Displaying #traffic-server/2015-11-17.log:

Tue Nov 17 00:16:29 2015  _klk_:Joined the channel
Tue Nov 17 00:17:04 2015  PSUdaemon:sbeards: i expected more facial hair...
Tue Nov 17 00:17:58 2015  sudheerv:haha
Tue Nov 17 00:23:26 2015  sbeards:heh
Tue Nov 17 00:23:42 2015  sbeards:I tried... but I cannot grow much
Tue Nov 17 00:24:20 2015  sudheerv:even this, is a result of getting married ;)?
Tue Nov 17 00:24:28 2015  sbeards:looks for transaction slot docs
Tue Nov 17 00:24:42 2015  sbeards:sudheerv: no, just laziness
Tue Nov 17 00:24:47 2015  sudheerv:sbeards: https://www.mail-archive.com/dev@trafficserver.apache.org/msg03969.html
Tue Nov 17 00:24:55 2015  sudheerv:lol
Tue Nov 17 00:25:25 2015  sudheerv:TSHttpTxnArgIndexReserve/NameLookup/Set/Get
Tue Nov 17 00:25:53 2015  sbeards:that is better than RTFM :)
Tue Nov 17 00:25:57 2015  sudheerv:ha! this is actually from Mr PSUdaemon
Tue Nov 17 00:26:02 2015  sbeards:or RTFCode
Tue Nov 17 00:26:08 2015  sudheerv:what is RTFCode?
Tue Nov 17 00:26:24 2015  sbeards:read the f*in code
Tue Nov 17 00:26:27 2015  sudheerv:haha
Tue Nov 17 00:26:36 2015  sudheerv:yeah, it is
Tue Nov 17 00:26:57 2015  sudheerv:http://www.slideshare.net/PSUdaemon/slides
Tue Nov 17 00:27:14 2015  sudheerv:these slides from Phil also look great
Tue Nov 17 00:28:46 2015  blattj1:Joined the channel
Tue Nov 17 00:28:55 2015  sbeards:sudheerv: thx
Tue Nov 17 00:29:39 2015  sudheerv:yw
Tue Nov 17 00:39:05 2015  _klk_:Joined the channel
Tue Nov 17 00:42:16 2015  sbeards:btw, our presentation is here: https://tr.im/ysearchplugins
Tue Nov 17 00:48:48 2015  zwoop:cool
Tue Nov 17 00:49:05 2015  zwoop:I guess I should email a reminder, but we'll ask all authors to add all slides to a document on the Wiki
Tue Nov 17 00:49:17 2015  zwoop:I'll try to setup the Wiki tonight, assuming I'm not passing out
Tue Nov 17 00:49:44 2015  _klk_1:Joined the channel
Tue Nov 17 00:50:48 2015  PSUdaemon:did not know he was the Txn slot expert...
Tue Nov 17 00:50:57 2015  PSUdaemon:good to know i have a niche
Tue Nov 17 00:51:05 2015  PSUdaemon:i'll be sure to put that on my yearly review
Tue Nov 17 00:51:07 2015  jpeach:PSUdaemon you *are* a txn slot!
Tue Nov 17 00:52:34 2015  amc:He's always looked more like a session slot to me.
Tue Nov 17 00:55:01 2015  sudheerv:lol
Tue Nov 17 01:05:41 2015  bcall:zwoop: slides - https://cwiki.apache.org/confluence/display/TS/Presentations+-+2015#Presentations-2015-2015FallSummit
Tue Nov 17 01:06:47 2015  PSUdaemon:aziz! light!
Tue Nov 17 01:06:56 2015  zwoop:bcall thanks!
Tue Nov 17 01:46:17 2015  biilmann:Joined the channel
Tue Nov 17 02:19:16 2015  dxu_:Joined the channel
Tue Nov 17 02:25:41 2015  reveller1:Joined the channel
Tue Nov 17 03:18:12 2015  ibrezac:Joined the channel
Tue Nov 17 03:19:21 2015  pyryq:Joined the channel
Tue Nov 17 03:19:26 2015  oag:Joined the channel
Tue Nov 17 03:25:39 2015  PSUdaemon:Joined the channel
Tue Nov 17 03:28:26 2015  biilmann:Joined the channel
Tue Nov 17 03:32:58 2015  ggherdov:Joined the channel
Tue Nov 17 03:35:28 2015  masaori:Joined the channel
Tue Nov 17 03:38:35 2015  _klk_:Joined the channel
Tue Nov 17 04:15:38 2015  psp:Joined the channel
Tue Nov 17 04:47:14 2015  dxu_:Joined the channel
Tue Nov 17 04:50:31 2015  boargod2:Joined the channel
Tue Nov 17 04:53:52 2015  felicity_:Joined the channel
Tue Nov 17 05:00:28 2015  msekimura:Joined the channel
Tue Nov 17 05:01:13 2015  dcarlin:Joined the channel
Tue Nov 17 05:01:42 2015  amc:Joined the channel
Tue Nov 17 05:02:40 2015  jsime:Joined the channel
Tue Nov 17 05:02:52 2015  Zerpex:Joined the channel
Tue Nov 17 05:04:38 2015  psp:Joined the channel
Tue Nov 17 05:18:08 2015  ihabunek_:Joined the channel
Tue Nov 17 05:18:24 2015  swoc:Joined the channel
Tue Nov 17 05:23:27 2015  dcarlin-:Joined the channel
Tue Nov 17 05:23:48 2015  skaji_:Joined the channel
Tue Nov 17 05:23:56 2015  lep2:Joined the channel
Tue Nov 17 05:25:37 2015  jpeach:Joined the channel
Tue Nov 17 06:01:43 2015  biilmann:Joined the channel
Tue Nov 17 06:29:24 2015  Randeep:Joined the channel
Tue Nov 17 06:32:24 2015  Randeeppr:Joined the channel
Tue Nov 17 06:42:15 2015  Randeep_pr:Joined the channel
Tue Nov 17 06:46:21 2015  Randeeppr:Joined the channel
Tue Nov 17 07:05:45 2015  Randeep:Joined the channel
Tue Nov 17 07:14:21 2015  dxu_:Joined the channel
Tue Nov 17 07:23:29 2015  biilmann:Joined the channel
Tue Nov 17 07:40:00 2015  boargod:Joined the channel
Tue Nov 17 08:06:47 2015  blattj:Joined the channel
Tue Nov 17 08:08:32 2015  blattj1:Joined the channel
Tue Nov 17 08:34:32 2015  es:Joined the channel
Tue Nov 17 08:39:26 2015  biilmann:Joined the channel
Tue Nov 17 08:46:16 2015  Randeep:Joined the channel
Tue Nov 17 09:06:16 2015  Lethalman:Joined the channel
Tue Nov 17 09:19:06 2015  Humbedooh:Joined the channel
Tue Nov 17 09:41:32 2015  dxu_:Joined the channel
Tue Nov 17 10:28:24 2015  niq:Joined the channel
Tue Nov 17 11:25:11 2015  Randeeppr:Joined the channel
Tue Nov 17 12:25:36 2015  Randeep:Joined the channel
Tue Nov 17 12:29:39 2015  Randeeppr:hi, I'm playing MPEG dash streaming from ats. Videos are playing but if I increase the number of stream, videos are stuttering. Its like in every 2s there is a kind of pause.
