18:02:04 #startmeeting 2015-12-09 discussion 18:02:04 Meeting started Wed Dec 9 18:02:04 2015 UTC. The chair is vbatts. Information about MeetBot at http://wiki.debian.org/MeetBot. 18:02:04 Useful Commands: #action #agreed #help #info #idea #link #topic. 18:02:04 The meeting name has been set to '2015_12_09_discussion' 18:02:23 #chair mrunalp wking 18:02:23 Current chairs: mrunalp vbatts wking 18:02:38 https://bluejeans.com/1771332256/ for posterity 18:03:59 #topic wking update on the CLI API 18:08:47 Video Attendees: mrunalp, philips, crosbymichael, duglin, julz, liangchenye, mikebrow_, neicy, estesp, RobDolinMS, tianon, wking, vishh 18:09:35 I just summarized https://groups.google.com/a/opencontainers.org/d/msg/dev/zKNoOV1RP-k/Q_aSoPnACQAJ 18:09:49 Feedback from runC and other runtime authors welcome 18:10:24 #action wking to PR opencontainers/specs to land the command-line API in the spec itself 18:10:48 (sorry I'm late, busy with launch of http://OpenLiveWriter.org/ this morning) 18:10:49 ^ will talk with mrunalp and Co. after the meeting to figure out how to orchestrate that 18:11:50 Brandon: trying to capture reasonable subset of operations on a container through CLI may be premature 18:12:16 Brandon: We should start with a harness and if that's successful, we can put into spec 18:12:28 Mrunal: Is this something that will evolve? 18:12:44 #chair RobDolinMS 18:12:44 Current chairs: RobDolinMS mrunalp vbatts wking 18:13:47 Julz: people are doing this, but reverse-engineering from runc 18:16:30 philips: suggesting don't merge it into opencontainers/specs until we have something that works 18:17:11 julz: the value of this goes down as long as its not official 18:17:29 julz: there are several people attempting to do this simultaneously 18:19:08 ... discussion ... 18:19:27 Julz: Better to iterate on one document together 18:19:44 Brandon: with testing of appc, we generate a bundle and test 18:19:57 Brandon: we're having to implement both internal spec and external spec 18:22:10 Trevor: This may not be that big after all. 18:22:40 Mrunal: All of the major implementers can continue collaborating. 18:23:03 Julz: It seems best to have in an official, foundation-owned location rather than Trevor's personal GitHub 18:23:04 I think the conclusion is "don't land it in opencontainers/specs yet, get some buy-in from runtimes, see if it works, and then land it" 18:24:00 Liangchenye: Test case writers can write commands down in test cases so problem can be solved. 18:24:16 liangchenye: the test suite can write an adaptor for runtimes that don't implement a common API 18:24:42 but it's sorta hard for test suite, right? :) and test suite doesn't belong to runtime author 18:24:46 lots of background noise 18:24:50 Yes @wking we already did this 18:25:00 #topic Roadmap updates 18:25:06 @wking, we call it 'factory' 18:25:06 liangchenye: Error: "wking," is not a valid command. 18:25:16 #topic Top-down approach 18:25:21 lk4d4: I agree, better to have the runtime authors write the adaptor. But it's a valid work-around until they do 18:25:30 Vincent: I'm hoping to organize documents we have 18:25:42 @lk4d4, yes, with a offical command line, it will be much easier. 18:25:42 liangchenye: Error: "lk4d4," is not a valid command. 18:25:59 Vincent: Docs about config file might reference others 18:26:10 lk4d4: yes, with offical command line, it will be much easier. 18:26:10 Vincent: 18:26:13 https://github.com/opencontainers/specs/pull/263 18:26:52 #link https://github.com/opencontainers/specs/pull/263 18:28:08 #topic Next milestones 18:28:09 #action wking will mock an alternative using Jekyll imports, https://github.com/opencontainers/specs/pull/263#issuecomment-163341758 18:28:23 Should 0.2.0 be a release? 18:28:47 Vincent: appologies for v0.1.0 --> v0.1.1 issue 18:29:05 Vincent: We have not gone through the repo to start tagging issues for 0.2.0 18:29:20 (wking - sorry for being too quick with the #topic lines) 18:30:13 also, https://vbatts.fedorapeople.org/oci.pdf is an example of the printable pdf .. 18:30:41 wking: Should we have a cadence that is high enough so folks can grab tags rather than commits 18:31:49 way too quiet 18:33:20 vbatts: suggests focusing on a single area per release, to get more movement 18:33:27 oops, vishh: ^ 18:35:01 #topic hashing with metadata 18:35:21 philips: guesses we end up with a just-filesystem hash and a filesystem+metadata hash 18:35:31 philips: leaning towards just using tar 18:35:50 mrunalp: suggests that^ for 0.3.0 18:36:03 mrunalp: suggests landing 0.2.0 now? 18:36:44 #endmeeting