23:00:15 #startmeeting weekly network team meeting, 6 Nov 23:00:15 Meeting started Tue Nov 6 23:00:15 2018 UTC. The chair is nickm. Information about MeetBot at http://wiki.debian.org/MeetBot. 23:00:15 Useful Commands: #action #agreed #help #info #idea #link #topic. 23:00:21 Let's spell "Nov" right! :) 23:00:25 evening! 23:00:27 https://pad.riseup.net/p/tor-netteam-2018.1-keep is the meeting pad 23:00:29 hej! 23:00:36 whom do we have tonight? 23:00:37 it's the 7th here :-P 23:00:39 * ahf giggles 23:01:18 :) 23:01:33 do we have teor in the room? :) 23:01:34 I made it! 23:01:38 yeah! 23:01:40 I see dgoulet, ahf, gaba. Do we have mikeperry, teor, catalyst, asn? 23:01:46 hi mikeperry ! 23:01:47 * teor is here 23:01:48 hi 23:01:52 whoa 23:01:56 very nice 23:01:58 I am hoping that asn is asleep :) 23:02:01 nice 23:02:07 ahf: I am hoping that you are okay with being awake :) 23:02:23 i'm perfectly ok with it 23:02:27 i think it's great we are doing this 23:02:35 me too 23:02:40 it's good to have you here, teor! 23:03:11 I am glad to be here. It's been a few years since I've been at a weekly meeting 23:03:19 \o/ 23:04:13 Before we start with the roadmap, we should remember our meta-priorities: 23:04:23 because sponsor 8 ends at end-of-year, we really need to get it all done... 23:04:36 ... and the most "not done" thing there right now is O2.5, improved bootstrap reporting 23:04:50 so if at all possible, those are the top priorities, IIUC. 23:04:52 do anybody doing the sponsor8 tickets need help with anything? 23:04:54 Right, gaba? 23:04:55 i hope to be able to jump in and help with some of that at latest next week 23:05:10 yep 23:05:37 ok 23:05:54 so, looking at the roadmap, are we up to date? Are we all working on stuff that's on there? 23:05:58 please please please reach out if any doubt/comment or help with planning it 23:06:20 ahf: fyi, I'm waiting on you finalizing the mainloop patch before jumping on reporting PT msg to the control port that is #28180 23:06:29 dgoulet: yep! 23:06:31 teor, catalyst: we should divide up the stuff on catalyst's list of subtickets for #28018 ? 23:07:02 teor, catalyst: or at least make sure we have them divided up enough to have a plan 23:07:06 nickm: we've been sending mail about it? 23:07:17 yes, and I've been seeing it -- i don't know if it's finalized though 23:08:07 ok i think #11966 is reviewing the patch from isis and running CI on it. we can clean it up when we have the tracker abstractions built. 23:08:24 #27100 needs new code, but probably not a lot 23:09:07 i can take #27100 23:09:16 it is a bit related to what i have on the plate already 23:09:38 both of these will possibly need spec changes. or maybe not, because we say that controllers shouldn't depend on the details of bootstrap reporting 23:10:09 teor, catalyst: are we on track to get all the stuff divided up and started? I'm happy taking on whatever neither of you wants, and/or backend things. 23:10:44 i reclassified a few tickets as sponsor8-can. these are things that i think aren't essential for the sponsor8 deliverable, but nice to have. 23:10:53 if people disagree with that, let's talk 23:11:30 catalyst: are they related to bootstrap? 23:11:44 yeah, children of #28018 23:12:01 was it OK for me to take #27100? i can just reassign it to me 23:12:11 I did the same with some HSv3 tickets, because that's the only system that will depend on a live consensus after #23605 is done 23:12:19 ahf: feel free to take #27100 23:12:27 oops #24661 23:13:08 next biggest win for UX is probably #24661 and related tickets. those will potentially improve our wrong-time-zone behavior by a lot 23:13:11 I also want to do #28203 in s8, so we can actually see bootstrap in chutney 23:13:24 that would be valuable 23:13:26 I am on #24661 and that rabbithole 23:13:43 Is there anything important that I should bang out here? 23:13:44 is the rabbithole big enough we should split the work further? 23:14:11 So far it is #24661 and #28319 23:14:18 Which is easier to do at the same time 23:14:58 If we decide to fix HSv3 as part of s8, then that might be a job for asn or dgoulet 23:15:34 teor: so HSv3 is still broken with large clock skews if we fix those two? 23:15:48 yeah 23:15:57 I put all the related tickets under #23605 23:16:07 teor: cool, thanks! 23:17:15 I think we might want a chutney mode where we run the clients a few hours skewed using libfaketime. That might be easy, or it could be really hard now that tor uses a lot of different timing sources 23:17:50 teor: that would be cool too, but yeah possibly difficult 23:18:19 After I think I've fixed the bugs, I will try, but set a time limit on it 23:18:21 It sounds like you don't need me here just now? :) 23:19:04 nickm: maybe take a look at the children of #28018 that are sponsor8-can and see if you agree if they're optional for the deliverable? 