14:59:41 <karsten> #startmeeting metrics team
14:59:41 <MeetBot> Meeting started Thu Mar 16 14:59:41 2017 UTC.  The chair is karsten. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:59:41 <MeetBot> Useful Commands: #action #agreed #help #info #idea #link #topic.
14:59:47 <karsten> https://pad.riseup.net/p/3M7VyrTVgjlF <- agenda pad
15:00:07 <karsten> please add more topics if you want.
15:00:33 <karsten> in the meantime, admire this new page: https://metrics.torproject.org/metrics-lib.html
15:00:41 <iwakeh> yes!
15:00:50 <iwakeh> and do the tutorials !
15:00:56 <karsten> oh, yes!
15:01:03 <iwakeh> rather after the meeting, though ;-)
15:01:43 <karsten> should we start, or wait a bit longer?
15:01:51 <iwakeh> anyone missing?
15:01:58 <karsten> hiro won't be able to join us today.
15:02:07 <karsten> but I have some status to report about onionperf.
15:02:12 <karsten> (later)
15:02:38 <karsten> okay, if anything else comes up during the meeting, please add it to the agenda.
15:02:41 <karsten> let's start.
15:02:45 <karsten> - Tor Metrics pages (karsten)
15:02:52 <karsten> - metrics-lib.html (exists; Git/Ant)
15:02:58 <karsten> that's the link I just pasted.
15:03:10 <karsten> the part that's missing is javadocs.
15:03:26 <karsten> iwakeh: did you have the chance to look into that yet?
15:03:44 <iwakeh> not yet, but the submodule should work
15:04:05 <iwakeh> stating that from the bird's eye perspective.
15:04:20 <karsten> yes, I don't see any major obstacles there,
15:04:31 <iwakeh> Basically, git checkout and some file copying.
15:04:37 <karsten> but there are probably a few details.
15:04:43 <karsten> right.
15:04:49 <karsten> ok.
15:04:50 <iwakeh> (ant docs run in between)
15:05:08 <iwakeh> I'll look at it this week.
15:05:22 <karsten> neat. and I think the other new pages are related.
15:05:27 <karsten> onionoo.html and collector.html.
15:05:28 <iwakeh> One idea
15:05:32 <karsten> yes?
15:05:37 <iwakeh> that just came to mind.
15:05:57 <iwakeh> could the intro pages from onionoo for example
15:06:12 <iwakeh> be used as the javadoc overview page, which
15:06:16 <iwakeh> is still missing.
15:06:29 <karsten> there's indeed some overlap.
15:06:34 <iwakeh> Just until we write a real intro.
15:07:07 <iwakeh> and, it would be nice to have the protocol in the javadoc
15:07:14 <iwakeh> regarding onionoo.
15:07:44 <iwakeh> I just have that in mind, too, when looking at the other topic.
15:07:50 <karsten> I could see how we're using existing content until we have more time to write new content.
15:08:13 <karsten> I'm a bit worried about keeping two pages that are almost the same.
15:08:26 <iwakeh> right, this way there is only one.
15:08:27 <karsten> because we tend to forget one of them and let it rot.
15:08:41 <iwakeh> oh, well, 'age'.
15:08:44 <karsten> heh
15:09:07 <karsten> sounds plausible.
15:09:12 <karsten> having only 1 page.
15:09:35 <karsten> regarding onionoo, I was wondering what to do with the table of contents.
15:09:37 <iwakeh> and, there we don't publish javadoc yet on a web site.
15:09:52 <iwakeh> toc?
15:09:53 <karsten> because none of the pages on metrics.tpo has a table of contents.
15:09:59 <karsten> the box with on-page links.
15:10:03 <iwakeh> ah
15:10:13 <iwakeh> doesn't hurt?
15:10:34 <karsten> maybe. but then we might consider having one on other pages, too.
15:10:42 <karsten> anyway, not the most pressing issue.
15:11:07 <karsten> how do we move forward? do you want to look at git/ant things for all three pages this week?
15:11:09 <iwakeh> true, wait for feedback.
15:11:32 <iwakeh> sure, it's one from the technology perspective.
15:11:44 <karsten> yep.
15:12:13 <karsten> okay, cool! maybe we can move these pages before AMS, because there will be a website discussion, and having metrics.tpo done would be nice.
