17:01:07 <micah> #startmeeting network team weekly meeting - august 08, 2022
17:01:07 <MeetBot> Meeting started Mon Aug  8 17:01:07 2022 UTC.  The chair is micah. Information about MeetBot at http://wiki.debian.org/MeetBot.
17:01:07 <MeetBot> Useful Commands: #action #agreed #help #info #idea #link #topic.
17:01:16 <micah> hello everyone!
17:01:19 <dgoulet> o/
17:01:32 <nickm> hellooooo!
17:01:43 <mikeperry> o/
17:01:47 <eta> o/
17:01:50 <Diziet> Hi
17:02:00 <micah> Welcome welcome to the weekly net team meeting, glad to see everyone so excited for another monday!
17:02:17 <micah> meeting pad: https://pad.riseup.net/p/tor-netteam-2022.1-keep
17:03:04 <micah> lets get things started, if anyone has any announcements to throw out there, please go ahead
17:03:41 <micah> For your information, ahf will be back next week. I'll be gone for the week next week, so we'll do a little tag team here
17:03:55 <GeKo> o/
17:04:36 <micah> dgoulet: do you want to give a release update?
17:05:37 <dgoulet> yes
17:05:59 <dgoulet> essentially, we are still doing some simulations to test couple patches. But we are ready to go once that is done.
17:06:10 <dgoulet> We'll be releasing 045, 046 (the very last one since EOL now) and 047
17:06:39 <dgoulet> many patches to help with the DoS in all possible ways especially in terms of memory pressure and congestion control
17:07:00 <dgoulet> I can't say when exactly but as soon as we can which I hope is this week
17:07:01 <dgoulet> ---
17:07:30 <micah> great, thanks - any help/reviews you need at this point, or are you good there now?
17:08:11 <dgoulet> yes I have 2 pending reviews that need a nickm if possible (https://gitlab.torproject.org/tpo/core/tor/-/merge_requests/610) and another one for arma2 (https://gitlab.torproject.org/tpo/core/tor/-/merge_requests/609)
17:08:29 <dgoulet> the rest is under control
17:08:40 <Diziet> nickm, eta: I'm deep in the channel padding thing; should be done by tomorrow.  That means I'm not reviewing your MRs.  I hope that's OK...
17:08:43 <micah> nickm: note ^
17:09:17 <nickm> micah, dgoulet: Ack. I should have reviews okay today.
17:09:33 <dgoulet> thanks
17:09:38 <nickm> Diziet: I can wait a day or two for reviews, but I'll eventually get kinda clogged up after that
17:09:41 <micah> ok great! thanks dgoulet and nickm! i'll try and draw arma's attention to that issue
17:10:03 <nickm> dgoulet: 610 is the only one for me, yeah?
17:10:13 <dgoulet> nickm: yes, it is assigned to you iirc
17:10:26 <Diziet> nickm: Noted.  I want to get this padding rework done while it's all in my head.
17:10:54 <micah> nice to see people focused on getting things done, excellent
17:11:15 <micah> last week we did an arti 1.0.0 milestone scoping sesion to really narrow things down for the anticipated sept 1st release
17:11:57 <micah> that was really productive, and I hope helps to continue that focus for getting there on time
17:12:15 <micah> there was the one thing on the agenda to talk about regarding the congestion control issue
17:12:16 <Diziet> It was kind of odd in that we seem to be on track to have done all the things that really need to be done.
17:12:28 <Diziet> (the Arti triage I mean)
17:12:52 <micah> its odd to feel good, right? :D
17:12:54 <micah> https://gitlab.torproject.org/tpo/core/arti/-/issues/88
17:13:13 <Diziet> micah: My first real software job had so much deadline crisis going on all the time...
