21:00:39 <vbatts> #startmeeting 2017-03-22 discussion
21:00:39 <collabot`> Meeting started Wed Mar 22 21:00:39 2017 UTC.  The chair is vbatts. Information about MeetBot at http://wiki.debian.org/MeetBot.
21:00:39 <collabot`> Useful Commands: #action #agreed #help #info #idea #link #topic.
21:00:39 <collabot`> The meeting name has been set to '2017_03_22_discussion'
21:01:17 <wking> vbatts: chair me?  Not sure how much I'll be able to type, but you never know ;)
21:01:39 <vbatts> no
21:01:41 <vbatts> ;-)
21:01:48 <vbatts> #chair wking mrunalp
21:01:48 <collabot`> Current chairs: mrunalp vbatts wking
21:02:05 <wking> #topic times for future meetings
21:02:08 <mrunalp> crosbymichael, stevvooe: Joining?
21:02:19 <stevvooe> mrunalp: i'll be there shortly
21:02:38 <wking> #topic runtime-spec 1.0 burn-down
21:02:50 <wking> vbatts: are we waiting for Docker and cri-o before a 30-day quiet period?
21:03:05 <wking> mrunalp: I don't think there's an official need for a 30-day quiet period
21:03:31 <wking> ^ I agree that there's no official need for 30 days of quiet before 1.0, but I still think it's a good idea
21:05:32 <wking> If nothing has changed since rc5, I agree that there's no point in cutting an rc6 (although I'm not sure what's landed since rc5)
21:05:46 <wking> crosbymichael: I don't think rc6 is very critical
21:05:56 <wking> crosbymichael: we're working on getting Docker up to date now
21:06:18 <wking> #link https://github.com/opencontainers/runtime-spec/issues/726
21:06:41 <wking> vbatts: we could vote for 1.0 and have 30 days of quiet afterwards
21:07:19 <wking> mrunalp: I don't think we want 30 days of quiet after 1.0.  It should be before
21:07:24 <vbatts> wking: only 8 PRs merged since rc5
21:08:10 <wking> crosbymichael: we're not clear on the 30-day, Member-notification period around releases
21:08:30 <wking> vbatts: the charter says the 30 days are after a new version
21:08:40 <wking> #link https://www.opencontainers.org/about/governance section 8.d
21:09:24 <david-lyle> will 1.0 be a tag or a branch?
21:09:54 <wking> crosbymichael: what happens if they reject after a release?
21:09:55 <vbatts> #link https://github.com/opencontainers/runtime-spec/issues/726
21:10:06 <vbatts> (already done)
21:11:17 <wking> RobDolinMS: Once we're happy, we can cut an rc6, send out notices, wait 30 days, and then tag it 1.0 if there were no problems, then 30 days for the charter's lawyer review
21:11:57 <wking> RobDolinMS: I think Chris reached out to Mike Dolan about when the 30 days are, and would rather have them answer that
21:12:12 <wking> RobDolinMS: I'd rather send cracra an email
21:12:16 <wking> crosbymichael: yeah, we can do that
21:12:23 <wking> mrunalp: cri-o will be done by next week
21:12:43 <cracra> [caniszczyk, Open Container Initiative] once you vote on v1.0 final and the vote is approved, the 30 day starts (note this doesn't mean that the release isn't final, it just gives members 30 days due to the IP policy)
21:12:47 <cracra> [caniszczyk, Open Container Initiative] hope that makes sense
21:13:43 <vbatts> cracra: so what does that mean if a member does not accept it?
21:14:19 <wking> cracra: after the release, members have 30 days to decide to leave or not
21:14:45 <wking> cracra: 1.0s are the same as any release.  You can just vote "we're going to tag rc6 (or whatever) as 1.0"
21:15:14 <wking> vbatts: what are the IP issues?
21:15:27 <wking> cracra: it's a patent grant largely orthogonal to releasing
21:15:58 <wking> cracra: it's a strange IP policy, but that's what we agreed to before starting
21:16:12 <wking> cracra: I've been pre-briefing folks, so I don't think anyone will be surprised by a 1.0
21:16:52 <wking> RobDolinMS: For example, some organization has a patent which is covered by the spec, and then that member would be granting a zero-cost license by accepting that spec version
21:17:19 <wking> cracra: I'd rather have the TDC focus on cutting the release, and we'll work behind the scenes to cover the patent thing
21:18:28 <wking> RobDolinMS: Hopefully companies with large patent portfolios will continue to stay in the OCI, and this gives implementers assurances that those patents won't be asserted against the implementations
21:18:49 <wking> ^ this makes me a lot happier :)
21:20:07 <wking> #630 and #720 are the only interesting changes since rc5
21:21:04 <wking> vbatts: the only thing new for 1.0 would be a 1.0.x branch, if we want to do that.
