15:00:08 #startmeeting metrics team 15:00:08 Meeting started Thu Mar 9 15:00:08 2017 UTC. The chair is karsten. Information about MeetBot at http://wiki.debian.org/MeetBot. 15:00:08 Useful Commands: #action #agreed #help #info #idea #link #topic. 15:00:14 https://pad.riseup.net/p/3M7VyrTVgjlF <- agenda pad 15:00:28 stargazer: if you want to discuss something metrics-related, feel free to add to the agenda. 15:01:35 what's missing, what's missing.... 15:03:36 should we start and keep appending topics while talking? 15:03:48 sure. that's fine. 15:04:02 i would like to know if metrics projects are gonna be considered for GSoC? 15:04:13 unfortunately, 15:04:38 not as we are quite busy currently and 15:04:49 during the time for GSOC. 15:05:01 yeah, we hope to have more time for mentoring next year. 15:05:12 oh 15:05:12 sorry! 15:05:36 alright, let's start with the first topic. 15:05:40 - Atlas blog post and comments (irl) 15:05:43 irl around? 15:05:44 well, that's alright.. Tor is a brilliant project and i'm sure i'll be here again next year 15:05:45 hello 15:05:49 * irl seems just in time 15:05:52 stargazer: sounds great! 15:05:58 stargazer: fine! 15:06:12 i'll log off then.. have a nice meeting... :) 15:06:14 irl: thanks for fighting the spam comments and for responding to the ham! 15:06:19 thanks, stargazer :) 15:06:22 the ham? 15:06:29 stargazer: i know debian does a lot of GSoC projects https://wiki.debian.org/SummerOfCode2017/Projects 15:06:31 the !spam 15:06:31 too late 15:06:50 ok, yes, we did get a lot of spam 15:06:52 well, bad ham might be spam. 15:07:04 we didn't get as much abuse as i was expecting, so that's good 15:07:21 i'm glad someone else handled the post-quantum crypto comments 15:07:32 i think overall the blog post went well 15:07:39 yes, I think so, too! 15:07:57 there does seem to be some demand for atlas without javascript, i'm wondering if we can add some noscript and links to alternative services 15:07:59 regarding comments, we usually leave those open for a week. 15:08:08 like use an onionoo cli tool or use onionite 15:08:30 hmm 15:08:38 but, they want the graphs 15:08:44 w/o script probably. 15:09:03 gsoc project: add matplotlib to an onionoo cli tool 15:09:06 done 15:09:13 heh 15:09:26 good idea! 15:09:31 :) 15:09:39 onionite generates ascii charts in

 tags, which i thought was innovative
15:10:00  i don't think that at the number of requests that atlas serves, this is practical for us to do
15:10:07  there's a good reason atlas is mostly client side
15:10:17  yes.
15:10:20  and we have good cache hit rates on the backend
15:10:43  well, up to you whether you want to add a link.
15:10:54  maybe, if you do, ask them first if they can handle the load.
15:11:01  yep, will do
15:11:12  i'll point them at our munin graphs also, so they have a clear idea
15:11:19  i don't want to run up their AWS bill or whatever
15:11:25  true!
15:11:32  do you have the munin link handy?
15:11:47  not handy, it's behind a password
15:11:49  I always forget how to put in user name and pass.
15:11:49  but i do have
15:11:59  https://people.torproject.org/~irl/volatile/varnish_cache-day.png
15:12:02  yes, but you can encode them in the url.
15:12:06  that's what happened when we did the blog post
15:12:42  it looks like it died for a little while, then was serving many many requests over the normal rate
15:12:52  (onionoo graphs, not atlas specific)
15:12:56  wow that's a lot of requests...
15:13:01  (:
15:13:04  https://tor-guest:password@munin.torproject.org/varnish-day.html
15:13:10  does that work?
15:13:32  yep
15:13:33  i can access that, yes
15:13:38  y
15:13:39  great.
15:13:56  onionoo is omeiense.torproject.org i think, not the other one
15:13:59  (or maybe it's both?)
