22:00:24 #startmeeting 2018-02-07 discussion 22:00:24 Meeting started Wed Feb 7 22:00:24 2018 UTC. The chair is wking. Information about MeetBot at http://wiki.debian.org/MeetBot. 22:00:24 Useful Commands: #action #agreed #help #info #idea #link #topic. 22:00:24 The meeting name has been set to '2018_02_07_discussion' 22:00:29 #chair mrunalp 22:00:29 Current chairs: mrunalp wking 22:02:19 #link https://github.com/opencontainers/runc/issues/1709 22:02:31 mrunalp: we can continue to discuss runc 1.0 asynchronously 22:02:50 mikebrow: I think we need another rc before the release 22:03:08 mrunalp: yeah, that makes sense. Especially with the delay since the last rc 22:03:21 mikebrow: and announce that this rc is likely to be the last rc before 1.0 22:03:24 mrunalp: yup 22:03:35 mrunalp: that's all I had 22:04:51 cracra (earlier): I'll likely reroll https://github.com/opencontainers/tob/pull/35 in the next week or so to address some of the open comments 22:06:32 Is there a new meeting invite with the conference number? 22:06:47 [caniszczyk, Open Container Initiative] tel:415-968-0849 22:07:31 * stevvooe says thanks 22:08:03 vbatts: there are a number of comments on the open distribution spec proposal. Last I looked it was pretty much synced up with my expectations 22:08:34 cracra: there have been a lot of comments on https://github.com/opencontainers/tob/pull/35, but I haven't gone through that yet 22:09:06 stevvooe: the specification is open to extention, but we want to bring things in without working on it, and can evolve the scope afterwards 22:09:32 stevvooe: we already had a public discussion on this specification 22:09:51 vbatts: and a lot of us were already involved in this discussion 22:10:31 cracra: my goal is not to do a formal vote until the new TOB is established, and that will take another week or so 22:10:50 mikebrow: I don't think there were big issues with the initial proposal. 22:11:04 mikebrow: the discussion has been about how specific to be with the scope 22:11:40 stevvooe: there are some PRs against the spec in distribution, and we can merge those pre-transfer, but we want those additions to land at some point (they're bandwidth limited) 22:12:08 vbatts: I haven't reviewed any PRs against the distribution repo except the OCI manifest one 22:12:27 stevvooe: the OCI manifest one doesn't touch the spec. I think the biggest change we'll want in the spec is decoupling the media types 22:13:18 stevvooe: the two PRs referenced from the current proposal were blocked by implementation issues, but those issues have been resolved by subsequent implementation changes 22:14:36 dmcg: some of the endpoints may be registry-specific. We need different handling for manifests and blobs 22:14:43 (something about linking, which I missed) 22:15:04 vbatts: things like mounting a blob that you don't have access too 22:15:23 stevvooe: I don't think these details need to be worked out in the proposal. The proposal just needs to allow for them later 22:15:48 vbatts: the scope table may be useful for guiding future extentions. But I'm in favor of making it narrow for a lateral move 22:16:13 stevvooe: we don't allow underscores in names, partly due to the backend. We might use characters to denote particular extentions 22:16:26 stevvooe: we don't allow underscores at the beginning of the name components 22:16:48 stevvooe: there's a concept of an afixing verb. Like tags and manifests. That would be an area for extention 22:17:06 stevvooe: we can do this in a safe way that we weren't as comfortable with early on 22:17:27 stevvooe: but I want to keep the scope of the proposal fairly narrow and let the maintainers work without introducing a bunch of new stuff 22:17:47 mrunalp: I'm stepping out now, and handing note-taking off to you 22:17:54 wking: okay 22:17:58 thanks! 22:18:27 vbatts|work: Any other topics? 22:19:10 stevvooe: image tools project release process 22:20:35 dmcg: What is the scope of the project? 22:20:54 stevvooe: It is challenging as goals of the PR are not clear 22:21:01 PRs 22:21:57 stevvooe: validation of bundle and output 22:23:01 vbatts|work: bundle conversion, bundle creating, assemble layers 22:23:09 vbatts|work: tools around basic image functions 22:23:26 vbatts|work: with one oci-image-tool 22:23:51 cracra: Use for certification to validate a bundle 22:23:55 stevvooe: We just clarify these goals 22:24:05 stevvooe: It seems premature to jump from 0.4 to 1.0 22:24:15 stevvooe: 0.5 would be more acceptable 22:24:47 stevvooe: used dockers archive package in containerd as image tools project wasn't ready 22:26:35 stevvooe: Question on a PR wasn't answered and was self-merged and used to satisfy some requirements for release. 22:26:55 cracra: 1.0 should require more discussion 22:27:19 dmcg: Is there roadmap/acceptance criteria for 1.0 22:27:27 stevvooe: There is one but w/o much feedback 22:27:52 stevvooe: Should clarify whether it is a CLI or a go pkg 22:28:25 stevvooe: Having a better workflow defined would help 22:30:55 cracra: Any other topics? 22:31:03 vbatts|work: Anyone from Intel, Hyper or Kata? 22:31:25 vbatts|work: VMs for OCI should be on this year 22:32:22 vbatts|work: What would be needed to support most generic VMs 22:32:37 david-lyle: Encouraging hyper to comment on the issue. 22:32:45 david-lyle: will track that down 22:34:57 vbatts|work: Could start fresh PR referencing that original PR 22:36:10 stevvooe: What are the missing pieces? 22:36:21 david-lyle: being able to specify kernel for the VM 22:36:28 stevvooe: PR that specifies the kernel path? 22:36:33 david-lyle: yes 22:36:49 vbatts|work: Some configs optional when running in a VM 22:37:38 mrunalp: Example of namespace specified to be host 22:37:54 david-lyle: will talk with sameo to figure out where we are 22:38:38 #link https://github.com/opencontainers/runtime-spec/pull/405/files 22:39:31 stevvooe: being able to specify kernel+guest image per container. As described on PR #405 22:40:41 stevvooe, mrunalp: I can give a kata update to the next OCI meeting to discuss this, if needed. 22:40:58 sameo: cool, sounds good! 22:41:57 #endmeeting