20:02:18 #startmeeting tc 20:02:19 Meeting started Tue Feb 4 20:02:18 2014 UTC and is due to finish in 60 minutes. The chair is ttx. Information about MeetBot at http://wiki.debian.org/MeetBot. 20:02:21 Useful Commands: #action #agreed #help #info #idea #link #topic #startvote. 20:02:24 The meeting name has been set to 'tc' 20:02:28 Our agenda for today: 20:02:41 #link https://wiki.openstack.org/wiki/Governance/TechnicalCommittee 20:02:57 #topic Defcore sub-committee questions about Projects vs Programs 20:03:03 zehicle: hi! 20:03:11 hey 20:03:18 zehicle: could you summarize what those questions or concerns are ? 20:03:22 #topic programs 20:03:35 yes 20:03:37 o/ 20:04:11 discussion is about basically about order/membership between programs and projects 20:04:26 and there's a follow-on about how programs are authorized 20:04:37 OK, I think there is confusion around the program concept. 20:04:45 As far as "core" in concerned, the board should really only be looking at projects imho. 20:04:51 My basic definition would be: 20:04:53 my understanding of OpenStack Programs is that they are meta constructs for OpenStack that abstract large areas of function 20:04:59 Projects are code repositories ("openstack/nova") 20:04:59 e.g.: compute, network, storage, etc 20:05:11 Programs are actually *teams* working on a number of those code repostories, aligned with a mission statement ("Infrastructure") 20:05:14 they also help protect OpenStack from name collisions 20:05:26 I think all the confusion stems from sometimes reusing the openstack functional name (i.e. "Compute") as the program name... 20:05:27 as such, a program would contain >0 projects 20:05:39 ...while one of the projects below them (nova) is also using that functional name (= "OpenStack Compute") 20:06:16 ttx, my understading is the OpenStack Compute _is_ the program name 20:06:33 the program name is "compute" 20:06:48 For docs, we have a list for conventional use https://wiki.openstack.org/wiki/Documentation/Conventions#Service_and_project_names 20:06:50 I guess we could rename some of the programs if that helped reducing the confusion 20:07:08 zehicle: basically, the team working on nova and python-novaclient is the compute team 20:07:22 #link http://git.openstack.org/cgit/openstack/governance/tree/reference/programs.yaml 20:07:22 what about glance? 20:07:22 the progral just reflects the structure of the dev teams 20:07:39 hold on 20:07:42 I see the wiki 20:07:43 the team working on glance and python-glanceclient is the image service team 20:08:24 so, are you suggesting that every project eventually becomes a program? 20:08:31 zehicle: no 20:08:46 zehicle: really, it's reusing the same name that creates the confusion 20:09:01 I don't think this matches how we're talking about OpenStack to users - jbryce said that he's focused on "capabliities" whichi are also described as compute, network, etc 20:09:16 zehicle: we couldn't be having this discussion if the team working on nova and python-novaclient was called the Albator team 20:09:25 s/couldn't/wouldn't/ 20:09:27 i think it makes sense to just consider a program to be a grouping of projects 20:09:38 i.e. a program is a set containing 0..n projects 20:09:40 vishy: right 20:09:49 that's where I was going 20:10:11 zehicle: but from the core / board perspective I think you should be looking at projects 20:10:20 as long as the group can contain 0 projects 20:10:31 ttx, no sure 20:10:43 zehicle: programs are just a way to reflect the dev teams organization 20:10:44 ttx: as far as I can tell they are ignoring projects completely and going with capabilities 20:10:49 the board has been working to use capabilities/tests to define core 20:10:52 zehicle, yes jbryce should be talking about capabilities, more than the groups implementing those 20:10:57 and structure reflects how we work, rather than the other way around 20:10:57 s/they/we (i suppose) :) 20:10:59 but there's still a question about use of the OpenStack trademark 20:11:22 zehicle: in program names ? 