17:01:23 #startmeeting OCI weekly meeting 5/18 17:01:23 Meeting started Wed May 18 17:01:23 2016 UTC. The chair is mrunalp. Information about MeetBot at http://wiki.debian.org/MeetBot. 17:01:23 Useful Commands: #action #agreed #help #info #idea #link #topic. 17:01:23 The meeting name has been set to 'oci_weekly_meeting_5_18' 17:02:15 mrunalp: chair me? Although I'm not sure I'll be fully focused the whole meeting 17:02:30 #chair wking 17:02:30 Current chairs: mrunalp wking 17:02:40 #topic create/start split 17:02:57 mrunalp: we want to bring the spec and runC PR together 17:03:09 julz_: I think duglin is in China this week, but I'm sure we can do that 17:03:31 crosbymichael: I don't think there are any issues as long as everyone's testing it. It works fine in my tests 17:03:38 mrunalp: I've done some testing, but will do some more 17:04:07 mrunalp: this is the only thing blocking 0.6, but RobDolinMS said he also has some stuff in the pipe for 0.6. Should we delay 0.6 another week? 17:05:16 RobDolinMS: I'm not sure how expensive it is to cut a release, but I expect some Windows input by Monday 17:05:31 mrunalp: it will take a few more days to update and review the create/start stuff anyway 17:05:43 crosbymichael: we can also cut 0.6 and wait on Windows for 1.0 17:05:53 mrunalp: either works. We can make a call once create/start lands 17:06:07 #topic https://github.com/opencontainers/ocitools/issues/66 17:06:08 #info We will expect some additional PRs re: Windows Containers early next week 17:06:48 mrunalp: ocitools#66 has a table of runtime-spec vs. test-suite coverage map 17:07:06 crosbymichael: before we do a 1.0 we should review the spec and double check MUST, SHOULD, etc. 17:07:35 crosbymichael: we don't want to have an accidental MUST that negatively impacts a particular use-case 17:07:50 #action Maintainers should take a look at the aforementioned PR and check if MUST/REQUIRED, SHOULD/RECOMMENDED, and MAY/OPTIONAL are accurate 17:08:18 crosbymichael: we could get a small group together after 0.6 and look at requirements line by line 17:08:33 mrunalp: that makes sense. Maybe we can set that up after next week's call 17:08:36 crosbymichael: sounds good 17:08:48 #info Mrunal suggests meeting next week and doing this 17:09:28 https://github.com/opencontainers/runtime-spec/pull/436 17:09:37 #topic remove solaris from full config 17:09:44 #url https://github.com/opencontainers/runtime-spec/pull/436 17:10:21 #topic no mount nesting on Windows 17:10:25 #url https://github.com/opencontainers/runtime-spec/pull/433 17:10:41 #link https://github.com/opencontainers/runtime-spec/pull/433 17:10:43 https://github.com/opencontainers/runtime-spec/pull/432 17:10:51 mrunalp: #433 is good with a squash 17:11:08 #topic fstab-options are Linux-specific 17:11:13 #link https://github.com/opencontainers/runtime-spec/pull/432 17:11:30 mrunalp: in runC we have some additions on top of the fstab format, and I think that should be allowed 17:11:58 RobDolinMS: The Windows folks haven't implemented this yet, so they don't want to get boxed in before they've had time to look at this in detail 17:12:53 mrunalp: we need extensions on Linux anyway 17:13:18 RobDolinMS: #432 just opts-out Windows for now, so I can clean this PR up and we can land it 17:13:24 #action RobDolinMS to update PR #432 17:13:25 https://github.com/opencontainers/runtime-spec/pull/427 17:13:31 #topic optional hooks 17:13:36 #link https://github.com/opencontainers/runtime-spec/pull/427 17:14:39 mrunalp: I agree with crosbymichael that we don't want this to be Go-specific 17:14:47 I'm just carrying over our existing Go-links 17:14:55 What should I link to instead? 17:14:59 wking: everyexecever.tumblr.com 17:15:03 Some options in https://github.com/opencontainers/runtime-spec/pull/427#discussion_r62376119 17:15:06 hah ;) 17:15:24 crosbymichael: we may not need a link, but you can link to execve 17:17:22 mrunalp: even though it may not be cross-platform, execve(3p) is probably practical enough 17:18:38 RobDolinMS: what if we follow #432 and say "in a platform-appropriate way", and then link execve(3p) for POSIX systems 17:19:02 sounds good to me^ 17:19:07 mrunalp: I think that makes sense 17:19:21 #action wking will reroll #427 with that wording 17:19:32 https://github.com/opencontainers/runtime-spec/pull/423 17:20:59 #topic dropping the "bundle" word 17:21:05 #link https://github.com/opencontainers/runtime-spec/pull/423 17:22:34 #link https://github.com/opencontainers/runtime-spec/issues/389 17:24:13 mrunalp: this probably needs more feedback and discussion, punt for now? 17:24:41 crosbymichael: just because the rootfs can be absolute doesn't mean we have to remove the "bundle" term 17:24:46 crosbymichael: it's still the number-one use case 17:25:01 mrunalp: yeah, it seems like a relative root path is the usual use case 17:25:23 But everyone agrees that #389 can be absolute 17:25:48 tianon: do we plan to split the config? 