16:59:50 <ahf> #startmeeting Network team meeting, 16th august 2021
16:59:50 <MeetBot> Meeting started Mon Aug 16 16:59:50 2021 UTC.  The chair is ahf. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:59:50 <MeetBot> Useful Commands: #action #agreed #help #info #idea #link #topic.
16:59:54 <ahf> hello hello hello welcome
17:00:08 <nickm> hello!
17:00:18 <dgoulet> hi
17:00:33 <ahf> our pad is at https://pad.riseup.net/p/tor-netteam-2021.1-keep
17:00:44 <ahf> let's give everybody a bit of time to check in
17:00:47 <mikeperry> o/
17:01:44 <juga> o/
17:02:18 <ahf> are folks doing OK on their boards?
17:02:31 <ahf> https://gitlab.torproject.org/groups/tpo/core/-/boards
17:03:10 <dgoulet> yes
17:03:22 <GeKo> hihi
17:03:45 <ahf> i am good too, the interviewing and the release took a bit more time than i had thought last week
17:04:00 <nickm> fine with kanban over here
17:04:50 <ahf> great. i think for the release section the work we are doing on getting 0.3.5, 0.4.5 and 0.4.6 out today have dealt with the most urgent of the tickets in there
17:05:12 <nickm> yup. I have a discussion item to go over  our remaining steps if that's cool?
17:05:45 <dgoulet> yeah same for me I have a question so could be the smae hehe
17:05:54 <ahf> shall we get those done now?
17:05:59 <dgoulet> sure
17:06:42 * jnewsome slips in after missing cal notification
17:06:47 <ahf> 2021-08-16 [nickm] Remaining stuff to do on release and timing it?
17:06:50 <ahf> jnewsome: o/
17:06:51 <dgoulet> quite simple, so how do we proceed with git tag considering I have the latest commits here but usually Nick signs the tag as well ? :S
17:06:59 <nickm> so here is what I'd suggest:
17:07:23 <nickm> when we're ready, you push all the commits.  Then I tag and sign and upload tarballs.  Then you do the rest.
17:07:38 <dgoulet> sounds good to me
17:07:48 <nickm> Let's do that after we're done with interviews today, so we aren't distracted
17:08:14 <ahf> hm, i am gonna be very much afk after the last interview tonight since it's ... 22 at that time i think for me :-/
17:08:19 <ahf> the "rest" is what? announcement?
17:08:30 <hiro> o/
17:08:40 <nickm> update website, tell tor-packagers, wait for website to update, announce on tor-announce, announce on blog
17:09:05 <nickm> then do post-announcement stuff listed in ReleasingTor.md: bumping versions to end with -dev, copying ReleaseNotes and ChangeLog to main, ...
17:09:23 <ahf> i also need to do the CVE update, but i guess that must wait to tomorrow
17:09:25 <nickm> I think that's all
17:10:14 <ahf> dgoulet: if i stick around for drafting announcement after interviews, are you up for publishing it ?
17:10:22 <dgoulet> no choice ;)
17:10:41 <ahf> XD i can stick around longer though, should be possible
17:10:43 <ahf> let's try
17:10:48 <ahf> ok!
17:10:48 <dgoulet> I mean I can't publish it... it requires www power I don't have
17:10:51 <dgoulet> but I can email easy
17:10:55 <ahf> ah
17:11:02 <nickm> no blog account?
17:11:19 <nickm> or can't update website?
17:11:24 <dgoulet> update website
17:11:38 <nickm> ok, that's fast and easy; I can do it if ahf doesn't hve time
17:11:58 <ahf> i don't have access i think. i think i have blog access
17:12:42 <nickm> ok
17:13:14 <ahf> ok, did we resolve both dgoulet's question AND nickm's first item on the pad?
17:13:23 <nickm> i think so!
17:13:28 <ahf> excellent
17:13:56 <ahf> we have no new things from other team's, except a ticket for nethealth that dgoulet filed
17:14:11 <ahf> tor#40448
17:14:13 <dgoulet> yeah but it was more as a reference for future so
17:14:23 <dgoulet> as in not tied to any urgency or milestone
17:14:27 <ahf> ya
17:14:32 <ahf> we have a discussion item more:
17:14:34 <ahf> 2021-08-16 [nickm] Plans for this week's arti meeting?