20:11:24 there's a question about how a project can use the OpenStack trademark 20:11:33 zehicle: or in project names 20:11:33 since we get into challenges w/ OpenStack Neutron, etc 20:11:41 agreed 20:11:48 Program Names are 1) generic and 2) can hold a trademark 20:11:54 zehicle, yes there is that trademark question and the TC is keen to see the board resolve it 20:11:58 so, OpenStack Compute should be protectable 20:12:20 zehicle: the program name is "compute", not "openstack compute", though* 20:12:23 if we agree that Nova is a member of Compute, then we've have a clear way to describe the project 20:12:38 OpenStack Compute Project Nova 20:13:01 "OpenStack Compute" is another name for the project 20:13:02 zehicle: hmmm... not sure about that 20:13:03 not the program 20:13:20 not according to what vishy was saying 20:13:27 "Compute" program has openstack/nova proiject, also known as "OpenStack Compute" 20:13:34 Nova is a member of Compute 20:13:43 but compute could have other memembers 20:13:44 Nova is a member of the compute program 20:13:57 and the Nova project is also known as OpenStack Compute 20:14:02 what is the objection to the Capital C? 20:14:03 the compute program could have other projects 20:14:05 Compute needs images to give capabilities though 20:14:13 whose official names may be OpenStack Foobar 20:14:18 why is compute better/different than Compute 20:14:23 indeed the compute program also holds the gantt and python-novaclient projects 20:14:38 zehicle: "compute" avoids using the trademark altogether 20:14:49 I've been told that Compute also includes glance 20:15:03 not sure that avoiding the trademark is needed/helpful 20:15:11 the compute program doesn't include glance 20:15:14 Glance is developed by a separate team, so Glance is in a separate program 20:15:16 since we specifically want to be able to say "OpenStack Foo" 20:15:21 so we have to deal w/ trademark 20:15:22 and OpenStack Compute is not another name for the glance project 20:15:37 I think we're back to the issue 20:15:41 I see glance as a project 20:15:45 that is a member of a program 20:15:46 it is 20:15:50 indeed 20:15:57 it's not the only member of a program, 20:16:05 and the program is not "Images" it's "Compute" 20:16:06 why should it be the only member? 20:16:24 annegentle: ? 20:16:28 * annegentle attempts to complete zehicle_at_dell's thoughts 20:16:29 zehicle_at_dell, it's the only member right now 20:16:41 well, python-glanceclient sort of 20:16:42 glance and python-glanceclient are the two members of the Image Service program 20:16:44 I am _not_ recommending any membership at this point, just using them as examples. that's a different discussion 20:16:48 markmc: well, not counting python-glanceclient 20:16:54 markmc, we agree on that on 20:16:57 ttx, yeah 20:17:04 but I'm not sure which programs we have 20:17:07 @link http://git.openstack.org/cgit/openstack/governance/tree/reference/programs.yaml 20:17:11 er 20:17:12 #link http://git.openstack.org/cgit/openstack/governance/tree/reference/programs.yaml 20:17:16 ^^ that's our list of programs 20:17:17 sorry, I know which are posted 20:17:24 zehicle_at_dell: that just reflects current reality 20:17:24 IOW, our list of teams 20:17:47 * zehicle_at_dell gets broom 20:18:06 zehicle: I'm starting to think we should have used DIFFERENT nati_ueno mes for the teams, to avoid confusion 20:18:17 names, not nati_ueno, stupid tab completion 20:18:26 perhaps 20:18:33 zehicle: note: we still can 20:18:37 zehicle: easy 20:18:54 I'm not sure that we need for projects to take on generic names 20:19:02 zehicle: would calling it the keystone team work better ? 20:19:04 maybe, but I think it's different 20:19:17 zehicle_at_dell: honestly for docs and a user perspective we have to call it something meaningful not code-named 20:19:18 rather than the identity team ? 20:19:24 zehicle_at_dell: the project is the thing we release, and that's the place where the trademark is important, right? 20:19:33 dhellmann: +1 20:19:39 zehicle_at_dell, 20 minutes in, I really don't know what the DefCore related issue you're getting at here is 20:19:48 zehicle_at_dell, maybe start again from that angle? 20:19:50 ah, circle is complete dhellmann 20:19:57 markmc: I believe that was the reason we suggested a F@F 20:19:57 all of the rest of this is just organizing people, and i'm not sure why that matters so much outside of the people being organized 20:19:59 f@f 20:20:03 annegentle: programs aren't really for consumption by docs or users though... really only openstack developers see them 20:20:42 TC asked the Board to approve then name change of Ceilometer to become a core prgram named OpenStack Telemetry 20:20:57 core project, right? 