17:26:00 tianon: if we do plan to do that, there is some sense in keeping "bundle" 17:26:18 JakeWarner|Work: I'm also concerned about splitting the config, and the unknowns of removing "bundle" 17:26:33 mrunalp: we had split, and then decided it didn't make sense, and merged them back together 17:26:43 tianon: but we might change our minds again and re-split later 17:27:38 tianon: I think it's reasonably clear that a relative roots is relative to the config path 17:28:44 mrunalp: I agree that having relative paths be relative to the config path makes sense 17:29:11 ^ I was saying that we want to be clear in the spec whether relative paths are relative to the config path or the runtime cwd 17:29:17 #action everyone discuss the PR ;) 17:29:28 https://github.com/opencontainers/runtime-spec/pull/418 17:29:42 #topic allow 'start' to not validate the config.json 17:29:48 #link https://github.com/opencontainers/runtime-spec/pull/418 17:30:34 mrunalp: this sounds ok to me 17:30:42 crosbymichael: sounds fine 17:31:17 https://github.com/opencontainers/runtime-spec/pull/412 17:31:23 #topic explicit container namespaces for uid 17:32:23 mrunalp: we can continue discussing in the PR 17:32:27 https://github.com/opencontainers/runtime-spec/pull/403 17:32:49 #topic capitalize MUST 17:32:54 #link https://github.com/opencontainers/runtime-spec/pull/403 17:33:06 https://github.com/opencontainers/runtime-spec/pull/397/files 17:33:10 #action RobDolinMS to reroll #403 to get wording updates 17:33:22 #topic cgroup namespaces 17:33:28 #link https://github.com/opencontainers/runtime-spec/pull/397 17:33:33 mrunalp: this just needs a wording update 17:33:52 #topic VM configuration updates 17:33:59 #link https://github.com/opencontainers/runtime-spec/pull/405 17:34:07 mrunalp: discuss or punt? 17:35:06 Anush: James updates the path stuff here. Do we need to do anything else? 17:35:14 mrunalp: so it seems like it just need more review 17:35:43 Anush: I don't see any outstanding requests. Either more review, or LGTM and merge 17:36:07 crosbymichael: I think moving the setting out of the top level would be good. Otherwise it looks good 17:37:37 crosbymichael: the top-level concept of a process is compatible across all systems 17:38:32 crosbymichael: in the context of the OCI, kernel parameters are not portable, so they should be in a sub-section 17:41:31 mrunalp: the logic is "if any platform cannot use your setting, namespace it, unless that would seem weird for asthetic reasons we're punting on defining" 17:41:39 wking - can you add an action for the 405 so that we can get to the bottom of it 17:41:41 works for my^ 17:41:59 Thx 17:42:18 #action anush_ to namespace VM-specific settings in #405 17:42:40 #action maintainers to review, LGTM, and/or merge #405 after the reroll 17:42:51 #topic repo organization 17:43:11 RobDolinMS: What goes into what repository? We have ocitools, runtime-spec, image-spec, some image-testing tools? 17:43:34 mrunalp: philips wants to keep the image-related tooling in the image repo (or out of ocitools at any rate) 17:43:46 mrunalp: I'm open to merging ocitools back into the spec if it makes sense 17:44:00 RobDolinMS: we should probably punt until we have philips 17:44:10 mrunalp: yeah, and it's not a big deal either way 17:44:22 RobDolinMS: if we keep using ocitools, should the TOB bless it? 17:44:36 mrunalp: was ocitools-blessing discussed in a TOB meeting? 17:44:49 crosbymichael: it's already official because it has an ocitools/ repo 17:45:05 RobDolinMS: I don't think ocitools has specifically been blessed, but nobody is raising concerns 17:45:16 crosbymichael: the repo predates the TOB 17:46:00 #link https://github.com/opencontainers/tob/issues/2 17:46:27 #link https://github.com/opencontainers/tob/issues/11 17:47:34 mrunalp: maybe we should raise it to the TOB? 17:47:48 RobDolinMS: maybe punt until next week, once philips and vbatts_ are present 17:48:09 RobDolinMS: so we can give the TOB something to vote on, instead of having the TOB sort it out amongst themselves 17:48:15 mrunalp: punting a week sounds good 17:48:27 #endmeeting