23:19:35 #action nickm reviews "can" children for #28018 and makes sure they are optional 23:19:38 okay 23:20:04 teor, catalyst: can you expand on the large clock skew "hs v3 broken" here? 23:20:12 (maybe after the meeting) 23:20:27 dgoulet: i think teor knows much more about it than i do 23:20:29 not sure what "fixing hsv3" means also for s8 :S 23:21:33 Is everybody else happy with the roadmap as it stands, and working on stuff fairly high on the list? 23:21:42 * ahf is 23:21:43 mikeperry, gaba : I think we should get the packet whitepaper on the roadmap 23:21:58 we should get it done at least a week before the mozilla allhands if we can 23:22:22 nickm: ok 23:22:43 gaba: is that okay with you? 23:22:45 ok, i will check with mikeperry later on that 23:22:46 yes 23:22:57 nickm: I have been meaning to do it asap but keep getting distracted by getting wtf-pad stuff done timely for review, etc. but it's on my list for sure 23:23:10 dgoulet: #27299 and #23764: make HSv3 clients accept a reasonably live consensus 23:23:29 teor: oh those, ok thanks 23:23:35 gaba, catalyst, teor: here is a query that shows the children of #28018 along with their sponsor fields: 23:23:38 https://trac.torproject.org/projects/tor/query?status=accepted&status=assigned&status=merge_ready&status=needs_information&status=needs_review&status=needs_revision&status=new&status=reopened&parent=~%2328018&col=id&col=summary&col=status&col=type&col=priority&col=milestone&col=component&col=sponsor&order=priority 23:23:49 nickm: I am not doing PrivCount, so I don't know how we feel about that 23:23:51 mikeperry: no worries 23:24:00 teor: you mean "not at all" or "not right now"? 23:24:14 not right now, not sure how long s8 will take 23:24:19 Either way, S8 O2.5 needs to take precedence 23:24:28 so it's the right call 23:24:30 * gaba step out for 5 sec 23:24:34 ok 23:24:52 if you need to take a break for a day or so to clear your mind at some point, then like feel free to poke privcount... but it's less urgent 23:25:33 I think we're ready to move past the roadmap 23:25:36 I will probably do the fallback directory mirrors in a break day 23:25:37 next is reviewer assignments 23:26:14 please take a look at the relevant spreadsheet page and see whether youcan do the stuff that you're listed for there 23:26:31 I see that dgoulet and asn have not assigned my #28330 branch. 23:26:36 IMO it's not _that_ big :) 23:27:07 ain't that big indeed just fun + new :) 23:27:25 so I wanted to leave it to anyone who wants it instead of force assigning 23:27:30 ah 23:27:44 okay, so the question becomes whether anybody likes reviewing that kind of thing :) 23:28:07 * dgoulet throws his name into consideration :P 23:29:16 okay, if nobody else wants it, then dgoulet gets it. 23:29:19 thanks, dgoulet! 23:29:31 o/ 23:29:44 any other issues with the reviews? 23:29:56 wrt leftover reviews from last week, is everything moving according to plan? 23:31:08 I don't know if we should merge #27490 23:31:38 It works as an experimental feature, but it might mess up pathbias 23:31:58 hm. mikeperry, you're the pathbias expert. do you have time to opine on that ticket? 23:32:23 Our options are "merge it as experimental" or "make it not mess up pathbias, then merge it" 23:32:31 * gaba needs to move next door and will be back very shortly. sorry. everything looks good so far 23:32:52 teor: I'm fine with both of those options. How hard is the second one? 23:32:53 hrmm. I don't understand at a glance the path bias conflict 23:32:58 but have not looked at the code 23:33:24 Tor clients on IPv4-only networks that set this option will fail half their connections to IPv6 relays 23:33:46 oh, that might mess up the guard code more than path bias 23:33:49 Would that trigger pathbias? 23:33:58 path bias only counts stuff as failures if your circuit makes it at least 2 hops first 23:34:04 oh right 23:34:33 Then I think we should merge it as experimental, and fix any guard bugs that happen 23:34:43 ok 23:35:06 the guard code has its own failure counters for deciding to switch to your filtered guard list, etc 23:35:14 o/ 23:35:17 separate from path bias 23:35:18 whoa 23:35:27 asn: hi! 23:35:31 hello :) 23:35:35 pretty late night but made it! 23:35:39 sorry for being late 23:35:44 * asn goes through backlog 23:35:45 you know this is optional for europe, right? 