20:21:03 sigh 20:21:04 core project. 20:21:08 there aren't core projects, anymore 20:21:11 we are not making projects core 20:21:12 just core capabilities 20:21:22 * zehicle_at_dell echos joshuamckenty 20:21:22 joshuamckenty: nor core programs, right ? 20:21:27 right 20:21:35 just integrated releases 20:21:43 the TC wanted the Board to allow the Ceilometer project call itself "OpenStack Telemetry" 20:21:44 there are three specific uses of the trademark 20:21:45 can we make sure we're all using the words with the same meantings before we decide we don't need to use the words? 20:21:48 I think that we could consider programs as core if we handle the membership issue 20:21:53 that we were trying to protect in the bylaws 20:21:58 which, based on our interpretation of the bylaws required board approval 20:22:08 1. Use of the mark by the community of developers to refer to the product of their efforts 20:22:08 we don't care whether that means making the project "core" or not 20:22:11 dhellmann: great idea :) 20:22:15 we just think it should be allowed use that name 20:22:16 we don't want core programs, because programs have multiple projects and not all of the projects in a given program are going to be core 20:22:16 zehicle: I think we shouldn't consider programs as core. They reflect how the team self-organize 20:22:22 2. Use of the mark of vendors to indicate compliance with some definition of "core" 20:22:36 the tests should look at capabilities provided by projects and should not think about programs at all 20:22:38 joshuamckenty, zehicle_at_dell: that is a cool approach, but on the other hand branding the projects as OpenStack definitely has some value 20:22:44 dhellmann, that's a big jump 20:22:59 so perhaps any project that implements some of said capabilites can use OpenStack in its name? 20:23:01 vishy: I'm just describing what's already been approved by the board 20:23:08 vishy, yes. it has value AND it's the only stick that we really have 20:23:09 zehicle_at_dell: when we say "program" we do not mean "computer program" we mean "team of developers" 20:23:13 so why is that a big jump? 20:23:17 so we need to maintain some goverannce on it 20:23:26 joshuamckenty, zehicle_at_dell, the creation of programs (i.e. "teams") is purely a technical governance issue 20:23:36 markmc: not if they use the mark 20:23:43 dhellmann, perhaps we sould say Program (not program) 20:23:44 markmc: but they don't 20:23:48 if you want to call them StackStack Program 20:23:49 joshuamckenty, we're not asking for programs to use the mark 20:23:50 joshuamckenty: ^^ 20:23:53 then there's no issue 20:23:54 zehicle_at_dell: I'm not sure case matters. :-) 20:23:59 zehicle: definitely agree -- that's why programs should stay out of it... since programs are a reflection of our ever-changing team organization, and not of "openstack" capabilities 20:24:00 dhellmann, it's big because you're assuming that integrated = program member 20:24:05 the mark is for projects imo 20:24:12 zehicle_at_dell: yes, that is a requirement from the TC 20:24:20 there has to be a team backing a project for us to integrate it 20:24:34 joshuamckenty, we've created the telemetry program and then *later* asking the board to allow the ceilometer project to use the OpenStack Telemetry name 20:24:39 vishy, that creates problems because we have to protect all the uses (which means ALL the projects) 20:24:49 markmc: you've got the order backwards, though 20:25:04 zehicle_at_dell: well that is the whole reason for having rules 20:25:06 joshuamckenty: what was the 3rd use? 20:25:08 because the "OpenStack Telemetry" program implies that the board has agreed that telemetry capabilities will be part of core 20:25:10 i'm not saying every project gets one by default 20:25:15 which it hasn't 20:25:18 joshuamckenty, no I don't 20:25:18 My understanding (no JD degree) is that we have an issue w/ brand protection if every project could eventually use _OpenStack_ 20:25:26 but there has to be a way for a project to be called OpenStack X 20:25:28 joshuamckenty, I've got the order exactly correct :) 20:25:36 if for no other reason then that's the way it works today 20:25:37 we'd have to vet every project name just in case it got inclubated 20:25:49 vishy: that's not a reason 20:25:54 joshuamckenty: it sounds like you think the board should be setting the technical direction, in terms of features, which is not a responsibility I understood the board to have. 20:25:56 joshuamckenty: actually it is 20:26:18 vishy, I think we'd get to a point where program membership is an additional gate beyond integrated 20:26:23 zehicle: if the telemetry program was called the doctorwho program, would we be having that discussion ? 