14:00:59 <nikhil_k> #startmeeting Glance Artifacts Sub-Team Meeting
14:01:00 <openstack> Meeting started Mon Jun 22 14:00:59 2015 UTC and is due to finish in 60 minutes.  The chair is nikhil_k. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:01:01 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:01:03 <openstack> The meeting name has been set to 'glance_artifacts_sub_team_meeting'
14:01:08 <nikhil_k> #chair ativelkov
14:01:09 <openstack> Current chairs: ativelkov nikhil_k
14:01:43 <ativelkov> hi :)
14:01:45 <kragniz> o/
14:01:48 <nikhil_k> #link https://etherpad.openstack.org/p/glance-artifacts-sub-team-meeting-agenda
14:01:49 <kragniz> thought I'd drop by
14:01:55 <ativelkov> welcome :)
14:01:56 <dshakhray> o/
14:02:01 <mfedosin> o/
14:02:01 <nikhil_k> all are welcome :)
14:02:23 <nikhil_k> It's a short (30 min) meeting for now
14:02:30 <nikhil_k> So, let's get started.
14:02:35 <nikhil_k> #topic General Updates
14:02:55 <ativelkov> So, this is a quick update on what we are currently doing.
14:03:31 <mfedosin> shifting to oslo.versionedobjects (in brief)
14:03:36 <nikhil_k> ativelkov: do you have any links?
14:03:55 <ativelkov> Based on the summit feedback (thanks to flaper87 for great idea) we've started to move artifact type definitions to oslo.versioned_objects
14:03:56 <nikhil_k> Like a spec, review or irc/ML conversation?
14:04:33 <ativelkov> nikhil_k: not yet. Right now I am working with oslo.versioned_object to add some features which are mandatory for us
14:04:40 <flaper87> ativelkov: w00000000000h0000000000
14:04:53 <flaper87> ativelkov: btw, have you seen my pings w.r.t artifacts?
14:05:01 <mfedosin> ativelkov made a couple of commit there
14:05:02 <flaper87> are there patches that need sanity checks from me?
14:05:08 <flaper87> mfedosin: you too ^
14:05:25 <nikhil_k> ativelkov: cool, please give us those links (later/offline okay too).
14:05:34 <ativelkov> flaper87: nope, didn't see any pings from you. My IRC bouncer is buggy as it turns out
14:05:55 <ativelkov> #link https://review.openstack.org/#/c/193077/
14:06:02 <flaper87> ativelkov: aaah :(
14:06:03 <mfedosin> flaper87, yes, but I want ativelkov to review them first
14:06:14 <ativelkov> #link https://review.openstack.org/#/c/194104/
14:06:27 <mfedosin> links are in agenda
14:06:27 <flaper87> mfedosin: ok, feel free to ping me on IRC when you guys are ready
14:06:43 <mfedosin> flaper87, sure, thank you
14:07:20 <ativelkov> So, whet oslo.versioned_object is currently missing the most are the constraints
14:08:00 <ativelkov> They have type validation and coercing, but no way to limit the length of the string, min/max value of integer, amount of items in the list etc
14:08:34 <nikhil_k> ativelkov: would it make sense to put those there?
14:08:42 <nikhil_k> or are you trying to make them configurable?
14:08:53 <nikhil_k> cool, that coerce bit seems to be doing a bit extra
14:09:08 <ativelkov> Of course this may be done at Glance level (i.e. by inheriting from base classes of oslo), but I want to check if these things are generic enough for reusing them widely
14:09:24 <nikhil_k> extra == (validation ++)
14:09:38 <nikhil_k> gotcha
14:10:12 <nikhil_k> yeah, some fields in glance are/may need to be generic enough
14:10:39 <ativelkov> And then there is a concept of "immutable field" (i.e. a field which is editable while the object is in "draft" state but becomes frozen once it is active) - but this is a naarow concept, we should keep it glance-only. I think
14:11:18 <nikhil_k> make sense
14:11:23 <nikhil_k> makes*
14:11:37 <ativelkov> So, my estimates are to have a working prototype of some oslo-based ArtifactType later this week
14:11:48 <nikhil_k> that's great news
14:12:04 <ativelkov> we may add some extra capabilities later if they are required
14:12:35 <ativelkov> So, any other questions on this topic?
14:12:49 <nikhil_k> ok. should we probably note down a few priorities for this cycle?
14:13:11 <nikhil_k> So, that they get listed right up in the liberty review list?
14:13:13 <sigmavirus24> That's awesome ativelkov
14:14:04 <nikhil_k> I guess, first one what I think is we need to move the API from EXPERIMENTAL to a supported state
14:14:13 <ativelkov> nikhil_k: my top priority is to improve API and reach agreement on it with API WG and all the interested parties
14:14:36 <mfedosin> api wg has many concerns about current state of artifacts api
14:14:43 <nikhil_k> yeah
14:14:50 <ativelkov> so do I :)
14:14:59 <mfedosin> +1 :)
14:15:12 <nikhil_k> should we discuss this around july 10?
