14:29:30 <karsten> #startmeeting metrics team 14:29:30 <MeetBot> Meeting started Mon Jan 8 14:29:30 2018 UTC. The chair is karsten. Information about MeetBot at http://wiki.debian.org/MeetBot. 14:29:30 <MeetBot> Useful Commands: #action #agreed #help #info #idea #link #topic. 14:29:37 <karsten> irl_work: you around, too? 14:29:38 <irl> hi 14:29:46 <iwakeh> hi irl! 14:29:53 <karsten> hi! 14:30:12 <irl> happy new year! i see we're still on the 2017/Q3 pad though 14:30:32 <karsten> fixed. 14:30:33 <iwakeh> true :-) new year new pad! 14:30:42 <karsten> okay. for next meeting! 14:30:49 <irl> heh 14:31:34 <karsten> https://storm.torproject.org/shared/Ou-1QRctynWbF4yedi-MfDsjImFMFSIEP20fbVGCPRa <- link 14:31:43 <iwakeh> still loading slowly the storm pads... 14:31:59 <karsten> hmm, do you think a new one will load faster? 14:32:22 <irl> perhaps, it does a lot of markup after it's loaded like author color highlighting 14:32:25 * karsten always wondered why the pad tells us how many bytes it takes on the server. 14:32:52 <karsten> okay, I'll create a new pad after this meeting. :) 14:33:06 <iwakeh> well, my sample of two shows no difference 14:33:14 <karsten> proven. 14:33:29 <iwakeh> oh, the second link was also for the old pad. 14:33:55 <iwakeh> we could erase the old stuff? 14:34:12 <karsten> I guess the size comes from history, not from currently visible stuff. 14:34:29 <karsten> let's not delete things. let's start a new pad. 14:34:31 <iwakeh> makes sense 14:34:57 <karsten> shall we start? 14:35:04 <irl> ok 14:35:08 <iwakeh> fine 14:35:14 <karsten> - Onionoo bandwidth graphs #24155 (irl) 14:35:25 <karsten> so, you have two ideas there? 14:35:47 <irl> so i guess that the 1 month graph is going to go away from onionoo? i haven't found any specifics about onionoo changes in tickets 14:35:56 <karsten> so, about that, 14:36:09 <karsten> I think (but am not 100% certain) that we never explicitly removed graphs earlier. 14:36:20 <karsten> that is, when tor switched from 15 minutes to 4 hours. 14:36:42 <karsten> we simply did not include those finer-grained graphs anymore, and when all (or most) relays upgraded, it didn't matter anymore. 14:37:09 <karsten> I only started looking before this meeting and would have to find the relevant code, 14:37:24 <karsten> but I think that if there's a really old relay, we'd provide a 1_week bandwidth graph for it. 14:37:47 <iwakeh> a question? 14:37:51 <karsten> assuming that this is correct, we'll not be making changes on onionoo's side. you'll simply drop the graph in relay search. for all relays. 14:38:03 <karsten> what's the question? 14:38:15 <iwakeh> onionoo only provides the data, so the graphs 14:38:19 <iwakeh> are supposed 14:38:23 <iwakeh> to disappear 14:38:30 <iwakeh> from relay search and onionoo 14:38:37 <iwakeh> will stop providing 14:38:45 <iwakeh> more finne grained data, correct? 14:39:05 <karsten> I think that onionoo stopped providing finer-grained data for relays that don't provide it anymore, 14:39:18 <karsten> but it keeps providing that finer-grained data for relays that still provide it. 14:39:28 <iwakeh> so, that is ok? 14:39:30 <karsten> the question is whether onionoo should stop doing the latter. I could imagine not. 14:39:34 <irl> onionoo doesn't provide the data anymore, but it also doesn't provide 3 month graphs for relays that have been online less than 1 month 14:39:51 <karsten> oh! 14:39:57 <irl> which means relay operators waiting months before they see bandwidth graphs 14:39:58 <karsten> that is a bug then. 14:40:04 <iwakeh> new ticket 14:40:15 <karsten> maybe it's already on a ticket and I didn't see it. 14:40:55 <iwakeh> do we need to adapt/check all bandwidth related reporting from onionoo then? 14:41:17 <irl> i can make a new ticket for that, ideally we would have 3 month and up always there, and finer grained ones can optionally not be there 14:41:19 <karsten> just bandwidth docs. 14:41:59 <karsten> yes, and ideally we'd do that without hard-coding the "3 months" part. 14:42:09 <iwakeh> is there the relay search ticket already? 