21:03:33 <mrunalp> #startmeeting OCI 31/8
21:03:33 <collabot> Meeting started Wed Aug 31 21:03:33 2016 UTC.  The chair is mrunalp. Information about MeetBot at http://wiki.debian.org/MeetBot.
21:03:33 <collabot> Useful Commands: #action #agreed #help #info #idea #link #topic.
21:03:33 <collabot> The meeting name has been set to 'oci_31_8'
21:03:37 <wking> mrunalp: chair me?
21:03:37 <mrunalp> #topic Updates on console/CLI API, if any.
21:03:42 <mrunalp> #chair wking
21:03:42 <collabot> Current chairs: mrunalp wking
21:03:44 <wking> thanks
21:04:28 <cyphar> I'm here
21:04:29 <cyphar> joining
21:04:48 <mrunalp> https://github.com/opencontainers/runtime-spec/pull/543
21:05:06 <wking> #topic SHOULD for reverse-DNS names
21:05:19 * vbatts here now
21:06:10 <wking> duglin: We have SHOULD language, why not use it?
21:06:18 <wking> duglin: lets change the SHOULD to a MUST.  What happens?
21:09:10 <mrunalp> wking, Should not impose on external spec authors
21:09:21 <mrunalp> duglin, This is an encouragement
21:09:30 <mrunalp> vbatts concurs
21:09:36 <RobDolinMS> Which Issue # or PR # is specifically addressing the reverse-DNS and SHOULD question ?
21:10:05 <duglin> https://github.com/opencontainers/runtime-spec/pull/510
21:10:29 <duglin> I need to add a line about reverse DNS
21:16:23 <stevvooe> " SHOULD   This word, or the adjective "RECOMMENDED", mean that there
21:16:23 <stevvooe> may exist valid reasons in particular circumstances to ignore a
21:16:23 <stevvooe> particular item, but the full implications must be understood and
21:16:23 <stevvooe> carefully weighed before choosing a different course.
21:16:23 <stevvooe> "
21:16:33 <stevvooe> https://www.ietf.org/rfc/rfc2119.txt
21:16:54 <wking> wking: Less "should not impose on external spec authors" than "why bother with a SHOULD we never intend to test?"
21:17:05 <wking> #topic image-spec 0.5.0
21:17:26 <wking> #link https://github.com/opencontainers/image-spec/issues?q=is%3Aopen+is%3Aissue+milestone%3Av0.5.0
21:18:01 <philips> I think we can get this one merged with some LGTM's from maintainers: https://github.com/opencontainers/image-spec/pull/226
21:18:14 <wking> vbatts: what's the timeline for 0.5.0?
21:18:21 <duglin> is vbatts cutting out for others too?
21:18:31 <wking> stevvooe: my big question is refs vs. tags, but philips has clarified that for me
21:18:48 <tophj> duglin: yes
21:18:49 <wking> philips: in that PR we just need another LGTM
21:19:01 <duglin> ok wasn’t sure if it was just my line
21:19:27 <vbatts> duglin: sorry. I'm thinking my internet at the house is silly.
21:20:22 <duglin> with today’s “great” technology you never know where the issues are.
21:21:29 <wking> we might want to move this hostname restriction to a runtime-spec SHOULD (I'll do that)
21:21:41 <duglin> not to rehash things but do we reserve org.opencontainers.* or org.opencontainers/*  or both?
21:22:19 <wking> duglin: I think folks liked the . form.  Folks thought / was weird with reverse-DNS (I'll dig up a link for you after the meeting)
21:22:42 <wking> stevvooe: what about different types of refs?  refs/tags, refs/names, ...?
21:23:04 <wking> philips: that's fine, or we could create a new directory called images/ or whatever
21:23:17 <wking> stevvooe: if we did that, we'd rename refs/ to tags/ and be done with it
21:23:24 <wking> vbatts: are we loosing the word refs?