14:15:27 <mfedosin> and it's really hard to implement cleint if we don't have stable api
14:15:36 <nikhil_k> with the API_WG or do we need to do that before?
14:15:37 <mfedosin> and it's really hard to implement client if we don't have stable api
14:16:05 <ativelkov> nikhil_k: I'd prefer to have some preliminary draft of the document with open issues and possible ways of making them better
14:16:30 <nikhil_k> ativelkov: sure, I am just trying to get a sense of how long we need to wait
14:16:43 <ativelkov> We've started working on such a document already
14:16:51 <nikhil_k> ok
14:16:56 <ativelkov> Will try to share it this week
14:17:22 <nikhil_k> ativelkov: sure, let's give it a few iterations within the team
14:17:53 <ativelkov> mfedosin: EXPERIMENTAL API is intended to be frequently changed, so its client should be changing as well
14:18:42 <mfedosin> actually yes, but I don't want to change it too often
14:18:52 <ativelkov> But we need to have some (may be in the most simple form) before we settle all the open API questions, so we may try it somehow
14:19:29 <nikhil_k> people are saying that if projects use them, then we need to keep supporting it. So, I think we need to move away from experimental in liberty
14:20:03 <nikhil_k> let me know when you think would be the best time to collaborate on the doc. I may be out for some travel from june 27-july2
14:20:04 <mfedosin> nikhil_k, hard, but possible
14:20:15 <nikhil_k> yeah
14:20:17 <nikhil_k> :)
14:20:42 <nikhil_k> ativelkov: one thing that is concerning atm is we need to mark one images api as supported and keep that in defcore for next decade or so
14:20:51 <ativelkov> nikhil_k: no objections from my side :) But are we going to support 3 majors of API at the same time? Otherwise it will depend on v1 deprecation
14:20:59 <nikhil_k> yeah
14:21:06 <nikhil_k> just coming to that ^ :)
14:21:26 <nikhil_k> basically the push from a few people is that we need to wrap v1, v2 on top of v3
14:21:54 <nikhil_k> that's really chanllenging given so much to do already. but we don't have to do this in liberty
14:22:03 <nikhil_k> I think v2 can be in def core
14:22:43 <nikhil_k> all of this is in flux
14:23:05 <mfedosin> what about domain in v2 btw? I have some thoughts how to improve it
14:23:32 <ativelkov> mfedosin: that's more an implementation detail
14:24:01 <mfedosin> I'll put my ideas in related etherpad then
14:24:02 <ativelkov> we may drop domain at any moment when we are ready too. API changes are much less flexible, so that is much more important
14:24:26 <mfedosin> ativelkov, agree with you
14:24:36 <nikhil_k> can't find a link: but look for email to openstack-defcore with sub: "Image APIs in Glance and Nova"
14:25:20 <nikhil_k> we just have 5 mins remaining.
14:25:41 <nikhil_k> ativelkov: mfedosin : please go ahead with whatever most important thing(s) you wanna discuss now
14:26:02 <ativelkov> Yeah, the question is about the client
14:26:21 <ativelkov> do we want to use v2's approach with jsonschema and warlock on top of it?
14:26:41 * nikhil_k -1 unless other way is way too complicated
14:27:30 <nikhil_k> but we may need to if v2 would be on top of v3 (I think) and we may need to support flexible/configurable schema
14:27:42 <ativelkov> The other way is to make glance-client work with type-less dicts of values for type-specific properties / blobs/ relations, and let the project-specific plugins interprete them
14:28:32 <mfedosin> we were discussing it all morning and have no decision, but I don't like warlock and stuff and want to find a better way to implement dependencies in client
14:29:09 <nikhil_k> I think that makes sense as long as project plugins don't go haywire. are we discussion to keep them in glance or in project repos?
14:30:24 <ativelkov> nikhil_k: in project repos, I guess
14:30:34 <nikhil_k> cool
14:30:35 <ativelkov> Glance will have only Images support as an example
14:30:38 <mfedosin> ativelkov, dicts means a little bit more work for users, but give more flexibility and much less work for us :)
14:30:45 <nikhil_k> ativelkov: I am with you
14:31:08 <ativelkov> We are out of time)
14:31:14 <nikhil_k> yeah
14:31:15 <nikhil_k> :)
14:31:19 <ativelkov> Let's continue in #glance
14:31:29 <nikhil_k> ativelkov: please run the meeting next week. I am out of town/may not be online
14:31:36 <ativelkov> nikhil_k: got it
14:31:42 <nikhil_k> THanks all!
14:31:43 <mfedosin> thanks all :)
14:31:45 <nikhil_k> #endmeeting