15:12:32 <iwakeh> doable.
15:13:06 <karsten> RaBe: by the way, I might not be in town on monday. maybe we can do the meeting via irc?
15:13:33 <RaBe> yes, sure!
15:13:36 <karsten> to do any remaining polishing.
15:13:43 <karsten> cool!
15:14:07 <karsten> move on?
15:14:14 <iwakeh> yep.
15:14:14 * karsten just added another topic after this one.
15:14:19 <karsten> - Atlas (RaBe)
15:14:23 <karsten> irl: you around?
15:14:34 <karsten> how's atlas going, 1.5 weeks after the blog post? :)
15:16:11 <karsten> RaBe: ^
15:16:27 <RaBe> well, there are some patches ready for review, and i'll see if I can do some more after they're merged :)
15:16:45 <karsten> sounds good!
15:17:21 <karsten> okay, no pressing questions/things to discuss here, right?
15:17:27 <RaBe> no
15:17:31 <karsten> ok :)
15:17:57 <karsten> okay, moving on:
15:18:01 <karsten> - OnionPerf progress (karsten)
15:18:26 <karsten> hiro set up 3 instances in the cloud plus 1 instance on a Tor VM.
15:18:30 <iwakeh> was there a naming problem?
15:18:39 <karsten> yes, with the Tor VM.
15:19:04 <karsten> we're still looking into that.
15:19:09 <karsten> well, into how to work around it best.
15:19:22 <karsten> there are other minor bugs in onionperf. or in the setup.
15:19:28 <karsten> https://people.torproject.org/~karsten/volatile/onionperf-2017-03-16.pdf
15:19:37 <karsten> ^ that's a quick plot of the three instances.
15:19:48 <karsten> so, they're all producing data now.
15:20:21 <iwakeh> complete/1000 (what unit?)
15:20:44 <karsten> it's hard to say when the setup will be bug-free enough to put data on collector/metrics.
15:21:00 <karsten> complete is milliseconds, so the displayed unit is seconds.
15:21:06 <karsten> log scale.
15:21:10 <iwakeh> are there tickets?
15:21:14 <iwakeh> for the problems.
15:21:19 <karsten> yes, on github.
15:21:57 <karsten> similar to the website, it would be great to have data by AMS.
15:22:04 <karsten> which seems doable.
15:22:12 <iwakeh> Can I help somewhere?
15:22:23 <karsten> yes. the collector patch.
15:22:29 * karsten finds the ticket number..
15:22:54 <iwakeh> #21272
15:22:55 <karsten> #21272
15:22:59 <karsten> so close!
15:23:02 <iwakeh> I won :-)
15:23:12 <karsten> you did.
15:23:28 <karsten> " iwakeh, patch looks good, pushed to ​my task-21272 branch. Thanks! Do you want to change anything else, or can I squash and merge?"
15:24:01 <iwakeh> oh, these should be two tickets.
15:24:10 <iwakeh> I'll reply after the meeting.
15:24:17 <karsten> and once it's merged, we'll need to decide when to deploy.
15:24:22 <karsten> yes, maybe even more than 2.
15:24:41 <iwakeh> true; deploy != release
15:24:43 <iwakeh> ?
15:24:57 <karsten> hmm, fine question.
15:25:07 <karsten> we could release, too.
15:25:09 <karsten> and then deploy.
15:25:20 <iwakeh> before AMS????
15:25:22 <karsten> unless we need to resolve many other tickets for that. ;)
15:25:47 <iwakeh> I think deploy without release is fine, as there are not that many tests...
15:26:12 <karsten> sure, we can do that, too.
15:26:20 <karsten> having the data is what matters most, I think.
15:26:28 <karsten> and putting it on metrics-web.
15:26:34 <iwakeh> Otherwise mirror operators feel forced to udate, too.
15:26:50 <karsten> which is easy and might not require code changes. I didn't look. (<- metrics-web changes)
15:27:02 <karsten> good point re: mirror operators needing to update.
15:27:21 <iwakeh> ok?
15:27:40 <karsten> one question is what we do if we learn that data is invalid, because of an onionperf bug.
15:27:55 <karsten> we could just delete that data.
15:28:03 <iwakeh> we're indending to collect in parallel for a while?
15:28:29 <iwakeh> yes, buggy data should be deleted.