Tue Nov 17 12:29:53 2015  Randeeppr:Anyone has experience similar??
Tue Nov 17 13:10:14 2015  reveller:Joined the channel
Tue Nov 17 13:40:47 2015  niq:Joined the channel
Tue Nov 17 13:40:47 2015  niq:Joined the channel
Tue Nov 17 14:14:56 2015  ibrezac:Joined the channel
Tue Nov 17 14:28:44 2015  esproul:Joined the channel
Tue Nov 17 15:24:12 2015  sudheerv:randeeppr: do you have access logs
Tue Nov 17 15:24:38 2015  sudheerv:sounds like some segments may be failing to download..
Tue Nov 17 15:43:02 2015  Randeeppr:Joined the channel
Tue Nov 17 15:49:26 2015  Randeeppr:sudheerv: yes I have access logs
Tue Nov 17 15:50:42 2015  sudheerv:Randeeppr: check the access logs for any errors
Tue Nov 17 15:53:12 2015  JSeymour:Joined the channel
Tue Nov 17 15:55:34 2015  Randeeppr:couldnt find any errors in squid.blog
Tue Nov 17 16:12:41 2015  bahumbug:Joined the channel
Tue Nov 17 16:24:22 2015  shinrich1:Joined the channel
Tue Nov 17 16:31:06 2015  luc:Randeeppr: are you sure, that your encoding is correct ?
Tue Nov 17 16:32:13 2015  luc:Randeeppr: i use ats for a thousand vod files without problem in dash, smotthstreaming or hls
Tue Nov 17 16:32:52 2015  Randeeppr:luc: yes only happening to some streams
Tue Nov 17 16:33:21 2015  Randeeppr:suppose if I play the same stream in two tabs in browser, one may work fine and other one may stutter
Tue Nov 17 16:37:25 2015  biilmann:Joined the channel
Tue Nov 17 16:37:53 2015  Randeeppr:mine are encrypted dash streams. I play them using shaka player. I do this to benchmark the hardware. Like how many concurrent streams it supports with its hardware
Tue Nov 17 16:38:03 2015  blattj:Joined the channel
Tue Nov 17 16:38:55 2015  luc:Randeeppr: i've not tested in the browser many streams (just one for test)
Tue Nov 17 16:39:22 2015  luc:Randeeppr: if i need to bench ats, i use a script that download each segments for each vod in parallel
Tue Nov 17 16:39:50 2015  jrushford:Joined the channel
Tue Nov 17 16:39:53 2015  blattj:hangout link for today: https://plus.google.com/hangouts/_/yahoo-inc.com/ats
Tue Nov 17 16:40:15 2015  Randeeppr:luc: actual devices to play streams will be set top boxes.
Tue Nov 17 16:41:35 2015  Randeeppr:luc: this live steams mpd will refresh in every 10s and the video chunks too. downloading segments will not work for this I think.!!!or!
Tue Nov 17 16:43:48 2015  luc:even with live stream you can simulate download chunck without decoding each files
Tue Nov 17 16:45:34 2015  Randeeppr:This idea didnt occur to me. I'll check this tomorrow. (I work in IST time)
Tue Nov 17 16:46:28 2015  Randeeppr:luc: Thank you :)
Tue Nov 17 16:46:42 2015  luc:well it's my end of the day for me ;)
Tue Nov 17 16:47:45 2015  Randeeppr:its 10:15pm here :) usually at my office time i dont see anyone here :P thanks again.
Tue Nov 17 16:48:58 2015  kang1:Joined the channel
Tue Nov 17 16:49:19 2015  kang2:Joined the channel
Tue Nov 17 17:02:29 2015  psp:Joined the channel
Tue Nov 17 17:09:10 2015  Randeep:Joined the channel
Tue Nov 17 17:10:32 2015  bcall:jpeach: on cb35e6ca143dd8618e0b73877853a32b3f6e4474 - is apply_options being called multiple times now?
Tue Nov 17 17:10:56 2015  jpeach:possibly, I looked at it and I thought it was ok
Tue Nov 17 17:12:27 2015  meeramn:Joined the channel
Tue Nov 17 17:21:29 2015  biilmann:Joined the channel
Tue Nov 17 17:24:57 2015  jpeach:hey Humbedooh?