23:36:01 its ok np 23:36:08 So if everything is working right, then most clients will fail most guards an average of once 23:36:20 hi asn :-) 23:36:49 Oh, in fact, it's better than that, because we only try to have one or two guards 23:36:50 (fwiw i really like the idea of having a weekly meeting with teor) 23:37:01 next agenda thing is rotations: catalyst on triage and mikeperry on CI. (This is after I re-rand the scripts yesterday. Please check the rotations carefully to make sure I didn't mess anything up) 23:37:51 asn: If you and teor (and maybe ahf?) want to do that, I think it would be awesome, but you should schedule it together :) 23:38:02 no i meant this one 23:38:11 asn: this one is monthly :) 23:38:28 Thursdays at 1100 UTC is our optional time 23:38:31 we have the option to schedule meetings thursday too for europe + australia :-) 23:38:42 (and everyone else who wants to join!) 23:38:43 yep sorry for failing to explain it, but i like the idea of having a monthly network team meeting with teor. 23:38:52 +1 23:39:00 cool 23:39:47 If we're set on rotations, then let's move on to discussion? 23:40:19 ahf: where are we with our intro-to-snowflake plans? 23:40:41 * gaba is back 23:41:11 nickm: i've started looking at it, i haven't written much down yet since the documentation that was around it was pretty decent 23:41:21 good 23:41:28 so building it etc. is pretty easy 23:41:39 ahf: were we going to schedule a "kickoff" / "intro to snowflake" meeting? 23:42:07 we talked about it, but what should we do during it? would it be to start spreading out all the non-assigned s19 tasks? 23:42:40 so, dgoulet and i talked with dcf on irc last week about s8 things, but he also mentioned some of the possibilities for a backup PT design 23:42:44 which is also an s19 task 23:42:50 one of them that was brought up is the HTTPS proxy 23:43:08 https://github.com/sergeyfrolov/httpsproxy 23:43:10 is the code 23:44:22 gaba: any thoughts there? 23:44:28 ahf: interesting 23:44:42 I think it would be mainly "let's all learn and think and talk about snowflake for a bit" 23:45:03 yes, I think it should be a start to look at what needs to be done 23:45:25 and learn and think about snowflake for people to work on it 23:45:26 maybe combine them? or maybe not? that would probably take over an hour 23:45:47 a presentation on snowflake with dicussion would be good 23:45:58 gaba and i talked about trying to encourage people to look at snowflake, so maybe we could try to coordinate *when* we do that and then maybe try to do it at the same time? 23:46:07 i don't know if that would work 23:46:21 i did it friday morning this week, which obviously wont work for everyone in here :-) 23:46:27 err, last week 23:46:52 yes, one morning a week for snowflake for people to get up to speed 23:47:04 (morning in whose timezone?) 23:47:08 hehe 23:47:11 MY MORNING :) 23:47:12 it could be a morning for US people, an afternoon/evening for me? 23:47:14 this time works for me 23:47:19 * catalyst looked at the client side of snowflake maybe a year ago 23:47:20 yes 23:47:40 how about friday at 15 UTC? :-P 23:47:46 * ahf totally dragged that number out of nowhere 23:47:54 the issue about friday is that is teor's saturday 23:48:00 ahhhh, right 23:48:20 We aren't going to be able to find a few hours that work for everyone 23:48:27 nope :-/ 23:48:44 what about people looking at snowflake on their own and then at the end of the month do a specific meeting on snowflake 23:48:58 gaba: i think that would be a good idea. at the end of november? 23:49:01 it could be for the last tuesday 23:49:01 possibly at this time again 23:49:04 yes 23:49:09 works for me 23:49:18 Sounds good 23:49:21 ok 23:49:23 ahf: could you send a reminder to the list? 23:49:23 so *everyone* on the team begins to spend a bit of time each week looking at snowflake? 23:49:30 +1 23:49:31 yes, if possible 23:49:32 What are we looking for? 23:49:33 nickm: yep, i assume the public list here? 23:49:38 yeah 23:49:50 teor: figuring out how it works, building it, getting it to run, see if it has problems, etc. is my understanding 23:50:05 yes 23:50:35 one part of this i'm unsure about with s19: is *eveyrone* on the network team going to be working on it? 23:50:44 Ok, is there a snowflake test framework 23:50:46 i have no idea what our planned allocations is for that sponsor 23:51:03 teor: that is above my current level of knowledge, sorry 23:51:08 teor: but worth looking into! 23:51:19 S19 is our main funding in the first half of next year IIUC 23:51:24 for all of us? 23:51:30 If there isn't, then the best use of my time is probably writing a PT mode in chutney 23:51:31 gaba: is that right? 23:51:39 i just worry a bit if we waste some people time on this too 23:51:51 asking everyone to spend some ... 