15:28:30 <karsten> yes, but we'll provide both torperf and onionperf data.
15:28:42 <karsten> and clients might not notice the difference.
15:28:44 <iwakeh> Label onionperf as beta?
15:28:58 <karsten> where?
15:29:00 <karsten> :)
15:29:02 <iwakeh> Until the next release?
15:29:08 <karsten> it's the .tpf file format.
15:29:18 <iwakeh> On the first deployment.
15:29:19 <karsten> @type torperf 1.0 or whatever.
15:29:38 <karsten> yes, but where would we label it as beta?
15:30:05 <iwakeh> Would it hurt to change the version?
15:30:16 <iwakeh> or type ?
15:30:38 <karsten> changing the type would break metrics-lib, for example.
15:31:07 <karsten> changing the version might break, too.
15:31:10 <iwakeh> hmm,
15:31:25 <iwakeh> it's different data.
15:31:55 <iwakeh> we didn't consider this in the beginning.
15:32:00 <karsten> yeah..
15:32:05 <iwakeh> but, actually it is different and
15:32:11 <karsten> it's different data using the same data format.
15:32:16 <iwakeh> should be marked as such.
15:32:16 * irl appears slightly late
15:32:27 <karsten> so that existing tools can use it without change.
15:32:32 <karsten> hi irl!
15:32:39 <karsten> okay, needs more thoughts.
15:32:47 <iwakeh> third ticket.
15:32:56 <karsten> we'll continue debugging onionperf while we think about that.
15:33:03 <iwakeh> yep.
15:33:13 <karsten> okay, moving on?
15:33:18 <iwakeh> sure
15:33:23 <karsten> - tech report (iwakeh)
15:33:30 <irl> just before tech reports
15:33:35 <iwakeh> ok
15:33:50 <irl> i'll say that tomorrow i'm writing a paper and for breaks i'll be merging atlas patches
15:33:56 <karsten> hehe
15:34:03 <irl> because writing a paper is not fun but atlas patches can be
15:34:21 <karsten> sounds great! and good luck with the non-fun part of that!
15:34:25 <irl> but nothing is currently a question, it's just merging patches
15:34:41 <irl> sorry i missed the actual atlas slot
15:34:52 <karsten> no worries. thanks for the update. :)
15:34:55 <irl> (:
15:35:11 <irl> iwakeh: over to you
15:35:17 <iwakeh> thanks :-)
15:35:34 <iwakeh> I'll mail some reading material tomorrow and
15:35:56 <iwakeh> just put 'TODO' on the roughest edges.
15:36:04 <karsten> sounds good!
15:36:19 <karsten> what's the plan for mailing the draft to others?
15:36:19 <iwakeh> I'll keep it below 40 p.
15:36:25 <karsten> including AMS attendees?
15:36:50 <iwakeh> Maybe we plan that after you
15:36:59 <iwakeh> read something more?
15:37:17 <iwakeh> Would be a good goal.
15:37:23 <karsten> yep. my hope is that we can maybe send it out on monday.
15:37:32 <karsten> could be an internal list.
15:37:52 <karsten> but it would be good to have it for discussions in AMS.
15:38:02 <iwakeh> definitely
15:38:06 <karsten> "didn't you read the bottom of page 37??? out here!!!"
15:38:18 <karsten> okay, cool!
15:38:31 <iwakeh> well, maybe I add even more for obfuscation.
15:38:42 <iwakeh> ;-)
15:38:44 <karsten> heh
15:38:53 <karsten> okay, moving on?
15:39:00 <iwakeh> fine.
15:39:01 <karsten> (nothing for the (...) part)
15:39:07 <karsten> - AMS preparation (karsten)
15:39:27 <karsten> https://people.torproject.org/~karsten/volatile/open-metrics-tickets-2017-03-13.pdf
15:39:34 <karsten> always good to start with a graph..
15:39:43 <karsten> I looked at open tickets earlier this week.
15:39:59 <karsten> this is one thing I'd like to talk about in AMS:
15:40:03 <iwakeh> oh, well but this is a bit like comparing apples and skyscrapers ...
15:40:18 <karsten> yeah, had to start somewhere.
15:40:24 <iwakeh> true
15:40:45 <karsten> I'd like to talk about the future of some components there.
15:40:55 <karsten> like Analysis and Censorship analysis.