14:42:11 <karsten> okay, I have an idea where in the code to look. 14:42:20 <karsten> good question. 14:42:29 <karsten> (sorry, gotta run for 2 mins..) 14:42:29 <irl> the relay search ticket was armadev complaining in #tor-project yesterday 14:42:50 <iwakeh> maybe link the two tickets? 14:43:23 <irl> arma5 ╡ (why does the '1 year' graph for this relay say 'no data available'? surely there are 3 months of data 14:43:26 <irl> ╡ available.) 14:43:52 <irl> but i wasn't sure exactly what changes we were making so i didn't turn this into a trac ticket yet 14:44:15 <iwakeh> ah, I understand. 14:44:19 <karsten> (back.) 14:45:44 <karsten> so, 14:45:54 <karsten> what about relay search displaying 1 month data at all? 14:46:10 * karsten is confused about tickets 14:46:15 <karsten> is there a ticket for that? 14:46:20 <iwakeh> you mean: skip these graphs? 14:46:49 <irl> now that i've had a good hack on the graphing code, i think we won't be dropping the 1 month graph. if it's not available then i'll hide the tab from view so it's not clickable. 14:46:52 <irl> this would be a new ticket. 14:46:58 <iwakeh> there is the onionoo ticket from above and a to-be-created one for relay search 14:47:36 <karsten> another option would be to display the last month from the 3_months data set. 14:47:51 <irl> can that be done in onionoo? 14:48:04 <karsten> weights data is even available for 1_week. 14:48:15 <karsten> fine question. related to your next one, I think. 14:48:20 <karsten> " - Extrapolating to form the wider scale graphs instead of not giving data" 14:48:27 <iwakeh> should there be a flexible time-frame query option? 14:48:46 <iwakeh> down to the minimum unit of X days? 14:48:55 <karsten> no, we wouldn't be able to prepare that. 14:49:09 <karsten> we're writing these files to disk and only handing them out. 14:49:12 <iwakeh> X=7 is a week. 14:49:31 <karsten> yes, but what we do right now is not flexible. 14:49:38 <karsten> we have a small set of hard coded Xes. 14:49:42 <iwakeh> ok. so receiver wuld need to infer the partial data. 14:49:58 <karsten> that's the question here. onionoo *could* prepare that data set. 14:50:10 <iwakeh> who else would benefit? 14:50:11 <karsten> it's redundant, though, and so far we didn't do it. 14:50:42 <iwakeh> and how much overhead on onionoos side? 14:50:43 <karsten> well, it would be easier to work with onionoo data. 14:50:47 <karsten> some. 14:50:57 <iwakeh> more/bigger docs. 14:51:02 <karsten> bigger docs. 14:51:07 <iwakeh> more calculation. 14:51:30 <karsten> let's create a ticket for that. 14:51:35 <karsten> and discuss more offline. 14:51:40 <irl> ok 14:51:40 <iwakeh> and move the discussion there. 14:51:45 <karsten> that's not clearly a good or bad thing. 14:52:26 <karsten> ok. 14:52:29 <karsten> made a note. 14:52:33 <karsten> moving on? 14:52:40 <iwakeh> fine 14:52:42 <irl> ok 14:52:45 <karsten> - 34c3 relay operators meeting feedback (irl) 14:53:06 <karsten> how was 34c3? :) 14:53:15 <iwakeh> crowded? 14:53:20 <irl> it was good, not too crowded 14:53:40 <irl> the new location worked quite well for it, except during the day when the huge glass hall felt a little like a washing machine trade show 14:53:58 <irl> but once it was dark everything was fine again with lasers and leds 14:54:02 <karsten> heh 14:54:38 <irl> there was a relay operators meetup and i made a note of a few things relevant to metrics 14:54:49 <iwakeh> cool 14:54:56 <karsten> ! 