Tue Nov 17 17:32:14 2015  bcall:jpeach: how often is it called? is it called before and after the connect()?
Tue Nov 17 17:32:42 2015  jpeach:it has to be called after setting the sockaddr since it looks at the address type
Tue Nov 17 17:34:34 2015  reveller1:Joined the channel
Tue Nov 17 17:38:33 2015  sbeards:any way I can add a link to our slides on this wiki? https://cwiki.apache.org/confluence/display/TS/Presentations+-+2015
Tue Nov 17 17:38:42 2015  sbeards:I don't see an edit button when I log in
Tue Nov 17 17:38:53 2015  bcall:jpeach: no jira ticket :) - dc82d62540a8e89c4439fd93e2d116cd9bd37fbb
Tue Nov 17 17:39:43 2015  jpeach:bcall: yes I am lazy
Tue Nov 17 18:03:06 2015  biilmann:Joined the channel
Tue Nov 17 18:03:13 2015  blattj:https://plus.google.com/hangouts/_/yahoo-inc.com/ats
Tue Nov 17 18:03:18 2015  jrushford:Joined the channel
Tue Nov 17 18:12:36 2015  es:Joined the channel
Tue Nov 17 18:15:04 2015  reveller1:Left the channel
Tue Nov 17 18:18:48 2015  niq:Joined the channel
Tue Nov 17 18:27:16 2015  Randeep_pr:Joined the channel
Tue Nov 17 18:29:23 2015  Randeeppr:audio is fine for everyone? or its just for me?
Tue Nov 17 18:29:42 2015  shinrich1:Is the main speaker a problem?
Tue Nov 17 18:29:45 2015  shinrich1:Or audience?
Tue Nov 17 18:29:47 2015  shinrich1:Or both?
Tue Nov 17 18:30:19 2015  shinrich1:Main speaker probably needs to move his mic up
Tue Nov 17 18:34:28 2015  Randeeppr:main speaker
Tue Nov 17 18:34:53 2015  Randeeppr:sometimes I dont hear anything he says
Tue Nov 17 18:35:21 2015  shinrich1:No better now? He moved up the mic. But he's a pretty soft spoken speaker anyway.
Tue Nov 17 18:35:46 2015  dcarlin-:shinrich1: other people complainted earlier on yahoo irc about audio cutting in/out
Tue Nov 17 18:35:53 2015  dcarlin-:how do the mics get onto hangouts?
Tue Nov 17 18:36:04 2015  shinrich1:Audio conference in
Tue Nov 17 18:36:35 2015  Randeeppr:better now
Tue Nov 17 18:37:03 2015  Randeeppr:but audio is cutting frequently
Tue Nov 17 18:37:27 2015  Randeeppr:when audience speak its very clear
Tue Nov 17 18:37:54 2015  shinrich1:I'll check battery on next speaker change.
Tue Nov 17 18:43:06 2015  shinrich1:Sound still ok? Working off the room speakers now.
Tue Nov 17 18:44:28 2015  Randeeppr:better.
Tue Nov 17 18:58:35 2015  ibrezac:Joined the channel
Tue Nov 17 19:07:18 2015  Randeeppr:last speaker had spoke about parent_proxy = false is that slide uploaded somewhere?
Tue Nov 17 19:08:16 2015  dcarlin:zwoop: are the slides being gathered from all the presentations?
Tue Nov 17 19:08:35 2015  dcarlin:Randeeppr: the current speaker is amc
Tue Nov 17 19:09:48 2015  Randeeppr:dcarlin: I meant the one who was showing slide. I think someone from comcast.
Tue Nov 17 19:10:41 2015  bcall:dcarlin: people should add them here - https://cwiki.apache.org/confluence/display/TS/Presentations+-+2015#Presentations-2015-2015FallSummit
Tue Nov 17 19:10:53 2015  biilmann:Joined the channel
Tue Nov 17 19:11:29 2015  dcarlin:bcall: doesn't seem terribly successful in the past :)
Tue Nov 17 19:11:36 2015  bcall:yes :)
Tue Nov 17 19:11:51 2015  shinrich1:We can hunt people down here…
Tue Nov 17 19:12:34 2015  ibrezac:Joined the channel
Tue Nov 17 19:13:06 2015  bcall:they can't leave until they give us slides
Tue Nov 17 19:14:20 2015  zwoop:bcall way to rule with an iron fist!!! +1
Tue Nov 17 19:22:19 2015  kang1:Joined the channel
Tue Nov 17 19:23:02 2015  ibrezac:Joined the channel
Tue Nov 17 19:40:55 2015  ibrezac:Joined the channel
Tue Nov 17 19:41:51 2015  jrushford:Joined the channel
Tue Nov 17 19:51:51 2015  _klk_:Joined the channel
Tue Nov 17 19:59:56 2015  pdm:hey-o, is the custom_error.so plugin some how built in or should I see it in the plugins dir if I have it available ?
Tue Nov 17 20:09:20 2015  kang1:Joined the channel
Tue Nov 17 20:25:59 2015  _klk_:Joined the channel
Tue Nov 17 20:32:42 2015  jpeach:pdm: custom_error must be some vendor plugin ... I don't see it in the apache repository
Tue Nov 17 20:33:04 2015  pdm:interesting, that is why I can't find it either
Tue Nov 17 20:33:15 2015  pdm:i'm tr ying to remap 404's to a 'pretty' web page instead of 'not found on accelerator'
Tue Nov 17 20:34:08 2015  pdm:this is what I found, but it wasn't computing to what I was seeing available to me https://cwiki.apache.org/confluence/display/TS/Per+Remap+Custom+Errors
Tue Nov 17 20:35:21 2015  sbeards:bcall: how do I edit that page? Do I need access?