15:14:07  that's what I wanted to mention in the next topic,
15:14:10  but I'll mention it now.
15:14:12  it's both!
15:14:20  ah cool (:
15:14:28  weasel was so kind to set up orestis as another front to omeiense.
15:14:41  he did explain which tool talks to which, but I forgot.
15:14:47  awesome
15:14:55  both omeiense and orestis are powered by the same onionoo java process.
15:15:05  and in theory, we could add more orestises.
15:15:20  though another plan would be to add another java process to orestis.
15:15:29  redundancy!
15:15:43  but for now, it seems we can handle load better.
15:15:48  good.
15:16:14  we got some odd things happening at high loads
15:16:25  where one document would load but the bandwidth documents would fail
15:16:29  right!
15:16:30  which caused all kinds of errors
15:16:31  those are fun.
15:16:46  we have tickets for this already
15:16:53  so this will be handled
15:16:57  great!
15:17:13  oh, one other thing on atlas
15:17:27  we're looking at splitting out a new column for not flags flags
15:17:37  sounds good.
15:17:40  we're thinking to call them properties, and these are synthetic
15:17:50  for now, these will be implemented in atlas
15:17:58  sure. my main concern was that the term "flags" is already taken.
15:17:58  but later, i'd like these to be implemented in onionoo
15:18:05  plausible.
15:18:30  if it's not too much effort, it would be great to prototype them in atlas and see which make sense, are useful, etc.
15:18:39  the properties being considered so far are: "not recommended", "ipv6" and another one i forgot
15:18:46  "offline"? :)
15:19:02  my favourite one though is armadev's suggestion: "tshirt" - this relay has earned a tshirt
15:19:06  RaBe: that was it
15:19:31  add a little t-shirt icon?
15:19:35  yep (:
15:19:39  +1
15:19:40  well, we already have "recommended_version" and "running".
15:19:40  well :-)
15:19:50  maybe a little t-shirt with an onion on it
15:19:51  so, "ipv6" and "tshirt" are missing?
15:20:27  the size needs to in the contact line then ;-)
15:20:30  karsten: my preference would be that onionoo gives an array of strings, and there's no logic in atlas for it
15:20:34  "this relay would currently qualify for a t-shirt"
15:20:46  irl: ah, sure.
15:20:47  but, troublesome if there
15:21:00  is a bug triggering the "earned.." message.
15:21:11  right. :)
15:21:19  well, let's just avoid bugs.
15:21:27  indeed
15:21:29  oh, sure :-)
15:21:48  so, can you add an onionoo ticket for that, so that we don't forget?
15:22:14  irl: ^
15:22:15  yes, i can, but i'll do it later today so i can have all the details in it
15:22:21  yes, sure.
15:22:35  it might not happen before AMS anyway, so no big rush.
15:22:47  ok
15:22:56  well, that's all from me on atlas unless there were any questions
15:23:14  no questions, but big thanks for doing the blog post and handling the response!
15:23:23  (:
15:23:37  shall we move on?
15:23:51  yep.
15:24:02  RaBe: of course, big thanks for contributing to atlas for the blog post. :)
15:24:16  - metrics-lib tutorials and web page (karsten|iwakeh)
15:24:27  so, we have some content.
15:24:35  which is ~80% done.
15:24:49  and we have a short todo list.
15:25:00  one todo item was to have syntax highlighting on the page.
15:25:13  only the code part.
15:25:18  RaBe already suggested something.
15:25:22  cool
15:25:27  what?
15:25:35  err, highlight.js?
15:25:53  https://highlightjs.org/
15:26:06  style: vs
15:26:20  there's a source code example on http://metrics-lib.cc-ltd.net/
15:26:22  I have some code here that I can add once we create this page.
15:26:27  hah, right, RaBe, even easier.
15:26:46  so visual.
15:27:04  colourful
15:27:04  code from RaBe, that is.
15:27:32  we can select of about 30 color schemes
15:27:38  or even create a own
15:28:08  right. but it seems we'll just need to make tiny changes for that and don't need to switch the library.