17:13:39 <micah> it was discussed at that meeting that this was set for that milestone, but not 100% of that needed to be done for that milestone, so it seems like a good idea to figure out what specifically in here should be part of that
17:13:40 <nickm> Diziet: I don't want to say that we are never in that place, but we haven't been there in a while
17:13:58 <micah> and what could be put off for later (possibly in discrete, separated issues)
17:14:21 <nickm> mikeperry would have good thoughts on arti#88 and whether there are any parts of it that can exist independently
17:15:25 <micah> because of the DoS, a lot of the congestion control test vectors haven't settled yet, so maybe things that don't require that immediately would be better to prioritize there
17:15:41 <mikeperry> I am working on a congerstion control tweak right now, but flow control could be done independetly of that. it has not changed
17:16:32 <mikeperry> I am almopst done with said congestion control tweak, but I n eed to run more sims.. jnewsome appears out tho, so we do not have extra runners to mnake that go faster :/
17:16:48 <nickm> you mean,  we can leave the current CC algorithms alone for now and apply the tweaks from tor when they are done?
17:17:14 <nickm> Do the new algorithms need any information that they weren't collecting previously?
17:17:16 <mikeperry> indeed you could do that. the tweaks I am doing are most important for exits and onion services
17:18:25 <micah> based on that, would it makes sense to make the flow-control design items in #88 be part of 1.0.0, and the rest could be surgically removed into another ticket that isn't part of that milestone?
17:18:56 <trinity-1686a> (jnewsome fwiw, I do see you. Not sure whether the people on the other side of the bridge do)
17:19:03 <nickm> so, iiuc most of this does not actually makes sense to turn on until it is all implemented
17:19:22 <nickm> IOW, we can't use the flow control part until the cc part is available, and vice versa
17:19:25 <mikeperry> trinity-1686a: aha thanks. I do not see jnewsome here on the irc side :/
17:19:34 <nickm> that said, we can build as much as possible and turn it on when it's all done?
17:19:56 <mikeperry> yeah, that would also work
17:19:57 <nickm> mikeperry: do you see particular items on #88 that can be deferred?
17:20:27 <micah> it seems like quite a bit of work to accomplish, test, and review in 3 weeks
17:20:53 <nickm> yeah
17:20:54 <micah> not to say eta can't do it, probably she can :D
17:21:11 <nickm> eta should also sleep and eat and relax
17:21:13 <mikeperry> nickm: yeah not really. that is already a stripped down set.. all of those things would need to be enabled together; they would need to be disabled if only some of them were present
17:21:27 <jnewsome> can people see me now?
17:21:28 * eta is a big fan of sleeping
17:21:32 <nickm> hi jnewsome
17:21:39 <mikeperry> jnewsome: yes! I see you now
17:21:48 <nickm> eta: "That's where I'm a viking!" ;)
17:21:50 <jnewsome> cool, seems like the bridged rooms may be broken :/
17:22:11 <nickm> eta: are there parts of #88 the you especially want to do or especially do not want to do at this point?
17:22:14 <jnewsome> mikeperry: i set up the ephemeral runners you asked for (2x)
17:22:32 <nickm> I'd be happy to pick up parts that you don't want, and there are some that I would actually enjoy :)
17:22:43 <eta> nickm: not especially; I mean I haven't done anything on flow control yet, so I think it'd make sense to continue with implementing vegas
17:22:58 <eta> and maybe try and get it to actually negotiate with the ntorv3 handshake
17:23:25 <nickm> ok; so if you want I could see how far I get on the flow control parts?
17:23:38 <eta> sure, that'd work
17:23:56 <nickm> and you'd, like, start doing the next parts of the cc work, like the estimators for RTT and bandwith-delay?
17:24:22 <nickm> (i'd suggest delaying "dcode, use, and pass consensus parameters for protocol negotiation" part until we hve diziet's padding thing merged
17:24:34 <nickm> since they may want (or not want) similar design)
17:25:16 <eta> yeah; I mean I think vegas doesn't require much estimation?
17:25:24 <eta> we have the RTT EWMA thing that just needs review
17:25:28 <nickm> ah cool
17:25:47 <Diziet> Is this mostly at the circuit layer?
17:25:59 <nickm> micah, eta: Do you want me to split off the flow-control part of #88 then?
17:26:05 <nickm> (into a new ticket)
17:26:11 <mikeperry> flow control is at the stream layer
17:26:14 <micah> ok, nickm will start hacking on flow-control parts, maybe it makes sense to carve out those from #88 into its own issue, so it can be assigned to nickm
17:26:22 <micah> nickm: +1
17:26:32 <eta> sure
17:26:33 <Diziet> mikeperry: Right, so, well above the channel stuff, so it may not conflict.