Tue Nov 17 20:35:35 2015  bcall:yes, you will need access
Tue Nov 17 20:35:52 2015  bcall:or you can give me a link to your slides and I can add it for you
Tue Nov 17 20:36:22 2015  sbeards:https://tr.im/ysearchplugins
Tue Nov 17 20:36:39 2015  PSUdaemon:jpeach: nice review
Tue Nov 17 20:36:50 2015  PSUdaemon:not that that isn't always the case
Tue Nov 17 20:36:56 2015  PSUdaemon:but i just happened to read that last one
Tue Nov 17 20:36:57 2015  jpeach:thanks PSUdaemon
Tue Nov 17 20:37:03 2015  bcall:sbeards: what do you want as the title?
Tue Nov 17 20:37:22 2015  sbeards:Yahoo Search ATS Plugins
Tue Nov 17 20:37:22 2015  jpeach:PSUdaemon: can you explain https://github.com/apache/trafficserver/pull/304 to me?
Tue Nov 17 20:39:17 2015  amc:Shouldn't you be able to do a pretty page using configuration?
Tue Nov 17 20:40:03 2015  pdm:I was trying seraching to find out how and this is what Icame up with so odn't know the right/easy way
Tue Nov 17 20:40:07 2015  bcall:sbeards: done
Tue Nov 17 20:40:12 2015  zwoop:sbeards send an email to Miles Libbey asking for write access to the Wiki
Tue Nov 17 20:40:17 2015  PSUdaemon:jpeach: i think we might be getting lucky in that we don't have all those other locks around
Tue Nov 17 20:40:18 2015  amc:Does the escalate plugin do this?
Tue Nov 17 20:41:05 2015  PSUdaemon:jpeach: so, essentially, the problem is real and verified
Tue Nov 17 20:41:10 2015  PSUdaemon:but maybe the solution in incomplete
Tue Nov 17 20:41:41 2015  PSUdaemon:does that help?
Tue Nov 17 20:42:12 2015  sbeards:amc: Y! escalate plugin intercepts certain status codes from a successful origin txn
Tue Nov 17 20:42:29 2015  sbeards:(we should OS that btw)
Tue Nov 17 20:43:13 2015  pdm:I see docs on the escalate plugin (cwiki page) but it has same info as the custom_error one I pasted above
Tue Nov 17 20:43:27 2015  PSUdaemon:it is weird that we have that mutex in there though and that doesn't fix it
Tue Nov 17 20:43:33 2015  amc:I think it's already OS - it appears to be in the ASF repo in the experimental plugins.
Tue Nov 17 20:43:46 2015  jpeach:PSUdaemon: test it?
Tue Nov 17 20:43:51 2015  pdm:I guess.. hmm I could have a 'final' map at bottom for this, all i want is if there is no matching remap 'do this'
Tue Nov 17 20:43:59 2015  sbeards:amc: Yahoo's is very different
Tue Nov 17 20:44:04 2015  pdm:I want to avoid 'not found on accelerator'
Tue Nov 17 20:44:21 2015  jpeach:pdm: use the error message templates?
Tue Nov 17 20:44:29 2015  sbeards:pdm: look at the body factory templates
Tue Nov 17 20:45:01 2015  pdm:thanks will take a look, I dont want to serve up a custom page, I want to redirect
Tue Nov 17 20:45:29 2015  amc:http://trafficserver.readthedocs.org/en/stable/reference/configuration/records.config.en.html#customizable-user-response-pages
Tue Nov 17 20:45:38 2015  sbeards:pdm: what we do is send a 404 still but then do a JS-based redirect or meta refresh
Tue Nov 17 20:45:42 2015  amc:For redirection, that's a bit different.
Tue Nov 17 20:46:10 2015  sbeards:pdm: see an example: https://search.yahoo.com/beer
Tue Nov 17 20:46:14 2015  PSUdaemon:TS-4025
Tue Nov 17 20:46:20 2015  pdm:sbeards: that seems ideal-ish, still give info (you made an error) but dont' leave them on a dumb page
Tue Nov 17 20:46:27 2015  sbeards:yep!
Tue Nov 17 20:46:42 2015  pdm:thanks i'll give this a shot
Tue Nov 17 20:47:09 2015  sbeards:also gives you a chance to beacon the error to your monitoring
Tue Nov 17 20:47:16 2015  sbeards:(if you need that)
Tue Nov 17 20:47:24 2015  PSUdaemon:jpeach: ?
Tue Nov 17 20:47:33 2015  PSUdaemon:jpeach: we run this in prod and has fixed our issue
Tue Nov 17 20:47:36 2015  PSUdaemon:if that is what you are asking
Tue Nov 17 20:48:07 2015  jpeach:I guess I don't understand the issue ... can you explain the root cause to me?
Tue Nov 17 20:49:05 2015  PSUdaemon:i'll have to look at it more in depth again
Tue Nov 17 20:49:18 2015  PSUdaemon:i don't recall exactly
Tue Nov 17 20:49:35 2015  PSUdaemon:it was only a code for the author because of hardening
Tue Nov 17 20:49:40 2015  PSUdaemon:for me it was mem corruption
Tue Nov 17 20:49:59 2015  PSUdaemon:so it's not obvious in the code where it's segfaulting, cause it's not
Tue Nov 17 20:52:30 2015  jpeach:I'm sure this will fix *something*, since it serializes on a global lock
Tue Nov 17 21:03:59 2015  mlibbey:sbeards. if you send me (mlibbey@apache) your cwiki username, will give you access to it.