15:41:05 <iwakeh> what's the difference?
15:41:09 <karsten> and also Compass, and whether this is in scope.
15:41:35 <iwakeh> Makes sense
15:41:37 <karsten> I think CA is just a bit more specific than A.
15:41:48 <iwakeh> like CA is not really metrics only,
15:41:56 <iwakeh> it is based on tor metrics.
15:42:12 <karsten> tbh, I haven't looked much at those tickets.
15:42:30 <iwakeh> It's somewhat about possible
15:42:35 <karsten> another goal would be to close more tickets.
15:42:36 <iwakeh> censorship events.
15:42:45 <iwakeh> always fine.
15:42:47 <karsten> yes, it is.
15:42:59 <Samdney> perphaps I can also help here. :)
15:43:09 <karsten> like focusing on defects and assigning enhancements to milestones.
15:43:15 <karsten> please do, Samdney. :)
15:43:42 <karsten> or did you mean help in AMS?
15:43:47 <iwakeh> defects need milestones too
15:43:49 <Samdney> both
15:44:26 <Samdney> we should talk in AMS, where I can help. ;)
15:44:31 <iwakeh> Could we do that in the meeting in AMS.
15:44:33 <karsten> okay!
15:44:44 <iwakeh> I mean the ticketing related things.
15:44:56 <karsten> yes, let's talk more about this in AMS.
15:45:01 <iwakeh> Many tickets are context for each other or need context.
15:45:18 <iwakeh> and the general development idea in the background
15:45:57 <karsten> okay, let's talk more about this in AMS. I don't yet know how much time we'll have for this on the 23rd/24th.
15:46:12 <iwakeh> During dinner?
15:46:13 <karsten> but maybe it's enough to get started and continue the more detailed discussion after AMS.
15:46:16 <karsten> hehe
15:46:27 <Samdney> :)
15:46:41 <iwakeh> right.
15:46:43 <karsten> another topic:
15:46:47 <karsten> operation of metrics services.
15:47:08 <karsten> my hope is that none of them will break while I'm in AMS.
15:47:15 <iwakeh> yes
15:47:23 <karsten> or that our collector sync strategy prevents any major problems.
15:47:33 <karsten> what could possibly go wrong.
15:47:49 <iwakeh> well, if the main collector goes away there two others.
15:47:51 <karsten> and of course I'll deploy the new version just in time before departing.
15:48:08 <karsten> right.
15:48:11 <iwakeh> well, life is a risk.
15:48:37 <karsten> heh, yes. but maybe we can come up with a better plan.
15:48:43 <karsten> for the future.
15:48:59 <karsten> okay, I don't have a nice graph for that one, sorry.
15:49:03 <iwakeh> didn't we plan that at the beginning of the year?
15:49:07 <karsten> what else should we prepare?
15:49:09 <karsten> yep.
15:49:15 <iwakeh> for ams?
15:49:18 <karsten> for our berlin meeting.
15:49:29 <iwakeh> when?
15:49:35 <karsten> the one that didn't happen.
15:49:59 <iwakeh> ah, ok.
15:50:00 <karsten> so, yes, we'll discuss this in AMS.
15:50:18 <karsten> what else?
15:50:50 <iwakeh> the changes coming up
15:51:06 <karsten> which changes?
15:51:09 <iwakeh> from what will be implemented from the tech report.
15:51:14 <karsten> oh, yes.
15:51:52 <karsten> sounds good.
15:52:18 <karsten> - OnionPerf second user model
15:52:25 <karsten> maybe we'll just use the onion service measurements.
15:52:31 <karsten> we should talk to rob about that.
15:52:39 <iwakeh> yes.
15:53:07 <karsten> okay.
15:53:23 <karsten> if something else comes up for AMS, let's use the mailing list.
15:53:41 <iwakeh> fine.
15:53:46 <karsten> next week there won't be an irc meeting.
15:54:14 <karsten> seems like we're done for today. :)
15:54:23 <iwakeh> ok.
15:54:39 <karsten> alright. thanks, everyone! :)
15:54:47 <Samdney> Bye!
15:54:50 <iwakeh> bye, bye!
15:54:53 <karsten> see most of you next week in AMS! bye.
15:54:58 <RaBe> bye!
15:55:04 <karsten> #endmeeting