeting is on the 11th, right. when do we start? 17:10:16 I can never find this stuff when I'm trying to plan 17:10:29 oh we were going to take that to email but i think none of us actually did? 17:10:31 last agenda I saw was that Rust stuff happens on the 9th and 10th 17:10:31 catalyst: okay… hmm… so essentially we don't care because [0, 2^32-1] is still a big enough range? 17:10:50 oh hm. 17:10:52 yeah i thought we were planning to be there on the 9th and 10th 17:10:55 nickm: there is a spredsheet there 17:10:56 nickm, can I join on that day ? 17:11:00 ok cool 17:11:22 which also brings up whether we should invite the rust people from mozilla for those days? 17:11:34 nickm: i'd rather not be there for more than a week; also if we start on the 9th i'll basically have to arrive on the 8th 17:11:40 haxxpop: let's discuss offline? We usually keep that day to the folks who are doing Tor fulltime, since it's all sponsors stuff. 17:11:53 nickm, alright 17:12:07 the 11th is going to be all sponsors stuff? 17:12:14 catalyst: hope your family is ok 17:12:15 All team meeting stuff IIUC 17:12:36 ok cool 17:12:44 mostly sponsor-organization-planning-etc 17:12:54 isabela: thanks! i hope so too; they say they're on relatively higher ground 17:13:05 :) 17:13:18 catalyst: if you can't make 9th and/or 10th IMO that's ok. I'm going to try to be there but I might have trouble on the 9th 17:13:19 fwiw i am trying to keep mozilla people informed so they know which days are helpful and which days are better to explore montréal or work remotely 17:14:09 catalyst: we rearranged the rust hacking to not have any crucial meetings/updates on days where some people are gone 17:15:02 catalyst: so the 9th and 10th will be toolchain setup (if needed) and i think a rust kata that komlo made 17:15:55 how about for after the main meeting? 17:16:34 the hack days generally have useful stuff and followup meetings from stuff brought up during the programmed meeting. 17:16:45 last i checked, the time after the meeting is much less structured thus far 17:17:13 i think there was a pad with detailed scheduling proposals including the rust stuff but i'm not sure how up to date it is 17:17:27 there was an email somewhere 17:17:37 it should probably be a pad/sheet 17:18:18 i think teor sent mail about it on 7/24 on the rust hackfest thread 17:19:24 we have 1. a spreadsheet for team meeting day on the 11 2. a pad to coordinate the rust hackfest stuff 3. a schedule that Alison is building from ppl's ideas which are for the close meeting days 4. a schedule for the open meetings days or hackdays which are the last ones 17:20:09 where 4. might be something done in a wall at the spot 17:20:13 ok. maybe let's send out the links to those on network-team about least once a week, since obviously we're really good at forgetting things? :/ 17:20:28 ok 17:20:37 thanks! 17:22:36 catalyst: I think for #23061, the right thing to do now is probably to tell the privcount math people (Aaron, Rob) that it's their problem to figure out how much precision we need in order to get a good-enough gaussian sample. 17:22:56 +1 17:23:54 we should make sure that we test that our gaussian/laplacian sample code can't do log(0) or 1/0.0 or anything for any of the possible outputs, but that's about it. 17:23:57 nickm: i think that's probably the right approach but people seem to be conflating 2**-53 sized steps with the holes in representable range between many of those 2**-53 steps and i think both are probably important to consider 17:24:43 ok. I think we should do both, though, in terms of what they mean for the sampled output. 17:25:12 oh, clearly i am out of the loop 17:25:29 probably from being gone for 1.5 weeks amidst others being gone 17:25:40 links to stuff would be great! 17:25:55 #23061 has some links, prop#280 is the application I'm most looking at. 17:26:22 I just announced a meeting on Sep 12 8:00 EDT to try to advance/resolve prop#280 17:26:28 the NSF is emailing my @tpo address saying that i receive funding from them and asking for demographic info? is this a normal thing? 17:26:38 yes. 17:26:46 ok 17:27:23 oh, they are asking for *roger's* demographic info 17:27:32 huh. I put in mine... 17:27:37 maybe ask roger what to do then 17:27:58 or at least i am supposed to log in with roger's last name??? 17:28:01 this is weird 17:28:16 I think this is one of those government programs that tries to track how much bias is going on in NSF grants. 17:28:21 dunno 17:29:04 i tried to log in and it says that someone else already used the temporary pin 17:29:16 ¯\_(ツ)_/¯ 17:29:20 oh well 17:29:33 * Samdney is reading the #23061 ticket ... 