14:55:14 <irl> the first was that relay operators don't know where to run relays, they have the goodbadisps page but it's not obvious where is oversubscribed 14:55:36 <karsten> fine question. 14:55:42 <irl> a mashup of that page with relay counts or consensus weight data would be really helpful to new operators or existing operators growing their family of relays 14:56:11 <iwakeh> true 14:56:39 <irl> the architecture i have for this in my head would involve the goodbadisps page being a lot more structured than it is now though 14:56:51 <irl> for now perhaps this goes in Metrics/Ideas 14:56:59 <karsten> yes! 14:57:08 <irl> i shall create a ticket then 14:57:14 <karsten> please do. :) 14:57:36 <irl> the next two topics are from nicoo 14:57:52 <irl> i'm not sure if you're aware of the infoscreens used at CCC 14:58:00 <iwakeh> not really 14:58:03 <karsten> nope. 14:58:22 <irl> they run software called infobeamer and essentially it goes through different screens showing information such as next talk time or public transport times 14:58:28 <iwakeh> https://info-beamer.com/ 14:58:30 <iwakeh> this? 14:58:33 <irl> it's scriptable in lua 14:58:36 <irl> yes, that's the one 14:59:02 <irl> nicoo wanted to have a screen showing metrics-bot style factoids around the halls, but we didn't have the time to do it 14:59:17 <iwakeh> neat idea! 14:59:26 <irl> i believe he is working on a module that could be reusable though and i will probably take this to EMF camp in August 14:59:43 <karsten> sounds fun! how can we help on the onionoo side? 15:00:02 <karsten> or, metrics side in general? 15:00:16 <irl> i think the onionoo side is all there already because it's just things that metrics-bot already does, just displayed on a big screen 15:00:26 <karsten> okay. 15:00:47 <irl> the final thing is rewriting tordnsel to use onionoo 15:00:59 <irl> which nicoo is working on, i believe in Go 15:01:17 <irl> he's going to speak to us before attempting any deployment but if there are any concerns that are immediate then we should tell him 15:01:31 <karsten> rewrite which part? 15:01:40 <irl> the DNS server 15:02:04 <karsten> ok. 15:02:13 <irl> from my understanding, we have exitmap run and then we import data into onionoo 15:02:25 <irl> the exit addresses can then be used to replace the dns server 15:02:32 <karsten> no, we're not using exitmap for this. 15:02:47 <irl> uh oh. what do we use? 15:03:14 <karsten> I think it's called TorDNSEL. I always found the names confusing. 15:03:23 <irl> oh dear 15:03:23 <karsten> we could ask arlo. 15:03:30 <irl> that's not good 15:03:40 <karsten> yes, it's all very unmaintained in that area. 15:03:46 <irl> so we're using the old tordnsel as a data source for onionoo which is a data source for the new tordnsel 15:04:05 <karsten> sounds like that's the new plan, yes. 15:04:10 <irl> well, at least we caught this early 15:04:28 <karsten> I mean, it could be that we're replacing a different thing that needs to be replaced. 15:04:36 <karsten> let's ask arlo. 15:04:40 <irl> the idea was to kill the old one 15:04:50 <karsten> ah, that won't work then. 15:05:01 <irl> can we get exit addresses from exitmap? 15:05:13 <karsten> I don't know. we should ask phw. 15:05:55 <irl> ok, i will let nicoo know that this plan needs more thinking and we can talk to arlo and phw 15:06:13 <karsten> sounds good! 15:06:49 <karsten> alright, moving on to the next topic? 