17:14:55 <nickm> Yes. Basically: I'm hoping we can plan more than last week; any thoughts?
17:15:11 <nickm> I could go over challenges and infelicities in the circuit timeout implementation
17:15:22 <nickm> I could sketch out a guard implementation (which we've gotta build next)
17:15:25 <nickm> I could talk about test coverage
17:15:35 <nickm> we could do something more collaborative
17:15:38 <nickm> we could brainstorm apis
17:15:48 <nickm> any other ideas?  I like the api brainstorming one a lot
17:15:54 <ahf> this is the wednesday meeting?
17:16:14 <nickm> yes
17:17:07 <ahf> circuit timer implementation is the thing you have looked into recently that involved a lot of interesting code that haven't been poked at much in recent times, no?
17:17:18 <ahf> where you had the set selection algorithms discoveries and what not
17:18:09 <dgoulet> API ++ for me
17:18:32 <nickm> yeah, it's the one i was implementing in arti over the past few weeks
17:19:19 <ahf> API is interesting for me too, let's go with that one then
17:19:38 <nickm> ok great!
17:19:57 <nickm> we won't be able to finalize anything but maybe we can brainstorm some user stories or something
17:20:15 <nickm> (this discussion item is solved to my satisfaction :) )
17:20:28 <ahf> yeah
17:20:35 <ahf> ok!
17:20:39 <ahf> i think we are at s61
17:20:59 <mikeperry> ok!
17:21:16 <mikeperry> so I put the updates on the pad
17:21:44 <mikeperry> main things to highlight is that jnewsome did some work to let us run multiple sims at once on the machine
17:22:02 <mikeperry> and hiro did some graphana graphing work of live data
17:22:28 <mikeperry> we should schedule a meeting this week to discuss how to get baselines from the sim and compare to live
17:22:58 <jnewsome> nod
17:23:09 <ahf> \o/
17:23:41 <hiro> yes now what I am missing is seeing if we want to use graphana to display onionperf data or a jupyter notebook so the script can also be used in a CI pipeline and used to compare results
17:23:46 <mikeperry> GeKo: I saw you and arma discussing experiment planning and schedule recording. it looked reasonable; I agree that at a high level, repeated trials is best, and documenting the dates of them
17:24:22 <mikeperry> hiro: did you discuss how to get the tgen models out of onionperf with acute?
17:24:40 <mikeperry> (should we try to get acute to attend our meeting? is she still doing onionperf work?)
17:24:45 <GeKo> great, i meant to point you to https://gitlab.torproject.org/tpo/network-health/team/-/issues/90
17:24:52 <jnewsome> btw here's a successful run from friday with 3 trials. tornettools draws errorbars when there are multiple trials: https://gitlab.torproject.org/jnewsome/sponsor-61-sims/-/jobs/32308/artifacts/file/public/tornet.plot.pages.pdf
17:25:20 <GeKo> mikeperry: i am quite happy with it for a v1 policy and start testing it in the current situation
17:25:26 <jnewsome> they're pretty wide since this was only a 1% sim and only 3 trials. i think we'll want to bump up to at least a 10% network for 'real' results
17:25:44 <GeKo> mikeperry: so, if there is anything you feel you want to add, now-ish would be a good time
17:26:12 <GeKo> otherwise i'll draft something this week and get back to rob and the other folks accordingly
17:26:12 <mikeperry> GeKo: do you have such a draft already? I do not see it on the ticket
17:26:16 <GeKo> no
17:26:27 <GeKo> i wanted to collect arma's and your opinions first
17:26:38 <GeKo> and start a draft with those included
17:26:58 <GeKo> i can do that this week and you can look over it by the end of it
17:27:02 <GeKo> and we take it from there
17:27:02 <mikeperry> GeKo: so I had an old schedule with the idea for a schedule when brainstorming things 1.5 years ago: https://gitlab.torproject.org/tpo/core/team/-/wikis/NetworkTeam/Sponsor61/PerformanceExperiments
17:27:37 <mikeperry> the first few paragraphs echo what arma said, I think. similar idea
17:28:03 <mikeperry> I don't have strong opinions on the actual schedule or coordination mechanism tho
17:28:27 <GeKo> i see, okay
17:29:16 <mikeperry> I wonder if the scripts from https://gitlab.torproject.org/tpo/network-health/team/-/issues/44 will also be useful for shadow investigtion or not
17:30:41 <mikeperry> with respect to work on Prop#324, I got a simple XON/XOFF implementation working for onion services
17:31:22 <mikeperry> dgoulet: I made a ticket with next steps, several of which I could use your help/input on: https://gitlab.torproject.org/tpo/core/tor/-/issues/40450
17:31:52 <dgoulet> mikeperry: yes yes!