17:27:03 <mikeperry> dgoulet is also the best PoC for the flow control stuff, though I can answer some questions too
17:27:51 <micah> ok, great. it might make sense to revisit where we are at next week to evaluate how things are going on this
17:29:12 <micah> remember to make your work visible in those issues so we know that its moving
17:29:24 <eta> ack
17:30:03 * nickm has edited the tickets, please feel free to edit more
17:30:08 <nickm> arti#534 is the new one
17:30:54 <micah> great, thanks for offering to help there nickm! if people have congestion control questions be sure to ask dgoulet or mikeperry
17:31:17 <micah> ok, other discussion items we need to hit here today?
17:31:55 * dgoulet is good
17:32:10 <micah> looking at AC and NH team issues, I dont see anything that they are looking for from the network team right now
17:32:54 <micah> so unless there are other discussion points, we could wrap it up
17:33:05 <Diziet> Nothing more from me
17:33:07 <mikeperry> some minor s61 updates
17:33:22 <micah> mikeperry: go ahead and drop those
17:34:08 <nickm> (do we need to arrange a meeting time for me diziet and mikeperry to thrash out remaining padding spec issues?)
17:34:17 <mikeperry> so I fixed an issue with onion service stat gathering in the sim. that is helping me tune slow start tweaks which are reducing queue usage at guards
17:34:31 <Diziet> nickm: I don't think the spec issues are critical.
17:34:37 <nickm> ok
17:34:51 <Diziet> They're mostly documenting the actual behaviours I think.  If there is anything wrong, then we can change to match, later.
17:35:27 <mikeperry> yeah I would prefer to wrap up these CC changes this week and focus on getting the release out.. already also gonna lose a day+ due to grant meetings
17:35:34 <mikeperry> hopefully next week will be more chill for that
17:35:46 <micah> yeah, ok with mikeperry booked up this week, so if its not critical, then lets revisit a meeting around that next week
17:36:01 <nickm> great
17:36:16 <micah> mikeperry: done re s61?
17:37:04 <mikeperry> I have a handful of more sims to run on this alg; I may ask for another set of 3 runners tomorrow or weds, jnewsome
17:37:17 <mikeperry> thanks for setting up the extra 2
17:37:23 <jnewsome> mikeperry: ok cool
17:37:29 <mikeperry> with another set of 3 we can look at the link-sharing results too
17:37:50 <mikeperry> but I probably won't have cycles for that today or tomorrow, so is fine
17:37:53 <jnewsome> mikeperry: should I go ahead and reserve them for tomorrow or wed now? might not be available if I wait
17:38:19 <mikeperry> oh yeah. sure
17:38:33 <jnewsome> ok, i'll reserve for wed
17:39:46 <mikeperry> juga is out, so I think that's the main thing. at some point dgoulet, geko, and I shoud sync about the dos attacks
17:39:59 <GeKo> sounds good
17:40:03 <dgoulet> indeed ^
17:40:06 <mikeperry> but geko just got back and is immediately being thrown into grantlife meetings, heh
17:40:08 <dgoulet> need to get GeKo up to speed
17:40:30 <mikeperry> so I think we can just call it for today
17:40:35 <GeKo> yeah today/tomorrow is not looking that good for me
17:40:42 <micah> ok - as I mentioned in the beginning, ahf will be back next week and I'll be gone. The overlap between the two of us is very narrow, so perhaps when he comes back, folks will need to get him up to speed on where things are at.
17:40:55 <GeKo> but otherwise just ping me and i can hop onto some bbb or something
17:41:15 <micah> i also just wanted to say a huge thanks to everyone for the hard work, the focus and care you are all taking around these things is impressive! please remember to be constructive in your merge requests reviews, and to keep an open mind when receiving that feedback. Collaboration is key and working together on that is how we'll get there.
17:41:32 <micah> and with that, we'll finish up
17:41:37 <micah> #endmeeting