15:06:52 <irl> yep 15:06:56 <iwakeh> sure 15:06:59 <karsten> - Monthly report (karsten) 15:07:16 <karsten> please take a look at the report items a bit further down on the pad. 15:07:27 <karsten> and please fill in anything that needs to be in the december report. 15:07:44 <karsten> irl: your item in particular needs a bit more work. 15:08:02 <iwakeh> the jetties ran fine so far, we could mention, too? 15:08:14 <irl> maybe sub bullet points? 15:08:15 <karsten> we did. 15:08:30 <karsten> last but one item. 15:08:52 <karsten> so, hmm, this is not the report format anyway. 15:11:35 <karsten> iwakeh: or what did you want to mention in particular? 15:11:38 <iwakeh> Am I still online? 15:11:42 <iwakeh> ah, yes :-) 15:11:46 <karsten> heh 15:11:54 <iwakeh> was so quiet :-) 15:12:01 <karsten> not on the pad. 15:12:55 <iwakeh> added a sentence there 15:14:10 <karsten> okay! 15:14:21 <karsten> if something else comes to mind, please add it after the meeting. 15:14:27 <irl> i feel like i have disproportionally too many links 15:14:29 <iwakeh> the kickoff 15:14:38 <iwakeh> sponsor13 15:14:42 <karsten> ah, 15:14:45 <karsten> I meant the report. 15:14:56 <iwakeh> ok ;-) 15:14:56 <karsten> if something else is missing in the monthly report, please add it later today. 15:15:01 <iwakeh> hehe 15:15:11 <karsten> and then I'll send the report tomorrow morning. if that's enough time. 15:15:30 <karsten> and yes, we can proceed with the next topic. 15:15:42 <iwakeh> oh, I didn't mean to hurry. 15:15:47 <karsten> no, it's fine. 15:15:54 <karsten> irl: moving on? 15:15:59 <irl> ok 15:16:05 <karsten> (and no worries about too many links. ;)) 15:16:07 <karsten> - Kicking off Sponsor 13 (karsten) 15:16:12 <karsten> it's january! 15:16:24 <iwakeh> yep 15:16:27 <irl> https://trac.torproject.org/projects/tor/wiki/org/sponsors/Sponsor13 ? 15:16:29 <karsten> when we got the grant we said we'd complete things in 2017 before starting this work. 15:16:35 <karsten> yes, looks good, irl. 15:17:04 <karsten> so, my plan was to resolve a few defects and make sure that everyone else is not blocking on me, 15:17:08 <karsten> and then dive into 13. 15:17:51 <karsten> does that make sense? 15:18:01 <iwakeh> How to proceed? 15:18:10 <iwakeh> Another meeting for discussing the 15:18:20 <iwakeh> various deliverables and the first steps? 15:19:07 <karsten> what are your plans, iwakeh, for the next weeks? 15:19:14 <karsten> do you have things you'd want to complete first? 15:19:18 <iwakeh> webstats needs to be finished. 15:19:59 <karsten> how about you finish webstats and I come up with a possible plan for 13 until the next team meeting? 15:20:12 <iwakeh> sounds like a plan. 15:20:23 <irl> for 1.1, does presentation include relay search? 15:20:29 <irl> trying to work out how implicated i am 15:20:54 <iwakeh> thisis only relating to the metrics-web 15:20:59 <iwakeh> that existed back then. 15:21:13 <irl> right, ok 15:21:32 <karsten> right. I think it's okay to mention it, but no need to go into details. 15:22:38 <karsten> alright. 15:22:41 <karsten> last topic? 15:22:43 <iwakeh> documenting relay search would be a good topic fr a next sponsor? 15:22:56 <iwakeh> fine to proceed. 15:23:24 <karsten> - Next meeting (karsten) 15:23:31 <karsten> in 1.5 weeks? 15:23:36 <karsten> or is that too long? 15:23:38 <iwakeh> 17th 15:23:49 <irl> the 18th? 15:23:59 <iwakeh> err, yes. 15:23:59 <karsten> yes, why not 18th? 15:24:02 <karsten> ok. :) 15:24:13 <irl> 18th is fine for me 15:24:25 <iwakeh> thursday calender week 3 is fine :-) 15:24:31 <karsten> heh 15:24:45 <karsten> perfect! 15:24:57 <karsten> I think that's all then. 15:25:08 <irl> (: 15:25:22 <karsten> thanks, and bye, bye! :) 15:25:26 <iwakeh> bye, bye :-) 15:25:29 <irl> bye! 15:25:31 <karsten> #endmeeting