Tue Nov 17 21:10:34 2015  gancho:Joined the channel
Tue Nov 17 21:16:16 2015  sbeards:mlibbey: done, thx
Tue Nov 17 21:27:12 2015  kang2:Joined the channel
Tue Nov 17 21:28:09 2015  blattj:Joined the channel
Tue Nov 17 21:31:17 2015  jrushford:Joined the channel
Tue Nov 17 21:40:30 2015  jpeach:y'know what would be cool? a diags option to print a stack trace on a diags message
Tue Nov 17 21:41:52 2015  jpeach:also, diagnostic on location might be helpful
Tue Nov 17 21:41:58 2015  PSUdaemon:jpeach: http://cdn.meme.am/instances/500x/37287023.jpg
Tue Nov 17 21:42:18 2015  jpeach:seriously
Tue Nov 17 21:50:57 2015  zwoop:jpeach I've done something like that sometimes, where I needed to see a stack trace at a particular place in the code. But I manually added calls to ink_stack_trace_dump()
Tue Nov 17 21:52:43 2015  Uri:Joined the channel
Tue Nov 17 21:55:54 2015  dcarlin:kichan: proxy.config.res_track_memory
Tue Nov 17 21:58:13 2015  kichan:cool
Tue Nov 17 21:58:57 2015  Uri:Client IP based debugging is already available via plugin -- why add it to the core??
Tue Nov 17 21:59:25 2015  Uri:Can someone ask the speaker?
Tue Nov 17 22:00:11 2015  Uri:@zwoop -- IIRC you also added an option in the header_rewrite for it
Tue Nov 17 22:00:24 2015  zwoop:uri yeah
Tue Nov 17 22:00:37 2015  zwoop:this still requires diags to be enabled though (uri as we discussed on email :)
Tue Nov 17 22:00:52 2015  Uri:So?
Tue Nov 17 22:00:53 2015  zwoop:just enabling diags, but an empty regex, has about a 50% penalty on performance
Tue Nov 17 22:01:12 2015  Uri:And this option overrides the debug_enabled flag?
Tue Nov 17 22:01:21 2015  zwoop:the IP one? I don't know
Tue Nov 17 22:01:23 2015  zwoop:amc would know
Tue Nov 17 22:01:31 2015  Uri:@amc?
Tue Nov 17 22:01:39 2015  Uri:^
Tue Nov 17 22:01:39 2015  sudheerv:Uri: the ip one is marked as debuggable in the netvc early on
Tue Nov 17 22:01:48 2015  zwoop:my guess is that it's not (i.e. I bet you still have to enable diags)
Tue Nov 17 22:01:52 2015  sudheerv:once the connection is marked, no regex needed
Tue Nov 17 22:01:58 2015  sudheerv:atleast, this was the version i wrote last year
Tue Nov 17 22:02:14 2015  sudheerv:since then, it has been changed a bit and i think it now supports regex i suppose
Tue Nov 17 22:02:14 2015  jpeach:I think the SSL wire trace config should be proxy.config.ssl.wiretrace.*
Tue Nov 17 22:02:27 2015  sudheerv:but, still it only runs regex on the session marked as debuggable
Tue Nov 17 22:02:29 2015  amc:Dr. H did that implementation, I'll bug her.
Tue Nov 17 22:02:59 2015  sudheerv:the main difference is running regex on *all* requests vs the only ones marked debuggable
Tue Nov 17 22:03:04 2015  jpeach:zwoop: is that using PCRE JIT?
Tue Nov 17 22:03:46 2015  zwoop:sudheerv right. I still think this requires diags enabled, right ?
Tue Nov 17 22:04:00 2015  sudheerv:zwoop: no, it's not required afaik
Tue Nov 17 22:04:06 2015  sudheerv:it uses the other setting debug.client_ip
Tue Nov 17 22:04:06 2015  zwoop:hmmm, ok
Tue Nov 17 22:04:12 2015  Uri:There's the session hook that happens very early on
Tue Nov 17 22:04:20 2015  sudheerv:Uri: yes, it can be done with a plugin
Tue Nov 17 22:04:24 2015  zwoop:well, the per txn tracing, as set via plugin APIs, also requires diags to be enabled
Tue Nov 17 22:04:28 2015  sudheerv:but, the point is to make it easily configurable for ops
Tue Nov 17 22:04:34 2015  amc:We did it that early because otherwise you can't discover SSL handshake errors.
Tue Nov 17 22:04:36 2015  sudheerv:rather than going with plugins which are not easy for ops
Tue Nov 17 22:04:43 2015  zwoop:sudheerv then how does it know which Debug() statement to print?
Tue Nov 17 22:04:49 2015  Uri:Why aren't plugins easy?
Tue Nov 17 22:05:08 2015  sudheerv:zwoop: it still runs the debug tags for the netvc that's already marked as debuggable
Tue Nov 17 22:05:10 2015  Uri:It can always be loaded, and hook according to config?
Tue Nov 17 22:05:29 2015  amc:it still can't see any of the original SSL handshake data or operations.
Tue Nov 17 22:05:57 2015  mlibbey:sbeards: wiki permissions changed
Tue Nov 17 22:05:58 2015  Uri:I said this before in github -- adding this logic inside the core just doesn't make sense to me - Plugin space is easier to expand
Tue Nov 17 22:05:59 2015  sudheerv:Uri: this is a philosophical argument, but, my view is that anything that can be directly configurable is far simpler than having to use plugins and separate config
Tue Nov 17 22:05:59 2015  jpeach:fwiw I also lean towards a plugin implementation
Tue Nov 17 22:06:09 2015  zwoop:sudheerv ok, I think I understand. This is sort of confusing since it's not behaving the same as the API for txn diags.