17:32:00 <dgoulet> mikeperry: true, I promised you that last week, ok anytime this week!
17:32:24 <mikeperry> nickm,dgoulet: I also made a clearer checklist of what is needed for the full end-to-end protover and negotiation work: https://gitlab.torproject.org/tpo/core/tor/-/issues/40444
17:33:07 <ahf> nice
17:35:13 <mikeperry> it might be tight to make sept 15th, esp since we are all also doing this hiring process, but with help, it still might be possible
17:35:34 <GeKo> mikeperry: i *think* acute has no hours left
17:35:50 <GeKo> but maybe she'd be able to attend a meeting anyway?
17:36:54 <GeKo> https://gitlab.torproject.org/tpo/network-health/metrics/onionperf/-/issues/40024#note_2745275 is the last i am aware of in that regard
17:37:08 <GeKo> juga: are you around by chance?
17:37:14 <juga> GeKo: yes
17:37:23 <GeKo> k, i wondered about https://gitlab.torproject.org/tpo/network-health/sbws/-/issues/40022#note_2747197 :)
17:37:39 <GeKo> that's something for the network-team to look at we need help with
17:38:10 <GeKo> but i am probably missing some details...
17:38:22 <dgoulet> lol the .exit hack of arma2
17:38:32 <dgoulet> I've seen that for "bermuda" :)
17:38:40 <GeKo> right :)
17:40:53 <mikeperry> aha, so such a thing would still have to be an actual relay that we would have to rate limit somehow otherwise, so it doesn't get used as a middle and mess with results
17:42:23 <GeKo> i am not sure why we can't hack that up just for ourselves for the sbws part
17:42:26 <mikeperry> I wonder if the ancient POSTDESCRIPTOR approach could work
17:42:34 <GeKo> and why this needs to be an actually feature in tor
17:42:42 <GeKo> but that could be part of the details i am missing :)
17:42:47 <mikeperry> it might also need UseMicroDescriptors 0, and some other hacks to use the descriptor tho
17:43:47 <mikeperry> GeKo: it certainly would be easier for deployment of sbws if it was something standard in Tor.. though I guess that also depends on how the deployment process works
17:44:43 <ahf> i have to run a bit to prepare for the hiring meeting in 15 min :-/ can people hilight me here when i need to stop the bot?
17:45:00 <dgoulet> sure
17:45:02 <mikeperry> yeah I am pretty much done unless there are more questions
17:45:05 <ahf> oh
17:45:08 <ahf> anything else? :-)
17:45:18 * dgoulet is good
17:45:23 <juga> mikeperry: GeKo: afiu, if we want to exit to localhost, we need arma2 hack, the POSTDESCRIPTOR/private is transversal
17:45:43 * GeKo is fine
17:45:44 <juga> we can continue discussing that somewhere else
17:45:50 * juga is fine
17:45:53 <mikeperry> jnewsome,hiro,acute, maybe Geko: we should try to sync later about shadow vs live comparisons, but today I am literally back-to-back meetings all daty
17:46:06 <GeKo> and i am done for today :)
17:46:08 <jnewsome> mikeperry: ok
17:46:28 <ahf> awesome, see you all around o/
17:46:30 <ahf> #endmeeting