21:21:08 <wking> mrunalp: I think we can defer that
21:21:51 <wking> crosbymichael: hopefully it quiets down after 1.0
21:22:31 <wking> vbatts: is a containerd update part of the milestone?
21:22:36 <wking> crosbymichael: it's what Docker is currently doing
21:22:49 <wking> crosbymichael: we need to port over the new TTY, etc. runc changes
21:23:20 <wking> vbatts: it would be good to get a vote out within a week or less
21:24:22 <wking> stevvooe: are we not following the milestone?
21:24:29 <wking> vbatts: right now the milestone is just #726
21:24:36 <wking> stevvooe: rc6 has seven issues
21:24:42 <wking> vbatts: we're not to image-spec
21:25:58 <stevvooe> https://github.com/opencontainers/image-spec/issues/588
21:26:02 <wking> #topic image-spec
21:26:07 <wking> https://github.com/opencontainers/image-spec/issues/588
21:26:30 <wking> stevvooe: I've never had problems with this, and don't know what we can put in the spec to cement these issues
21:26:40 <wking> stevvooe: I'm not sure where to take this, and look forward having cyphar on the call
21:27:15 <wking> RobDolinMS: The spec should define what it is, not how to implement it
21:27:42 <wking> stevvooe: the request is for an implementer's note.  I think the data structures are obvious, but that could be my familiarity
21:27:53 <wking> stevvooe: the data structures define the algorithms, not the other way around
21:28:03 <wking> vbatts: there's nothing about algorithms in the current note
21:28:20 <wking> stevvooe: right.  That's what this issue is asking for.  Do we need to encode that in the spec
21:28:47 <wking> vbatts: an encoutered media type that is unknown should be safely ignored...
21:29:04 <wking> stevvooe: I think the issues are "how to handle names that aren't unique?" and platform selection
21:29:11 <wking> vbatts: there's been lots of discussion about that
21:29:32 <wking> stevvooe: I think the conclusions are a little hyperbolic, but the core is a concern that some things may be underspecified
21:29:57 <wking> crosbymichael: the more you start dictating implementations, the more it hurts implementers
21:30:08 <wking> #link https://github.com/opencontainers/image-spec/issues/622
21:30:16 <wking> ^ confusion about what 'features' means
21:31:04 <wking> stevvooe: I think we'll miss wide if we try to explain the datastructure math, and could be inconsistent with what the data structure actually is
21:31:16 <wking> stevvooe: effectively what crosbymichael said
21:33:38 <wking> stevvooe: maybe merging the spec would help, so it's harder to selectively read the spec
21:33:51 <wking> stevvooe: e.g. read one section but miss something important in another section
21:34:40 <wking> stevvooe: so I'll keep pointing at code in containerd about how to deal with this.  There aren't many ways to deal with this
21:35:40 <wking> #link https://github.com/opencontainers/image-spec/pull/620
21:35:46 <vbatts> https://github.com/opencontainers/image-spec/issues/618
21:35:50 <wking> stevvooe: do we want to make that change in the spec too, but just in Go?
21:35:55 <wking> * but -> or
21:35:58 <stevvooe> https://github.com/opencontainers/image-spec/pull/620
21:36:51 <wking> crosbymichael: the split didn't make much sense to me when I read the spec
21:36:57 <wking> stevvooe: yeah, and cyphar ran into this as well
21:37:27 <wking> stevvooe: for example, if your visitor is looking at an index, but is calling a handler that doesn't care if the descriptors are ManifestDescriptors or Descriptors
21:38:20 <wking> stevvooe: currently as the spec reads it would just be ignored (when parsing JSON)
21:38:25 <wking> vbatts: like JSON Schema?
21:38:30 <wking> stevvooe: yeah.  Do we need to do that?