Tue Nov 17 22:06:15 2015  sudheerv:zwoop: yeah
Tue Nov 17 22:06:35 2015  Uri:And that's how you end up with an FTP/NTLM/Kerberos implementation inside the core
Tue Nov 17 22:06:39 2015  Uri:shudders
Tue Nov 17 22:06:49 2015  sudheerv:Uri: i think it's a question of tradeoff
Tue Nov 17 22:06:55 2015  Uri:Everything is
Tue Nov 17 22:07:02 2015  sudheerv:i think these are things which are useful to *everyone*
Tue Nov 17 22:07:07 2015  sudheerv:and not use case specific
Tue Nov 17 22:07:12 2015  sudheerv:hence, it makes more sense to have it in the core
Tue Nov 17 22:07:14 2015  sudheerv:vs a plugin
Tue Nov 17 22:07:21 2015  sudheerv:atleast, that's my view :)
Tue Nov 17 22:07:21 2015  jpeach:can the core be changed to make this use case better/possible/reaonable?
Tue Nov 17 22:07:25 2015  sudheerv:we could agree to disagree
Tue Nov 17 22:07:26 2015  Uri:Sure - I agree it's useful (I added the infra for supporting this)
Tue Nov 17 22:07:53 2015  Uri:But the idea was for the logic to be in plugin space
Tue Nov 17 22:07:57 2015  jpeach:sudheerv, Uri: notice also that this is yet another request classification mechanic
Tue Nov 17 22:08:26 2015  sbeards:mlibbey: thx!
Tue Nov 17 22:09:01 2015  Uri:@jpeach yep
Tue Nov 17 22:09:38 2015  Uri:And I honestly think doing these in plugin/external process space instead of complicating the core for all users makes much better sense
Tue Nov 17 22:09:52 2015  amc:I agree user space is easier to expand but ultimately there is data you want you can't get if it has to be turned on by the plugin.
Tue Nov 17 22:09:58 2015  Uri:Just like you wouldn't insert geoip lookups into the core
Tue Nov 17 22:10:01 2015  sudheerv:Uri: jpeach: my view is that if *every* ats deployment will find something useful, it should be in the core
Tue Nov 17 22:10:28 2015  shinrich1:Plus the code change for the client_ip debug is pretty minimal
Tue Nov 17 22:10:28 2015  sudheerv:sure, plugins are great
Tue Nov 17 22:10:31 2015  amc:I don't think utility is a good metric - header_rewrite is useful in basically all deployments but it's in a plugin.
Tue Nov 17 22:10:44 2015  sudheerv:but, the point is the commonality
Tue Nov 17 22:11:05 2015  sudheerv:amc: i dont think i agree with that
Tue Nov 17 22:11:05 2015  Uri:@amc - logs emitted before the session start hook?
Tue Nov 17 22:11:11 2015  sudheerv:utility should be very importantt
Tue Nov 17 22:11:17 2015  sudheerv:factor in making a decision
Tue Nov 17 22:11:28 2015  amc:I think it should be in core if and only if it can not be done in a plugin.
Tue Nov 17 22:11:36 2015  Uri:@amc +100
Tue Nov 17 22:11:42 2015  sudheerv:well, i disagree :)
Tue Nov 17 22:11:45 2015  jpeach:but avoiding and reducing complexity is the most important
Tue Nov 17 22:11:47 2015  sudheerv:but, so be it :)
Tue Nov 17 22:11:57 2015  biilmann:Joined the channel
Tue Nov 17 22:12:55 2015  Uri:Sorry for picking on this -- but I spent some time with a 2006 version of YTS that had all sort of fun things in the core.... making me into something of a minimalist
Tue Nov 17 22:13:12 2015  sudheerv:Uri: I understand your point of view..
Tue Nov 17 22:13:23 2015  sudheerv:just saying that something that's super useful to ops should be provided as a native option
Tue Nov 17 22:13:25 2015  Uri:(which fits my location pretty well -- at DockerCon EU :-)
Tue Nov 17 22:13:30 2015  sudheerv:and not as a plugin which requires additional config
Tue Nov 17 22:13:45 2015  sudheerv:the value and utiilty of debug things are just huge
Tue Nov 17 22:13:59 2015  Uri:I'd definitely support configuring a debug plugin as default
Tue Nov 17 22:14:02 2015  jpeach:I am very sympathetic to the minimalist view, but there can also be significant complexity in many plugins that do similar things or inteact in complex ways
Tue Nov 17 22:14:11 2015  sudheerv:Uri: what does that mean really? :)
Tue Nov 17 22:14:23 2015  sudheerv:a plugin enabled by default is equivalent to having it in the core
Tue Nov 17 22:14:24 2015  sudheerv::)
Tue Nov 17 22:14:30 2015  Uri:Not really
Tue Nov 17 22:14:50 2015  Uri:If I find a bug/have any objection/just feel like it I can easily disable it
Tue Nov 17 22:15:07 2015  sudheerv:you can do the same in the core too :)
Tue Nov 17 22:15:09 2015  sudheerv:it's configurable/optional
Tue Nov 17 22:15:16 2015  sudheerv:not enabled by default
Tue Nov 17 22:15:35 2015  jpeach:let me put it another way ... the plugin API *should* be powerful enough that this could be reasonably done as a plugin
Tue Nov 17 22:15:38 2015  sudheerv:infact, it's (slightly) more work to disable a plugin vs turning off a records.config option
Tue Nov 17 22:16:23 2015  Uri:@sudheerv - The code will always be there. Will always have to be taken into account, will inflate our (already massive) core
Tue Nov 17 22:16:48 2015  sudheerv:Uri: sure, i think the tradeoff here pays off, since the code change was super small
Tue Nov 17 22:16:55 2015  amc:jpeach - Possibly. You'd need a hook that is called immediately after the TCP connect before any data transits.
Tue Nov 17 22:16:57 2015  sudheerv:shinrich1 has all the details, but, the code change is really really small
Tue Nov 17 22:17:23 2015  Uri:How about putting this to a vote?