20:26:24 joshuamckenty: you can't go back and uncall things OpenStack Compute 20:26:26 it is all over the internet 20:26:28 dhellmann: it is precisely the responsibility vested into it by by the bylaws 20:26:30 zehicle_at_dell: we can change the names at incubation time if there is a conflict -- that's part of the incubation process, right? 20:26:37 ttx, yes 20:26:54 even if ceilometer still asked to be called OpenStack telemetry ? 20:26:56 zehicle_at_dell: why make it more complicated then the board must approve using OpenStack in the name of the project? 20:27:01 vishy: true, but that doesn't mean we need a process that allows additional cases 20:27:10 joshuamckenty: agreed 20:27:14 joshuamckenty: that's now how I read 4.1 20:27:25 but involving programs etc. just makes it more complicated 20:27:43 when it could be simply stating integrated projects do not get automatic access to the trademark 20:27:48 vishy, I don't remember the board ever doing that except the predefined ones in the by-laws 20:27:49 dhellmann: we can argue over the language if you want, but I can speak to the intent of the drafting committee 20:27:51 it must be approved by the board 20:27:51 programs are and should stay a reflection of how our teams self-organize 20:27:58 joshuamckenty: later, then 20:28:02 because we can't mandate how teams self-organize 20:28:05 I see programs as irrelevant to this discussing, they're just internal team structure. Defcore only cares about released software, i.e. projects. 20:28:18 mikal: +1 20:28:18 mikal: agreed 20:28:22 DefCore cares about use of the mark that's confusing 20:28:30 and the equation of project to program 20:28:32 are the programs using the mark? 20:28:33 mikal: the confusion is because swe used the same names (the functional description) 20:28:36 So can we ban the word "program" from this conversation to reduce confusion? 20:28:36 but programs aren't using the mark 20:28:36 mikal, so how does that match to vishy comment that programs have >0 projects 20:28:41 does anyone else feel like there actually isn't any debate here, and we just need to state what is actually true 20:28:51 that projects can only use the mark if approved by the board 20:28:53 joshuamckenty: the programs are NOT using the openstack trademark though 20:28:55 and programs cannot use it 20:28:56 zehicle_at_dell: every program has at least one project 20:29:05 vishy, yes 20:29:09 (That I can recall) 20:29:11 joshuamckenty: but just the functional description. 20:29:15 great 20:29:16 simple problem, simple answer 20:29:19 mikal: some programs have 0 projects 20:29:20 well, okay 20:29:21 mikal, can you give me an excample of a program w/ >1 project? 20:29:24 mikal: we have discussed some programs that won't have their own project (UX) 20:29:29 then we just need to make it more clear in community documentation 20:29:29 i.e. "compute program", not "openstack compute program" 20:29:31 jeblair, which ones? 20:29:36 Oh true 20:29:36 zehicle_at_dell: nova, gantt, python-novaclient 20:29:44 that the existence of a program does not indicate that the projects inside it will become part of openstack 20:29:49 joshuamckenty, pointers to where it's not clear are welcome 20:29:53 joshuamckenty: oh, definitely nOT 20:29:54 joshuamckenty: correct 20:29:54 zehicle_at_dell: we've given you an example several times: the compute program has nova, gantt and python-novaclient 20:30:06 oh, it looks like "Release cycle management" did grow a project, but it started without one 20:30:09 joshuamckenty: by doing what? 20:30:18 joshuamckenty: the "release managemnt" program, for example, will NEVER EVER have an openstack name attached to one of its projects 20:30:22 i think we are actually all agreeing we just need to make clear statements about what exists becasue it is confusing 20:30:27 and we have positively discussed that we would be happy to add a ux program with 0 projects (but the team for that hasn't formed yet) 20:30:50 I think we're on a tangent again 20:30:53 https://wiki.openstack.org/wiki/Programs 20:31:06 What I am getting from this is that the board wants to be the only one who can assign the openstack amrk to things 20:31:10 This line is particularly confusing: "The Technical Committee has a mandate to review programs that wish to be included as official OpenStack programs." 