21:39:26 <wking> vbatts: if someone is using a descriptor outside the manifest, where we don't carry additional fields, having 'platform' doesn't make it invalid
21:39:36 <wking> vbatts: so maybe no, because it's testing an assumed-broken use case
21:39:50 <wking> stevvooe: so we can merge #620, and we can adjust the spec later if we need
21:40:39 <wking> #topic meeting times
21:40:49 <wking> RobDolinMS: 8am Pacific and 5pm Pacific
21:40:59 <wking> RobDolinMS: There's a proposal to alternate even-odd
21:41:22 <wking> vbatts: start that next week?  But next week is KubeCon
21:41:37 <wking> vbatts: 8am is terrible?
21:41:46 <wking> RobDolinMS: 8am is dropping off kids and walking the dog ;)
21:42:02 <wking> #link https://github.com/opencontainers/runtime-spec/pull/719
21:42:09 <wking> crosbymichael: hopefully post 1.0 these calls are less critical
21:42:21 <tianon> crosbymichael: "same"
21:42:22 <wking> crosbymichael: I think they're still useful for conflict resolution
21:42:26 <tianon> xD
21:43:19 <wking> RobDolinMS: So starting next week?
21:44:40 <wking> RobDolinMS: Wed 3-29 lots of people will be in Europe, so favor Europe then
21:44:47 <wking> I'll update my PR to do that^
21:45:06 <wking> #topic 1.0 releases
21:45:19 <wking> vbatts: for the runtime, we're mostly through the final burn-down
21:45:56 <wking> vbatts: we want to see more testing.  Docker and cri-o, although containerd, etc. would be nice
21:46:04 <wking> vbatts: we want to see good test coverage there next week
21:46:13 <wking> vbatts: then put up a 1.0 vote
21:46:50 <wking> vbatts: the vote will happen like normal, with a tagged release, and then the 30-day stuff happens out of band
21:47:13 <wking> vbatts: for the image spec, there's an rc6 milestone to burn down, and then it will go through the same 1.0 proces
21:47:25 <wking> vbatts: so a bit behind the runtime spec, but the same 1.0 process
21:47:36 <wking> vbatts: best case scenario is a runtime-spec 1.0 in two weeks
21:47:49 <wking> vbatts: I don't know what an ideal time would be for image-spec
21:48:01 <wking> vbatts: closing it out before April 15th woudl be a beautiful thing
21:48:46 <wking> tianon: I'll take a stab at pushing an ICS to match a README
21:49:26 <wking> that's for minutes?  I can do that
21:49:56 <tianon> <3
21:49:58 <wking> vbatts: other issues?
21:50:05 <wking> #topic image-spec issues
21:50:13 <wking> stevvooe: I just closed #493
21:50:19 <wking> stevvooe: let me know if you disagree
21:50:32 <wking> stevvooe: the charset stuff in #599 needs help, if anyone has time.
21:50:37 <stevvooe> https://github.com/opencontainers/image-spec/issues/599
21:50:42 <wking> stevvooe: do we have a help-wanted label?
21:50:49 <wking> vbatts: can you manage labels without write access?
21:50:55 <wking> no
21:51:04 <wking> stevvooe: help-wanted exists
21:51:47 <stevvooe> https://github.com/opencontainers/image-spec/issues/599
21:52:10 <tianon> https://twitter.com/DockerCuties/status/839614192436948992
21:52:54 <stevvooe> https://github.com/opencontainers/image-spec/issues/415
21:53:03 <wking> stevvooe: we still need Microsoft validation here^
21:53:15 <vbatts> tianon: <3
21:53:16 <wking> RobDolinMS: is that the one with some email last week?
21:53:19 <wking> stevvooe: yeah
21:53:28 <vbatts> vbatts|cutie
21:53:32 <tianon> ++
21:53:33 <wking> RobDolinMS: I'll forward that to some Windows folks today
21:53:45 <wking> stevvooe: I think we can take #400 of the 1.0 milestone
21:53:48 <wking> vbatts: so post-1.0?
21:53:55 <wking> stevvooe: I think we have enough to go on
21:54:05 <wking> stevvooe: you can sign a hash, and that's enough for now
21:55:01 <wking> RobDolinMS: You can assign #415 to me, but make a note that I'm asking for colleague import
21:55:04 <wking> * input
21:55:26 <stevvooe> https://github.com/opencontainers/image-spec/pull/492
21:55:28 <wking> stevvooe: do we merge #492?  That's the only one I can see taking extra time
21:55:59 <wking> stevvooe: I think cyphar needs another pass on this
21:56:12 <wking> stevvooe: and I was going to have some of our people take another peek at it
21:57:03 <wking> #endmeeting