15:59:52 <karsten> #startmeeting metrics team
15:59:52 <MeetBot> Meeting started Thu Nov 29 15:59:52 2018 UTC.  The chair is karsten. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:59:52 <MeetBot> Useful Commands: #action #agreed #help #info #idea #link #topic.
16:00:17 <karsten> https://storm.torproject.org/shared/5h1Goax5eNusxjXJ_Ty5Wl7hFR1uqCReUiN8xdlBG8T <- agenda pad
16:00:24 <karsten> anything that's missing?
16:01:02 <irl> not from me
16:01:12 <karsten> ok. gaba, are you around now?
16:02:03 <karsten> if not, let's move the first topic to the end.
16:02:25 <karsten> * next retrospective: dec 3rd
16:02:40 <karsten> sounds like the date works for all of us.
16:02:56 <irl> did we decide a time?
16:03:00 <karsten> what time should we pick? 16:00 UTC? 17:00 UTC? 18:00 UTC?
16:03:17 <karsten> I don't think we did.
16:03:46 <irl> 16:00 or 17:00 are ok, 18:00 is probably not
16:04:04 <karsten> okay.
16:04:19 <karsten> maybe gaba joins us later today. if not, email.
16:04:33 <karsten> moving on.
16:04:35 <karsten> * open reviews: #28353, #28137, #28305 (karsten)
16:04:47 <karsten> these are smaller reviews, I hope.
16:04:59 <irl> aah, but they are not all needs_review
16:05:05 <karsten> can you take a look and maybe do these reviews if they're indeed quick?
16:05:10 <karsten> possible, yes.
16:05:25 <karsten> some are needs_comment_by_irl. :)
16:05:37 <irl> i look at my query regularly to spot things that need review
16:05:44 <irl> https://trac.torproject.org/projects/tor/query?status=merge_ready&status=needs_review&reviewer=irl&or&status=merge_ready&status=needs_review&owner=!irl&component=%5EMetrics&reviewer=&col=id&col=summary&col=reviewer&col=component&col=status&col=owner&col=type&col=priority&col=milestone&col=changetime&order=priority
16:05:49 <karsten> yes, I know.
16:06:24 <karsten> I'm asking, because november is almost over, and maybe something can still be resolved before EOM.
16:06:46 <irl> ok, november is over for me in 1 hour because tomorrow is a public holiday
16:06:51 <karsten> oh.
16:07:14 <karsten> let's forget about it then.
16:07:19 <karsten> next week.
16:07:32 <irl> ok, i can try to do them on the 3rd
16:07:40 <karsten> ok.
16:07:58 <karsten> let's move on to the next item then.
16:08:02 <karsten> * blog post (karsten)
16:08:08 <karsten> did you see my graphs?
16:08:18 <irl> yes i did
16:09:14 <irl> the first 3 are quite easy to understand i think
16:09:23 <irl> the one that compares databases is going to take more explaining
16:10:21 <karsten> okay, we can leave that out then.
16:10:55 <irl> the comparison between relays and cw is a very compelling argument for using cw as a better metric than relays
16:11:33 <karsten> would you want to draft some text for the post? you could also author it, if you want.
16:12:00 <karsten> agreed re: relays vs. cw.
16:12:16 <irl> yes, i can look at doing this
16:12:32 <irl> is that huge spike the time that everyone was at ccc and suddenly all the relays appeared on google cloud?
16:13:33 <karsten> heh, I don't know. could be!
16:13:53 <irl> i have memories of this happening but i don't remember which year, i'll see if i can find out.
16:14:02 <karsten> probably worth investigating, because that's probably a question folks would ask.
16:14:06 <irl> it is nice that the CW graph is unaffected by this change
16:14:11 <karsten> yep!
16:14:25 <karsten> so, the post is supposed to be published around dec 7.
16:14:50 <irl> i think i can draft some text on the 3rd, but you might have to finish it because sbs is coming to aberdeen for the active measurement / ooni hackathon
16:15:03 <irl> which is the 5th/6th/7th
16:15:13 <karsten> okay, that sounds good to me.
16:15:17 <irl> cool (:
16:15:29 <karsten> please prioritize this over the other reviews we talked about earlier.
16:15:38 <irl> ok
16:15:44 <irl> that is helpful
16:16:03 <karsten> okay, great!
16:16:11 <karsten> we ran out of topics, I think.
16:16:41 <karsten> unless you have more?
16:17:07 <irl> no, the collector work is keeping me busy and i think you answered all of the latest round of questions in your last email
16:17:21 <karsten> okay, cool!
16:17:23 <irl> oh, except one
16:17:26 <karsten> yes?
16:17:49 <irl> do you still think we should have the feature that does perform the downloads of /*/all given how little extra data we get out of it?
16:17:58 <irl> or should we ignore it until we have a better mechanism?
16:18:37 <karsten> are we going to import from cached descriptors, synced over from dir auths?
16:18:51 <karsten> hmm, maybe that's unrelated, though.
16:18:54 <irl> i'm not familiar with this mechanism
16:18:57 <irl> do we currently do that?
16:19:00 <karsten> yes.
16:19:35 <irl> this seems like a thing we should find a way to not do
16:19:45 <karsten> yes.
16:19:55 <karsten> we import gabelmoo's cached descriptors.
16:20:14 <karsten> gabelmoo puts them on a web server, and we download them from there.
16:20:49 <karsten> but it's probably unrelated, because your analysis shows that downloading /*/all doesn't give us much anyway.
16:21:00 <irl> right
16:21:03 <karsten> we need a different way to download really all descriptors.
16:21:20 <irl> ok, i will put this into the next steps then for dir-spec extension
16:21:26 <karsten> so, I think it's the right way not to download /*/all.
16:21:32 <karsten> sounds good!
16:21:44 <irl> ok then, that's everything from me (:
16:21:50 <karsten> cool!
16:22:12 <karsten> have a good long weekend! bye. :)
16:22:18 <irl> bye!
16:22:22 <karsten> #endmeeting