15:28:17  great!
15:28:20  so, we can start with this and improve from there.
15:28:27  is there line numbering possible?
15:28:50  I was wondering about that, too.
15:28:56  but if it's js, what about users without js?
15:29:06  grey
15:29:07  we cannot refer to line numbers if js is required for showing them.
15:29:20  true.
15:30:16  unrelated questions:
15:30:20  - okay to leave out tutorials from metrics-lib.git?
15:30:35  yes
15:30:40  - okay to create https://metrics.torproject.org/metrics-lib.html with new content?
15:30:47  b/c the readme refers to them
15:31:07  hmm?
15:31:12  404?
15:31:23  what do you mean above with
15:31:24  15:31:06 < iwakeh> b/c the readme refers to them
15:31:39  >>  okay to leave out tutorials from metrics-lib.git
15:31:42  (the metrics-lib.html page does not exist yet.)
15:31:58  you mean refer from metrics-lib's README to that URL?
15:32:00  ah, that was a question/topic.
15:32:29  yes, sorry.
15:32:30  yes, didn't we discuss that on the pad?
15:32:37  ah, maybe.
15:32:53  just some questions coming to mind 5 minutes before the meeting.
15:32:59  ok?
15:33:22  alright, I should maybe look at the todo list from that other meeting first. :)
15:33:33  ;-)
15:33:37  there's a JS only line numbering plugin for highlightjs, http://wcoder.github.io/highlightjs-line-numbers.js/ - if we want line numbers for non js users, we'll have to put it in the source code and just style it some way...
15:34:00  uh, no , rather js.
15:34:19  maybe, users without js copy the code into an editor and
15:34:28  can follow the description that way?
15:34:50  if the code description refers to line numbers.
15:34:58  do we have to refer to line numbers?
15:35:02  or even just count the lines if it's a short excerpt...
15:35:44  how about we leave them out for now?
15:35:48  yes.
15:35:53  ok.
15:35:59  we have good comments in the code.
15:35:59  I wouldn't refer to line numbers. If something changed in code ... I would refer to the function names ect...
15:36:17  right.
15:36:23  good point.
15:36:32  and some js less.
15:36:41  ok, so no line numbers for now :)
15:36:48  yep.
15:37:13  okay, should we move on to the next topic, also web-related?
15:37:22  sure.
15:37:28  - onionoo.html and collector.html on Metrics (karsten)
15:37:46  so, we have redesigned index.html files in both git repos.
15:37:59  and we discussed somewhere to put them on metrics.tpo.
15:38:13  should we move forward with that?
15:38:28  I know I said something about waiting for the rebranding discussion to be done,
15:38:31  we could just try the submodules.
15:38:38  but it seems like that's going to take longer.
15:38:50  git submodules.
15:38:56  yes,
15:38:58  and
15:39:20  the rebranding shouldn't make us hide the pages until a decision is reached.
15:39:29  right.
15:39:43  if we change names, we'll put in a redirect.
15:39:54  and the redirect on, say, https://onionoo.torproject.org/ will just be updated.
15:40:25  so, regarding the git submodule approach,
15:40:41  what's the best way to add this to ant?
15:40:56  index.html are just there.
15:41:03  or do you refer to javadoc
15:41:05  ?
15:41:18  ant will have to take the index.html from its submodule and take everything inside 
15:41:22  The example is the approach for metrics-base
15:41:39  and add some jsp and then put it in the .war file.
15:41:45  ah
15:42:22  can't index.html include a file with just the  contents?
15:42:42  not if we want to keep it usable for mirrors.
15:43:37  speaking of, we should probably rip out the metrics design parts from onionoo.git and collector.git and make sure the pages still look good with default bootstrap.
15:43:42  also for mirrors.
15:43:46  Or, ant combines these files?
15:44:13  you mean the ant in onionoo.git just combines it differently than the ant in metrics-web.git?
15:44:27  Oh, you want plain style for mirrors?
