17:02:43 #startmeeting weekly discussion 17:02:43 Meeting started Wed Oct 28 17:02:43 2015 UTC. The chair is vbatts. Information about MeetBot at http://wiki.debian.org/MeetBot. 17:02:43 Useful Commands: #action #agreed #help #info #idea #link #topic. 17:02:43 The meeting name has been set to 'weekly_discussion' 17:02:50 #chair RobDolinMS wking 17:02:50 Current chairs: RobDolinMS vbatts wking 17:03:00 Thanks vbatts 17:03:06 is there no sound without a mod? 17:03:17 crosbymichael: correct 17:03:21 ok 17:03:28 we can use https://bluejeans.com/5531707434 17:03:34 I have a 800 number we can use if you want 17:03:42 that's mine, at least for this week. 17:03:57 I'm in https://bluejeans.com/1771332256/; should we switch? 17:04:08 RobDolinMS: yes 17:04:11 for today 17:04:14 #topic Challenge with audio 17:04:32 #info switching to https://bluejeans.com/5531707434 for audio channel 17:04:45 got it 17:04:53 so on the phone I just hear music 17:05:00 duglin: hmmm 17:05:13 oh, never mind here we go 17:05:20 Phone Dial-In 17:05:21 + 800 451 8679 17:05:23 Meeting Id 17:05:25 5531707434 17:05:27 k 17:06:04 Yes. 17:06:26 We seem to have successfully switched audio. 17:06:43 +q 17:06:48 #topic Confirming agenda 17:07:19 #topic Separation 17:07:31 #info This topic came-up during hack day last week 17:07:44 We spent time waling through fields last week 17:08:02 ... and implications of if not present 17:08:34 (Aside: If you're attending the weekly OCI Dev ConCall please type something (i.e. hello) to be recorded for attendance) 17:09:33 vbatts suggestion some fields only used by a runtime (i.e. hooks) 17:10:59 Who is the 919 phone number talking? 17:11:29 discussion about writing a copy of the config to the /run/opencontainers/containers/{id} directory 17:11:43 919: Very few things are immutable 17:12:05 919: many fields appearing in both config.json and runtime.json 17:12:50 if there is a config.json, use it. If there isn't a config.json use fall back to default.json, so folks can use default.json as an immutable copy 17:13:25 919: Is it accurate assumption that files are read-only? 17:13:48 are you guys using the same bluejeans linkfor the meeting? 17:14:11 anuthan: this week we're at https://bluejeans.com/5531707434 17:14:13 @anuthan, we had audio issues so switched to: https://bluejeans.com/5531707434 for audio channel 17:14:13 RobDolinMS: Error: "anuthan," is not a valid command. 17:14:35 Thank you 17:15:46 vbatts: If there is a default.json, some fields should be ignored 17:17:54 personally I have no problem with a single config file (no config.json / runtime.json split, and no config.json / default.json fallback) 17:18:46 and I would have the runtime ignore all environment variables, so there were no environment variables that impacted the launched container 17:20:12 919: Someone hands you a bundle, you explode it out, you should be able to immediately verify hash 17:20:25 919: For how long after should you be able to verify hash? 17:20:31 919: you probably don't need those hash/sigs to still apply after local modification 17:21:06 crosbymichael: As soon as you've taken action(s) will be hard to verify hash. 17:21:11 +1 to not jumping through hoops to get some sort of immutable config 17:23:43 919: rather than thinking mutable vs. immutable, 17:24:03 919: Things you don't port to another host go into runtime 17:24:14 ^ that's what we're doing now with runtime.json and config.json, but there's a Solaris thread asking to get those runtime.json things distributed with the bundle 17:25:44 I can buy into the split if we do it on portable vs non-portable. Then we never have to worry about overriding values between two config file - there’s no overlap. 17:26:03 i've got to duck out now. 17:26:13 RobDolinMS: good luck on policing the topics! 17:26:30 vbatts: thanks :) 17:27:57 crosbymichael: The base bundle should be a portable artifact that can be put on any server and run 17:28:59 919: We have a few proposals 17:29:34 #info duglin volunteers to write-up idea of switch from mutable/immutable to portable/non-portable 17:29:41 #action wking to write up single config approach 17:30:49 #topic Separation of create / start phase 17:31:00 #info suggestion to table this 17:31:24 crosbymichael: This may be more technical than spec and may be platform specific 17:32:10 if you have a distinct create operation, you can drop the pre-start hooks, which makes a simpler spec 17:32:15 #info duglin asks if spec will allow for create as separate operation from start 17:32:29 #info crosbymichael suggests we should look at use cases 17:33:30 #info Suggested next step is for someone to draft proposal for this 17:33:42 #action julz to write-up proposal 17:34:28 #topic More hackdays 17:34:48 #info duglin would like more; or at least another 17:35:24 #info RobDolinMS would like another as well 17:36:02 #action RobDolinMS to kick-off thread about doing another hackday 17:36:04 (sounds good) 17:36:42 LOL is that a dog or a baby? 17:36:54 #topic Required vs. Optiona fields 17:37:06 wking: What's a good way to have that discussion? 17:37:49 Solaris will be willing to participate 17:37:49 RobDolinMS: Suggest ensuring platforms are well represented in discussion 17:39:24 #info we should think on this and table to next week 17:39:52 #topic Roadmap 17:40:11 #info crosbymichael thinks he has pretty good participation so far 17:40:26 (see GitHub discussion on the PR) 17:40:39 woo hoo! 17:40:47 #endmeeting