20:31:19 So we need to be careful to not call something "OpenStack Foo" until the board has been consulted, yes? 20:31:20 mikal: there is a specific mandate for the community to use the mark 20:31:26 but not to refer to componets 20:31:28 joshuamckenty: agreed, we need to clarify that 20:31:29 joshuamckenty, sounds accurate to me 20:31:30 but just the community itself 20:31:38 they're still official programs 20:31:43 so, we have an unending list of programs? 20:31:46 joshuamckenty: I agree that page is a bit confusing from a mark perspective 20:31:48 even if we don't use the mark with them 20:31:52 they're programs of OpenStack, but not "OpenStack programs" 20:31:57 :) 20:31:58 so lets edit it to clarify 20:32:02 joshuamckenty: I'll take that 20:32:13 "official programs of OpenStack" 20:32:17 "official programs of OpenStack" 20:32:19 or even 20:32:21 right :) 20:32:27 awesome 20:32:28 "TC-recognized prograls of OpenStack" 20:32:32 programs 20:32:35 how did the offiical programs become official? was there a TC vote? 20:32:43 zehicle_at_dell: yes 20:32:43 Yes 20:32:44 yes 20:32:56 there was a mailing list process for existing ones to submit, IIRC 20:33:40 * ttx edited the wiki 20:33:47 So I still don't understand the ask from the board here 20:33:53 joshuamckenty, resolved in under 30 minutes with no need for a f2f :) 20:33:55 Is it just that you want us to stop using the openstack name? 20:34:05 mikal, that programs don't use the mark 20:34:11 mikal: in a confusing way yes 20:34:11 just stop calling them "OpenStack Telemetry", yes 20:34:11 it creates confusion 20:34:13 That's it? 20:34:19 but that does bring us back to *projects* using the name 20:34:31 joshuamckenty: in retrospect, we should have called those programs by the codename 20:34:38 i.e. the nova program 20:34:39 from an official documentation perspective, then the second change request is for the documentation conventions to stop using OpenStack Telemetry, right? 20:34:44 I think the gap is larger than I thought 20:34:48 and avoid confusion with the trademark altogether 20:34:48 the TC would still like Ceilometer project to be able to use the OpenStack Telemetry name 20:34:57 What if we just always have "team" at the end of program names... "Compute Team", "UX Team", etc etc 20:35:30 and the TC doesn't mean by that request that Ceilometer should be a required part of OpenStack clouds 20:35:32 mikal: i like that because it helps us continue to use the generic name which i think has value 20:35:36 mikal: blame mordred for suggesting "program" instead of "team" 20:35:40 what happens if there's another project that wants to use the same program name? 20:35:44 that way a project that has capabilities in core can be called "OpenStack Compute 20:35:44 without overlap on the program 20:35:44 Awesome, can we cover one other defcore issue? 20:35:48 for example another approach to orchtestration? 20:35:54 ttx: if we have to use 'nova', ok, but i think it's good that 'compute program' doesn't mean 'just nova'. 20:35:54 markmc: yes, that's my second question from this 20:35:55 joshuamckenty: quick yes 20:35:58 would Orchestration has two projects? 20:36:08 zehicle_at_dell: we would ask them to collaborate with the existing team and possibly create another project 20:36:13 markmc: ceilometer can only be "openstack telemetry" if the board picks a capability it provides to be in the core set? 20:36:15 markmc: we're trying to have those two things mean the same thing 20:36:15 So would it be fair to say that "The integrated release of OpenStack includes Telemetry in the Havana version" ? 20:36:18 zehicle: if the same team works on it yes 20:36:19 #link https://review.openstack.org/#/c/55375/2/resolutions/20131106-ceilometer-and-heat-official-names 20:36:28 wherever possible 20:36:29 e.g., OpenStack means that some foo capabilities are part of core 20:36:43 mikal: yes 20:36:50 joshuamckenty: what's the second issue ? 20:36:51 that's the current theory 20:36:56 designated sections 20:37:00 So we can never call something "OpenStack X" until after a capability is blessed? 