1/14th of their workweek on this might not be ideal 23:52:07 err, 1/10th 23:52:17 Well, not for me, because I'll be PrivCount? Or is there not much Sponsor V left? 23:52:26 mm, right now we don't have anybody taht can work on it other than ahf and we need people for snowflake work 23:52:54 yes teor 23:53:03 maybe the right question is who's really excited about the upcoming snowflake work and wants to start looking into it? 23:53:11 or maybe want to look into to figure out if they are really excited about it 23:54:17 privcount can be a small part of s19 if we are out of v 23:54:19 IMO 23:54:23 yes 23:54:50 i'm gonna send an email to our internal list and to tor-project that we are going to have this meeting on the last tuesday of this month at this time (23 utc) 23:54:51 we need at least 3 people for snowflake so this will be work for the first couple of months in 2019 23:54:57 yes, thanks 23:54:58 and what we will go over things about snowflake 23:55:06 and people should try to look a bit into it before joining :-) 23:55:07 ideally 23:55:12 yes, i think this is a good start 23:55:15 great 23:56:48 pfew! On to more discussion? 23:56:59 quick one: 0.4.0! 23:57:07 oh, yeah. 23:57:10 no objection so far so I say we should bump it 23:57:23 okay, I'm on it, unless somebody objects now 23:57:33 very nice 23:57:41 looks like teor has questions. 23:57:47 wrt #28096, I'd be happy to help 23:58:05 teor: is "some time this week " soon enough, or does it need to be sooner? 23:58:06 We currently run Appveyor on Windows Server 2012, but Windows Server 2016 also works. Should we add it? We get better coverage, but it doubles the build time. #28318 23:58:24 nickm: any time this week is fine, there's no hurry on that patch 23:58:29 teor: ack, ty 23:59:04 teor: I have no opinion -- appveyor is already slow enough that I don't wait around for it. (Can we throw money at them? Does travis have a timeline for windows support?) 23:59:49 Re: #28096, wur Windows versions will continue to be a bit wrong on clients until we do #28097 23:59:57 didn't travis just announce windows support? 00:00:11 yeah, but not the bits we need. No C compilers yet. 00:00:19 I guess we could install one. 00:00:20 oh, early beta it seems 00:00:32 https://docs.travis-ci.com/user/reference/windows/ 00:02:49 ok, anything else in the meeting today? 00:02:57 wrt #27376, anybody willing to opine about build documents? I'd suggest that we ask our application-team friends , since they're they ones who do cross-compilation 00:03:10 *the ones 00:03:13 oops, not the right ticket 00:03:57 #26376 00:04:17 okay, same question about that one 00:04:36 what are the tradeoffs between them? Should we just concatenate them and let somebody else edit them into shape? :) 00:05:15 i still don't get #26376 much. i think the only one i can think of that uses that environment is hello71 00:05:19 I honestly think ahf's is much better 00:05:28 i think everyone else is using the homebrewed versions of the dodgy makefile i wrote 00:05:49 But I also think we are confused about how we want to support Windows builds in general 00:05:57 yes 00:06:01 Or at least, I am confused 00:06:17 did Hello71 end up deleting the MXE stuff? 00:06:23 Yes 00:08:27 How can we move forward? Decide which Windows build systems we want to support? 00:09:12 honestly my preference, if i had the time to do it, would be to just clean up or rewrite Hello71's doc to be what we want, instead of continuing to go back and forth 00:09:12 In practice, we support whatever we do in CI, and whatever Tor Browser does. Everything else is maintained. 00:09:27 i think that we are less in need of deciding on this one because we are not the ones shipping binaries: the application team does 00:09:43 so i think from my point of view the important angle is to get an environment where you can debug CI issues and normal issues 00:09:54 on whatever we are using for our normal work environment 00:10:10 yeah, i think this doc is primarily useful to us when we need to work on debugging windows build stuff on our local machines 00:10:49 yeah 00:11:14 i wrote the makefile thing once because i was pretty impressed by nickm's fedora setup with building cross compiled binaries and running tests in wine 00:11:51 Ok, so in my case, my travel laptop does Windows, and I need a build that will run on Windows. But for most of us (including me), cross-compiling and Wine is the way to go. 00:12:24 yeah, i have a windows machine now, but i use the makefile too via the windows linux support subsystem 00:12:27 in a debian container 00:12:34 Except for fun tickets like #28096, where we needed to test on a real Windows client and server 00:12:37 (so many levels of weird cross environments when doing that) 00:12:52 ahf: fun race conditions, too 00:13:43 i need to checkout soon, i'm sitting and falling a bit asleep :-S 00:14:11 should we call this meeting and continue the cross-compilation discussion on #tor-dev? 00:14:25 yes :) 00:14:25 Sure 00:15:02 okay. Thanks everybody! I'll see you all online! 00:15:04 #endmeeting