17:30:51 haxxpop: is your fuzzing code mergeable? I'd like to get it into master soon 17:31:33 Everybody: there are a few unreviewed tickets left in review-group-22. dgoulet and asn have done a bunch of them. If anybody can get the remainder, that would rock. I think I've already reviewed all the stuff I didn't write 17:32:35 it looks like the two remaining onesare #23149 and #22407 17:32:39 *ones are 17:32:57 also #23061 but I think that's getting pretty heavy attention already 17:36:07 oh we should setup our rotation for Sept. 17:36:10 ok i can take #23149 17:36:19 isis: do you have anybody who's helping review/discuss/whatever on the moat stuff? (And is #22817 the right ticket ID?) 17:36:28 catalyst: thanks! 17:37:06 can anybody take #22407 ? 17:37:15 it's a feature folks have been asking for for simply ages 17:37:32 nickm, yes but there is only one file. in fact, it's from dgoulet :) 17:37:55 ok. Where is it? Can I merge it now? :) 17:38:15 a moment it's on oniongit. I'm getting a link for you now 17:38:45 thanks 17:39:01 one it's merged, oss-fuzz will start poking it 17:39:43 catalyst: wrt the state-machine stuff -- do you think you'll be able to get a rough design sketch some time this week that we can talk about when I'm back at full speed? I'd like to help this advance any way I can 17:39:59 yes plz someone take #22407 so we can move forward with review-group :) (related to msg on top of pad) 17:39:59 nickm, I also generated the corpora. Do you want that as well ? 17:40:05 haxxpop: awesome 17:40:06 sure 17:40:09 i'll take #22407 17:40:16 isis: thanks! 17:40:20 woo! thx! 17:40:27 * nickm sets reviewer fields 17:40:29 np! 17:40:38 * armadev catches up from backlog 17:40:50 isis: you can fill out the nsf questions or ignore them, i think either is totally fine 17:41:08 oh it's #22871 17:41:11 oops 17:41:26 np 17:41:27 and nope, i don't have anyone looking at stuff 17:41:37 i listed a big pile of people as working 0 months on Sponsor3, on the theory that maybe it'll make it less surprising when we list non-0 numbers for some of the people in the future 17:42:09 (it's ambiguous what the report is even asking for. it asks who worked on the project. which can be bigger than the part nsf is funding. it's all kind of vague.) 17:42:11 for #15967 i think mcs et al. are looking at it 17:42:17 isis: ok -- maybe when you're happy with it, throw it on oniongit or something and ask folks for feedback? Even though we're not experts, we should at least try to get our feet wet there. 17:42:30 nickm, https://oniongit.eu/haxxpop/tor/tree/fuzzing-hsv3 17:42:38 haxxpop: ty! 17:42:49 nickm, https://oniongit.eu/haxxpop/fuzzing-corpora 17:43:01 i'll need to do a revision of that, which should be fast because i already wrote down what needs to be done, and then the moat stuff should also be fast since it's just writing a little server 17:43:15 nickm, the second is wrong. This is the correct one https://oniongit.eu/haxxpop/fuzzing-corpora/tree/fuzzing-hsv3 17:43:41 ok 17:43:47 isis: excellent 17:44:37 wrt 031 stable -- there are a bunch of tickets still pending that we should defer or not. I'm particularly concerned about #22752, which seems blocker-ish for windows relays 17:44:55 I'm hoping I get time to fix that in the next 24 hours, and put out a release some time over my break... 17:45:04 but the target looks harder and harder 17:46:51 haxxpop: just merged the fuzzing stuff. let's see what happens. :) 17:47:06 nickm, :D 17:48:54 Should we have some kind of pull request and continuous integration ? I think it will make our life easier 17:49:45 CI is a work-in-progress; maybe you can help hiro and others get some answers on #22891 ? 17:49:56 for pull requests we're mainly using trac still 17:51:20 (any more topics for this meeting? I think it's gone well, and we've got a great week ahead...) 17:51:32 I think trac is just the bug tracker. Can it be used to create a pull request ? 17:51:34 also I have another meeting in 8 min so I could use a quick break.... 17:51:34 nickm: 031! 17:51:43 dgoulet: I was just talking about 031? 17:51:45 Sept 5th is next monday 17:51:52 oh sorry 17:51:52 yup; we won't make sep 5th 17:52:02 ack 17:52:04 I propose _not_ delaying the timeline for 0.3.2 though 17:52:11 perfect 17:52:23 (sorry I missed the last bits above) 17:52:34 also we should look to see what went wrong with stabilizing 0.3.1, and whether we should do anything different for future release series 17:54:41 anything else for today? I'll be distracted for the next 2 hours or so but I hope I can be on IRC lurking 17:54:53 nothing from me <: 17:55:30 * dgoulet is good 17:55:32 ok. thanks, all! Be well! Ping me if you want me for anything. :) 17:55:34 #endmeeting