20:37:15 we need the PTLs to decide what code sections are designated sections 20:37:16 mikal: the board decides how the trademark is used 20:37:27 markmc: has the board taken that up yet? 20:37:30 In that case, can't we just say to the board "bless this name, this project provides the following capabiities we would like to have considered for the core set"? 20:37:44 jeblair, no, it's gotten all mixed up in this defcore stuff 20:37:46 ttx: I get that, I'm trying to work out how we interact over it though 20:37:47 joshuamckenty: what is "designated sections"? 20:37:51 what does same team mean? if there are two projects, would they be the same team? 20:38:03 zehicle_at_dell: team == program 20:38:21 zehicle_at_dell: yes. A team can work on more than one project at a time. I thought we'd covered that. 20:38:22 https://wiki.openstack.org/wiki/Governance/CoreDefinition 20:38:24 zehicle: if they share the same core reviewers, then they are the same team 20:38:26 dhellmann: ==^ 20:38:42 section 4 20:38:51 as a hypothetical example, 20:38:56 back 20:38:57 since I'm not supposed to use the term "plugin", 20:38:58 I don't think you're leaving room for parallel efforts 20:38:59 sorry about that 20:39:21 nova might make the "conductor" a designated section 20:39:24 joshuamckenty: are you seriously asking for line numbers within specific files? 20:39:31 if we're saying team == program because they are commonly managed 20:39:38 no, module names or some such is probably fine 20:39:42 joshuamckenty: how does a section differ from a capability? 20:39:47 zehicle_at_dell, I think you're on to another topic altogether, unrelated to DefCore 20:40:00 capabilities don't have to be implemented with the same code unless that code is a designated section 20:40:06 zehicle_at_dell, i.e. "how the TC should run its technical meritocracy" 20:40:09 e.g., neutron plugins 20:40:09 Wait what? 20:40:10 joshuamckenty: tying core branding to code organization is going to make a mess when we refactor something 20:40:14 and floating ip support 20:40:27 markmc, asking questions to resolve word choices 20:40:31 You want us to list the blocks of code that implement floating IP support? 20:40:36 no 20:40:38 the reverse 20:40:39 zehicle_at_dell: parallel efforts as part of the integrated release is a poor choice imo, but this is a tangent 20:40:48 You want us to list the blocks of code which don' 20:40:49 the blocks of code that vendors MUST INCLUDE to use the mark 20:40:53 t implement floating IP support? 20:40:56 but it does seem to me that I should be able to ask questions about it without that push back 20:41:09 mikal: they want a list of modules that must be included to still call the thing e.g. nova 20:41:15 joshuamckenty: that's a lot more granular than my original expectation 20:41:16 all of them 20:41:21 jeblair: +1 20:41:23 jeblair: nope 20:41:27 joshuamckenty: I was expecting "all of nova except the hypervisor drivers" 20:41:28 because some of them are hypervisor specific 20:41:30 to prevent people from rewriting the whole thing and still calling it OpenStack 20:41:35 mikal: exactly! 20:41:39 each project gets to designate 20:41:44 joshuamckenty: no, the hypervisor specific stuff is behind a clear line in the code 20:41:46 neutron will be less, probably 20:41:47 joshuamckenty: makes sense to me 20:41:48 zehicle_at_dell, we're trying to avoid tangents, is all 20:41:49 we left it to the TC to make recommendations about how to handle it 20:42:02 joshuamckenty: curious why this is left up to the PTLs 20:42:05 joshuamckenty: so "cinder-create, delete, ...." 20:42:05 glance is likely 100% as well 20:42:11 well 20:42:12 markmc, I don't think we resolved the issues that started this 20:42:12 joshuamckenty: literally call out all of the API calls? 20:42:12 vishy: who would you suggest instead? 20:42:19 joshuamckenty: doubtful there are driver backends for glance 20:42:21 jgriffith: not api calls, code modules 20:42:21 and now we're way off 20:42:25 Because modules doesn't make any sense 20:42:33 jgriffith: why not? 20:42:36 joshuamckenty: tc would be another choice 20:42:37 vishy: joshuamckenty: this does sound more product manager type questions than tech leads 20:42:37 cinder has drivers 20:42:46 i think this topic is very interesting but probably merits a clear/detailed proposal for us to review 20:42:48