21:58:33 #startmeeting 2019-01-30 discussion 21:58:33 Meeting started Wed Jan 30 21:58:33 2019 UTC. The chair is vbatts|work. Information about MeetBot at http://wiki.debian.org/MeetBot. 21:58:33 Useful Commands: #action #agreed #help #info #idea #link #topic. 21:58:33 The meeting name has been set to '2019_01_30_discussion' 21:59:34 @vbatts hello! 21:59:34 RenaudWasTaken: Error: "vbatts" is not a valid command. 22:00:06 vbatts: looks like he isn't on IRC, should I shoot him an email? 22:02:46 RenaudWasTaken: sure thing. he shows up on IRC from time to time 22:04:40 #topic distribution-spec RC 22:04:43 #link https://github.com/opencontainers/distribution-spec/milestone/2 22:08:15 date was fairly arbitrary, suggested by cra 22:09:19 OT: Is the "OCI Image Spec Weekly Call" deprecated in favor of this "OCI Dev Call"? 22:12:22 zendril: oh, yeah. it's basically always been whole of OCI dev call 22:13:04 @vbatts|work: here's the comment from stevvooe: https://github.com/opencontainers/distribution-spec/issues/26#issuecomment-454992064 22:13:05 estesp: Error: "vbatts|work:" is not a valid command. 22:14:36 on the UUID, the language should say something like MAY or SHOULD be UUID like rfc (but is still just a `string`) 22:15:06 on the /v2/_ping endpoint, it was largely ust a legacy header for detecting version 22:15:17 but for OCI distribution, it ought not be required 22:18:11 "magic" headers ought be needed. Particularly for static site use cases. 22:19:45 if anything the payload of a file like _ping or version should have structured version (potentially even feature flag listing) 22:20:29 https://github.com/opencontainers/distribution-spec/issues/26 22:21:33 #link https://github.com/opencontainers/distribution-spec/issues/26 22:26:07 ruh roh. Accepts header must be set for ManifestList in order to get an actual list, otherwise you only get the manifest 22:27:26 #link https://github.com/docker/distribution/issues/2395 22:29:03 if the name:tag referenced is a manifestlist, then it should always return _that_ object 22:34:24 #topic jimmi Z strawman for arbitrary objects 22:34:29 #link https://groups.google.com/a/opencontainers.org/d/msg/dev/oguystbnnw4/mKJK4IzSDAAJ 22:35:07 jimmi: crone's proposal is orthogonal. OCI ought to be flexible enough to allow it. 22:35:56 all these mime types, like a Helm or CNAB tarball 22:36:05 Current Helm manifest: https://gist.github.com/jdolitsky/99c7364a892c8e350c85314705a9a7c1#chart-manifest 22:36:05 (mattfarina ^^) 22:36:14 #link https://gist.github.com/jdolitsky/99c7364a892c8e350c85314705a9a7c1#chart-manifest 22:41:08 in any case I think it makes sense for Helm and CNAB to align on manifest layout 22:41:19 jdolitsky: agreed 22:42:47 et 22:45:09 ohman. search and mimetypes 22:46:26 i reckon we're kinda on search API topic now 22:46:32 #topic search API 22:46:52 the registry ought to be dumb to all the client permutations 22:47:18 but clients ought to collate the mimetypes and such, based on what the client supports 22:47:34 manifest layout does not matter so much from the perspective of a tool downloading its own content. I think search / common search API is actually why this is important 22:47:50 fair 22:48:10 mimetypes, but also the annotations on the name:tag 22:48:26 dang, possibly even annotations on the descriptor even 22:48:43 there ought to be a hierarchy of relevance 22:49:23 -> manifest annotation -> descriptor mediaType -> descriptor annotation 22:53:37 https://github.com/mattfarina/image-spec/commit/4012edfeaa6114d84875b5d680257debb1f07a90 22:53:59 #link https://github.com/mattfarina/image-spec/commit/4012edfeaa6114d84875b5d680257debb1f07a90 22:56:30 how to have structured annotations, and the opaque string formating 22:57:02 #link https://github.com/opencontainers/image-spec/blob/master/annotations.md 22:58:28 org.opencontainers.image.description Human-readable description of the software packaged in the image (string) 23:02:12 #action vbatts set up call for next wednesday (Feb 6th) 23:04:52 mikebrow: I really hope you're the last to leave on purpose just to get more of that sweet sweet hold music 23:05:07 you know it 23:05:07 #action err'body while this call is fresh, please reply on the email for today's agenda the topic you have that didn't get addressed 23:05:33 we'll try to group them. For either handling on the mailing list or a github issue/PR 23:05:40 #endmeeting