Tue Nov 17 22:17:33 2015  sudheerv:sounds good
Tue Nov 17 22:17:46 2015  sudheerv:i think that's reasonable :)
Tue Nov 17 22:18:20 2015  amc:This is the reason the SSL stuff was done with hooks instead of directly in the core, e.g. SSL_PREACCEPT_HOOK.
Tue Nov 17 22:18:25 2015  shinrich1:There is a diff on the bug. I made some tweaks. I'll get a pull request out for broader discussion probably Wednesday or Friday.
Tue Nov 17 22:18:42 2015  Uri:RE making pull requests mandatory -- We're talking about every single commit, no matter how trivial -- right?
Tue Nov 17 22:18:48 2015  amc:Right.
Tue Nov 17 22:18:55 2015  amc:Every *code* change.
Tue Nov 17 22:19:02 2015  sudheerv:Uri: i think autoconfig stuff, docs etc should be exceptions :)
Tue Nov 17 22:19:09 2015  sudheerv:(atleast, i hope that's the case)
Tue Nov 17 22:19:22 2015  sudheerv:amc: yes
Tue Nov 17 22:19:26 2015  sudheerv:agree - *code* change
Tue Nov 17 22:19:28 2015  sudheerv::)
Tue Nov 17 22:19:32 2015  Uri:Autoconfig is a very fragile area
Tue Nov 17 22:19:38 2015  Uri:Agree regarding docs
Tue Nov 17 22:20:00 2015  sudheerv:yeah, but, given that, it's mostly zwoop that handles the pain, it's probably easier on him to not create a PR every time :)
Tue Nov 17 22:20:09 2015  Uri:(autoconfig changes broke the build on various setups quite often)
Tue Nov 17 22:20:50 2015  Uri:@zwoop & @jpeach - you're the guys that usually handle this area - your thoughts?
Tue Nov 17 22:22:55 2015  jpeach:build breakages are usually someone fixing a local problem without considering other platforms; I think they have a higher threshold than docs because the testing is harder and there is breakage
Tue Nov 17 22:22:56 2015  sudheerv:Uri: btw, i think there will be a vote on that too..
Tue Nov 17 22:23:07 2015  sudheerv:jpeach: agree
Tue Nov 17 22:23:19 2015  sudheerv:but, also need to consider the *overhead*
Tue Nov 17 22:23:25 2015  jpeach:I usually just merge "obviously correct" docs changes
Tue Nov 17 22:23:42 2015  jpeach:sudheerv: build changes have the highest overhead on test
Tue Nov 17 22:23:52 2015  sudheerv:yeah, if they break :)
Tue Nov 17 22:23:59 2015  sudheerv:what's the % of breaks
Tue Nov 17 22:24:02 2015  sudheerv::)
Tue Nov 17 22:24:02 2015  Uri:They break quite often
Tue Nov 17 22:24:18 2015  sudheerv:ok..
Tue Nov 17 22:24:29 2015  jpeach:yup, which is why we needed to invest in the build farm ... we could not keep it stable without that
Tue Nov 17 22:25:03 2015  Uri:Is it strong enough to do a build-on-pull request?
Tue Nov 17 22:25:11 2015  sudheerv:yeah, we could go for a vote on this - but, the primary reason for the PR i think was *release* quality improvement
Tue Nov 17 22:25:20 2015  sudheerv:if the build broke, technically, there's no release yet :)
Tue Nov 17 22:25:27 2015  sudheerv:anyway, i'm fine with whatever everyone agrees
Tue Nov 17 22:25:38 2015  Uri:likewise
Tue Nov 17 22:25:49 2015  jpeach:Uri: probably; the problem is the git hook integration w/ ASF policy :-/
Tue Nov 17 22:26:28 2015  Uri:What do you mean? It can receive the hook directly from Github
Tue Nov 17 22:26:40 2015  Uri:(that's what I currently do for my setup)
Tue Nov 17 22:27:17 2015  jpeach:we need asf infra to set up the github webhook ... maybe zwoop can get them to do that?
Tue Nov 17 22:28:11 2015  Uri:@zwoop -- don't you have have admin rights for the github project?
Tue Nov 17 22:28:20 2015  jpeach:nope
Tue Nov 17 22:28:34 2015  Uri::-(
Tue Nov 17 22:43:26 2015  princess_susan:hehe
Tue Nov 17 22:48:02 2015  princess_susan:Thanks Dan!
Tue Nov 17 22:49:54 2015  PSUdaemon:Joined the channel
Tue Nov 17 22:50:53 2015  shinrich1:Joined the channel
Tue Nov 17 22:58:21 2015  es:Joined the channel
Tue Nov 17 23:00:02 2015  es:Left the channel
Tue Nov 17 23:08:59 2015  _klk_:Joined the channel
Tue Nov 17 23:12:59 2015  niq:blames zwoop
Tue Nov 17 23:13:13 2015  zwoop:obviously
Tue Nov 17 23:13:17 2015  dcarlin:Left the channel
Tue Nov 17 23:14:26 2015  niq:re: pull requests for everything
Tue Nov 17 23:14:41 2015  niq:Idea: just fine.
Tue Nov 17 23:14:54 2015  niq:Decision-making process: erm .....
Tue Nov 17 23:18:12 2015  Uri:The description for PFS isn't accurate
Tue Nov 17 23:18:37 2015  Uri:Brute force attacks are still possible - it's just that they are needed per stream
Tue Nov 17 23:18:55 2015  Uri:and the compromise of the private key won't compromise everything
Tue Nov 17 23:20:22 2015  Uri:BTW @jpeach and @zwoop -- Do you really force Apple ATS to only support PFS ciphers?
Tue Nov 17 23:20:40 2015  Uri:(At least that's what ssllabs says)
Tue Nov 17 23:20:58 2015  jpeach:Uri: what domain did you check?