21:23:36 <duglin> ‘cause I see kube using things like:   pod.alpha.kubernetes.io/init-containers
21:23:36 <wking> stevvooe: yes, unless you see a problem
21:23:57 <wking> vbatts: not a big problem.  But I kind of like the flexibility of generic refs, with tags just being one type of ref
21:24:27 <wking> stevvooe: the Docker save format is a little wild-west at the top level, so I sympathize with that argument
21:24:48 <wking> stevvooe: but we can merge the tag note
21:25:01 <wking> vbatts: I like keeping things under refs/
21:25:21 <wking> philips: I think that closes #173, now that the language is clear
21:25:41 <wking> philips: stevvooe, do you need help on #208?
21:26:11 <wking> stevvooe: I was waiting to see if folks had a good reason for hash sharding
21:26:17 <wking> philips: I didn't see any reasons like that
21:26:24 <wking> stevvooe: can someone else do the code for #208?
21:26:41 <wking> philips: runcom?  Can you do the code?
21:26:44 <wking> runcom|afk: yes
21:27:50 <wking> vbatts: should parsers fail if they find duplicate entries in a tar stream?
21:28:04 <wking> vbatts: shuold you keep a set of all entries you'd seen, so you can detect duplicates?
21:28:21 <wking> philips: I think this is pretty much a MUST NOT
21:28:43 <wking> vbatts: if you have a duplicate entry, you get undefined behavior, and shame on you
21:28:57 <wking> stevvooe: I was mostly concerned about memory if you keep a set of all entries
21:29:32 <wking> vbatts: I don't think we need to require that, just make it undefined
21:29:57 <wking> Do implementers have to check this?  "MUST raise an error" vs. "handling is undefined"
21:30:36 <wking> vbatts: "handling is undefined", so it's up to the implementer
21:30:47 <wking> stevvooe: O(N) for entries is mostly small
21:31:40 <wking> vbatts: I'll put together a vote for 0.5.0 tagging once we get these out
21:31:50 <wking> stevvooe: were there major objetions to #212?
21:31:55 <stevvooe> https://github.com/opencontainers/image-spec/pull/212
21:31:57 <wking> philips: no major objections
21:32:05 <wking> stevvooe: this is mostly cleanup
21:32:09 <wking> philips: I think it's ok
21:32:46 <wking> #topic next image-spec milestone
21:33:03 <wking> philips: move #212 into 0.5.0?
21:33:09 <wking> stevvooe: yeah, I can fix that today
21:33:22 <wking> philips: what about extensions?
21:33:28 <wking> duglin: I should be able to get that done soon
21:33:54 <wking> philips: I'll take a swing at #220
21:34:08 <wking> philips: we need to figure out what else has to happen before 1.0 stuff starts rolling
21:34:14 <wking> philips: other discussion points?
21:34:24 <wking> #topic separate tool repos
21:34:36 <wking> #link https://github.com/opencontainers/tob/pull/18
21:34:43 <wking> mrunalp: RobDolinMS, do you want to talk about this?
21:34:47 <philips> https://github.com/opencontainers/tob/pull/18
21:35:15 <wking> mrunalp: I don't know what happened at the face-to-face, but this PR looks good to me
21:35:35 <wking> philips: motivation is just to decouple tooling from specs so they can iterate independently
21:35:58 <RobDolinMS_> I'm on the line and can hear folks, but sounds lik you guys can't hear my audio
21:36:52 <RobDolinMS_> The PR currently lists the proposed repos as: runtime-tools and image-tools
21:36:52 <crosbymichael> hit the magic buttons to unmute
21:37:01 <crosbymichael> i forget them
21:37:31 <wking> https://github.com/opencontainers/tob/pull/18/files#r76834447
21:37:39 <wking> philips: no strong opinion, but we should pick one
21:37:52 <wking> mrunalp: so we'll wait for the TOB
21:38:01 <RobDolinMS_> Someone (Antonio?) had asked for them to be plural
21:38:13 <wking> #endmeeting