15:44:35  yes.
15:44:37  I think we'll need it.
15:44:50  or a minimal style that is not green and purple.
15:45:01  but, low maintenance effort.
15:45:13  okay, ticket maybe?
15:45:17  add minimal style to make thing readable.
15:45:21  yes.
15:45:21  yes.
15:45:48  okay!
15:45:48  well, wouldn't that be just removing the tor stylesheet file and keep the default bootstrap css? :)
15:46:01  that sounds like low maintenance effort!
15:46:01  the javadoc but
15:46:10  the index.html is different?
15:46:48  javadoc is just metrics-lib, right?
15:47:00  and we don't have metrics-lib mirrors.
15:47:04  the others too.
15:47:18  do you want to put them on metrics.tpo, too?
15:47:21  but the javadoc is not on the web-page.
15:47:28  right.
15:47:35  no, they are no apis
15:47:40  right.
15:47:51  what we could do:
15:48:09  not sure if we want to, but we could apply the shiny metrics style only for the javadocs generated by metrics-web.
15:48:14  and keep the default otherwise.
15:48:35  but it doesn't really matter.
15:48:51  ticket!
15:48:53  i think we
15:49:00  can just keep the javadoc as is.
15:49:05  works for me.
15:49:23  alright, moving on?
15:49:26  This is metrics' content not mirror content, the javadoc.
15:49:29  yes.
15:49:46  except when somebody modifies the code on their mirror.
15:50:01  but, it's not on the web, so what.
15:50:07  - OnionPerf integration into CollecTor (karsten|iwakeh)
15:50:08  bthey won't publish the javadoc?
15:50:16  ok next topic.
15:50:26  javadoc on an onionoo mirror?
15:50:30  or collector mirror?
15:50:40  noone would want that.
15:50:40  they'll publish the modified spec.
15:50:43  hehe
15:50:49  OnionPerf!
15:50:54  !
15:51:00  I think we're making progress on the ticket.
15:51:11  step by step.
15:51:12  and hiro is not here, I think.
15:51:13  yes.
15:51:40  btw, the cool thing about onionperf is that we'll get two things out of it:
15:51:49  a replacement for the torperf instances
15:51:57  and new measurements to onion services.
15:52:03  both are already running.
15:52:10  which again?
15:52:21  the 'new'
15:52:36  ah, the same static downloads but to an onion service.
15:52:47  50KB, 1MB, 5MB.
15:52:51  ah, true.
15:53:10  this could be our
15:53:11  5.2. Develop and deploy at least one more user model in addition to the current model.
15:53:20  fine :-)
15:53:23  and people do care about onion service performance.
15:53:32  yes.
15:53:45  anyway, I think we're doing good on the ticket.
15:53:50  let's keep the discussion there.
15:54:01  ok
15:54:08  last topic on the list:
15:54:10  - Tech report draft (iwakeh)
15:54:22  Quite some distractions,
15:54:29  but the week
15:54:30  uhmmmm ;)
15:54:38  is not over yet.
15:54:50  true!
15:54:55  ;-)
15:55:18  Maybe, I should keep the email client off ... :-)
15:55:20  okay, let me know if you have any questions, and of course if you have something to read.
15:55:25  hehehe
15:55:28  yes, sure.
15:55:36  sorry for the distractions.
15:55:48  many things happening in parallel right now.
15:55:48  All that work without readers would be bad .
15:56:06  yes, multitasking.
15:56:20  my hope is that we'll have enough time to finish everything between AMS and the end of june.
15:56:32  that's the plan.
15:56:38  and if we can finish off some stuff now, even better.
15:56:53  right.
15:57:03  okay, looks like we're done.
15:57:08  yep.
15:57:09  12 minutes later than usual.
15:57:15  sigh
15:57:23  :-)
15:57:26  cool. thanks, everyone!
15:57:32  :)
15:57:34  have a good one. bye!
15:57:36  thanks, and bye bye!
15:57:37  bye!
15:57:38  bye
15:57:40  #endmeeting