Tue Nov 17 23:21:22 2015  Uri:Not a domain -- As a client -- hold on, I'll send a link
Tue Nov 17 23:22:13 2015  biilmann:Joined the channel
Tue Nov 17 23:22:17 2015  Uri:https://www.ssllabs.com/ssltest/viewClient.html?name=Apple%20ATS&version=9&platform=iOS%209
Tue Nov 17 23:22:35 2015  Uri:I assume thats really ATS :-)
Tue Nov 17 23:23:02 2015  zwoop:that's not ATS :)
Tue Nov 17 23:23:15 2015  zwoop:ATS means something else in this context I'm pretty sure
Tue Nov 17 23:24:42 2015  Uri:'k - out of curiosity, any idea what it means?
Tue Nov 17 23:25:07 2015  jpeach:https://developer.apple.com/library/ios/documentation/General/Reference/InfoPlistKeyReference/Articles/CocoaKeys.html
Tue Nov 17 23:25:29 2015  jpeach:scroll down to "Requirements for Connecting Using ATS"
Tue Nov 17 23:25:46 2015  jpeach:ATS == App Transport Security
Tue Nov 17 23:26:00 2015  jpeach:it is also not Apple Type Services :)
Tue Nov 17 23:27:00 2015  Uri:They are usurping our acronym!
Tue Nov 17 23:27:09 2015  Uri:cries foul!
Tue Nov 17 23:27:21 2015  jpeach:seconded!
Tue Nov 17 23:28:12 2015  zwoop:+1
Tue Nov 17 23:28:31 2015  Uri:Look - It's even mentioned in the next slide :-)
Tue Nov 17 23:31:17 2015  zwoop: :)
Tue Nov 17 23:31:45 2015  sudheerv:I got confused to when they used ATS :)
Tue Nov 17 23:32:00 2015  zwoop:yeah, how dare they
Tue Nov 17 23:32:11 2015  sudheerv:s/to/too/
Tue Nov 17 23:32:13 2015  sudheerv::)
Tue Nov 17 23:32:16 2015  zwoop:There's a mechanic in Denver, "ATS Diesel", I should go after them
Tue Nov 17 23:32:31 2015  sudheerv:in Denver of all places!
Tue Nov 17 23:32:35 2015  zwoop:http://www.atsdiesel.com/ats2/
Tue Nov 17 23:32:38 2015  zwoop:maybe it's worldwide
Tue Nov 17 23:32:47 2015  sudheerv:it;s like challenging you, zwoop :)
Tue Nov 17 23:32:53 2015  zwoop:exactly
Tue Nov 17 23:33:28 2015  sudheerv:why on earth are they calling themselves ATS :)
Tue Nov 17 23:33:36 2015  sudheerv:the website doesn't give any reasons
Tue Nov 17 23:33:42 2015  sudheerv:or full form
Tue Nov 17 23:38:08 2015  Uri:Encrypted SNI would hurt a lot of MitMs that need it to generate the fake cert
Tue Nov 17 23:38:27 2015  jpeach:also a lot of people that just do transpory layer optimization
Tue Nov 17 23:39:36 2015  Uri:Well - considering the TLSv1.2 adoption rate TLSv1.3 probably won't be out for another decade ;-)
Tue Nov 17 23:39:44 2015  biilmann:Joined the channel
Tue Nov 17 23:41:45 2015  zwoop:uri seems about right
Tue Nov 17 23:45:03 2015  sudheerv:Uri: why are you not here :-/?
Tue Nov 17 23:45:14 2015  Uri:DockerCon EU
Tue Nov 17 23:45:20 2015  sudheerv:ah, ok
Tue Nov 17 23:45:21 2015  [Jok]:shinrich1: TCP Fast Open is supported in iOS9/OS X 10.11 would you consider it? ;-)
Tue Nov 17 23:46:04 2015  Uri:And besides, I'm not really doing a lot of ATS work these days -- mostly fiddling with containers/SaaS providers/Network Overlays....
Tue Nov 17 23:46:09 2015  jpeach:TS-2520
Tue Nov 17 23:46:13 2015  sudheerv:cool
Tue Nov 17 23:46:16 2015  jpeach:[Jok]: ^^^
Tue Nov 17 23:46:25 2015  [Jok]:thx jpeach
Tue Nov 17 23:46:30 2015  zwoop:needs a very modern version of the kernel to work though (beacuse the option # changed)
Tue Nov 17 23:46:37 2015  zwoop:like, Linux v4.x
Tue Nov 17 23:46:44 2015  Uri:Talk about DWAVE
Tue Nov 17 23:46:49 2015  jpeach:zwoop: it changes?
Tue Nov 17 23:47:05 2015  zwoop:yeah, it was experimental for a long time, and finally got an IANA number
Tue Nov 17 23:47:08 2015  zwoop:iOS / OSX uses the IANA number
Tue Nov 17 23:48:49 2015  dstates:Joined the channel
Tue Nov 17 23:49:02 2015  dstates:sup
Tue Nov 17 23:50:01 2015  Uri:claps
Tue Nov 17 23:51:26 2015  kang1:Joined the channel
Tue Nov 17 23:52:03 2015  Uri:Well -- There are claims that what they do isn't really quantum computing -- but considering their investors/customers it can't be complete @#!#$
Tue Nov 17 23:52:43 2015  [Jok]:TCP Fast Open is HTTP (not https) right? (it'd require that openssl to support it for having that benefit if I understand correctly)
Tue Nov 17 23:52:55 2015  jpeach:right
Tue Nov 17 23:54:52 2015  Uri:Can someone ask regarding userbase of the alternatives?
Tue Nov 17 23:55:17 2015  Uri:I know that BoringSSL has Google behind it - but what about the others?
Tue Nov 17 23:57:42 2015  PSUdaemon:libressl is openbsd

Comments