15:00:00 <nils> #startmeeting modularity_wg
15:00:00 <zodbot> Meeting started Tue Aug 30 15:00:00 2016 UTC.  The chair is nils. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:00:00 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
15:00:00 <zodbot> The meeting name has been set to 'modularity_wg'
15:00:00 <nils> #meetingtopic Weekly meeting of Modularity WG
15:00:00 <nils> #chair dgilmore haraldh langdon tflink
15:00:00 <zodbot> Current chairs: dgilmore haraldh langdon nils tflink
15:00:14 <nils> #topic Roll Call
15:00:20 <nils> .hello nphilipp
15:00:20 <cpacheco> .hello cpacheco
15:00:20 <zodbot> nils: nphilipp 'Nils Philippsen' <nphilipp@redhat.com>
15:00:24 <zodbot> cpacheco: cpacheco 'Courtney Pacheco' <cpacheco@redhat.com>
15:00:31 <jkurik> .hello jkurik
15:00:32 <zodbot> jkurik: jkurik 'Jan Kurik' <jkurik@redhat.com>
15:00:43 <langdon> .hello langdon
15:00:44 <zodbot> langdon: langdon 'Langdon White' <langdon@fishjump.com>
15:00:50 <sct> .hello sct
15:00:51 <zodbot> sct: sct 'Stephen Tweedie' <sct@redhat.com>
15:01:12 <nils> #chair sct
15:01:12 <zodbot> Current chairs: dgilmore haraldh langdon nils sct tflink
15:01:32 <clime> hey
15:02:19 <nils> #topic Agenda
15:02:19 <nils> #info Containers (langdon, sct)
15:02:19 <nils> #info Badge
15:02:34 * threebean waves
15:02:36 <dgilmore> hey all
15:02:52 <nils> does anybody want to add something to the agenda?
15:03:12 <asamalik> .hello asamalik
15:03:13 <zodbot> asamalik: asamalik 'Adam Samalik' <asamalik@redhat.com>
15:03:19 <moto-timo> .hello ttorling
15:03:20 <zodbot> moto-timo: ttorling 'None' <TicoTimo@gmail.com>
15:03:30 <threebean> nils: pardon.. can you link to the agenda?
15:03:35 <nils> threebean: sure :)
15:03:48 <nils> http://piratepad.nl/modularity-wg-agendas
15:05:01 * langdon just realized he had an action from last time :(
15:05:13 <nils> threebean: hmm, not sure this needs to be in the agenda itself :)
15:05:26 <nils> or should it go into the log?
15:05:52 <threebean> nils: heh, just for future reference so we can link to it for people who don't know it.
15:06:00 <nils> ah
15:06:06 <langdon> It's on the wg wiki page
15:06:33 <nils> ok, nothing additional for the agenda apparently
15:06:52 <nils> #topic Containers (langdon, sct)
15:07:07 <nils> langdon, sct?
15:07:10 <langdon> With threebean and lkockman here can we get an update on Fedora stg?
15:07:21 * langdon typing to slowly
15:07:29 <nils> langdon: I'll tack it onto the end, sec
15:07:33 <langdon> K
15:07:35 <nils> #undo
15:07:35 <zodbot> Removing item from minutes: <MeetBot.items.Topic object at 0x456c5250>
15:08:10 <nils> threebean, leave them #info please for easy copy'n'paste :)
15:08:50 <threebean> sorry, yes.  "helping"  :p
15:08:52 <nils> #info Update on Fedora stg infra (lkocman, threebean)
15:08:56 <threebean> langdon: sure, I can answer questions.
15:09:01 <nils> that was for the record, and the agenda :)
15:09:10 <nils> ok, first topic coming up...
15:09:17 <nils> #info Containers (langdon, sct)
15:09:35 * nils yields the floor
15:09:42 <sct> Heh
15:09:52 <sct> So
15:09:55 <sct> containers are great
15:09:57 <langdon> sct, this was meant to discuss parallel installation.. sorta
15:09:59 <sct> everybody should have one
15:10:10 <sct> but they are complicated too
15:10:11 <sct> so
15:10:18 * langdon doesn't have sliced bread yet.. will containers work for me?!?!
15:10:28 <moto-timo> and a new concept to many that apparently flumoxes them
15:10:36 <sct> short answer is, Langdon and I are trying to figure all the options for installing multiple variants of dependencies at the same time...
15:10:56 <sct> we've got different solutions --- traditional library parallel install (but that only works for major versions, doesn't work at the errata level)
15:11:03 <sct> SCL-type solutions, containers etc
15:11:16 <sct> and the matrix of possibilities gets pretty large
15:11:37 <langdon> and the matrix of pain
15:11:44 <sct> so Langdon and I are still refining it.
15:12:12 <sct> So TL:DR; it's complicated and a WIP, we'll come back once we've got a more consumable map of the options.
15:12:50 <langdon> probably a ML item
15:12:51 <sct> The one thing we are trying to avoid is a big bang,
15:13:09 <langdon> and pain
15:13:10 <sct> where everything needs to be perfectly modularisable in all possible combinations all at once
15:13:22 <moto-timo> singularity
15:13:26 <langdon> yes
15:13:30 <langdon> skynet
15:13:30 <nils> moto-timo: beat me to it
15:13:31 <sct> it will likely be multiple options and we can work on different options over time.
15:14:36 <nils> But we can probably assume that we need to use whatever method of separating modules to be able to install side-by-side, right?
15:14:58 <langdon> right.. but .. do different use cases have different choices?
15:15:09 <sct> nils: Yes and no.
15:15:10 <langdon> can there be a pool of shared libraries that are reduced over time?
15:15:12 <nils> I mean, that e.g. a set of packages is completely disparate, not stepping on each others toes, would be the exception.
15:15:16 <langdon> does everything have to be a container
15:15:18 * asamalik wants to have 3 different versions of docker running at the same time, each in a different container provider
15:15:19 <sct> nils: It's a matter of getting the granularity right.
15:15:44 <sct> nils: eg. I really don't expect us to package every part of Gnome so you can choose any version of any Gnome app and they all work perfectly together
15:15:50 <moto-timo> flatpak?
15:15:58 <nils> yeah, where we take care of getting shared bases right, in-sync, we can avoid that probably
15:16:03 <sct> there are still going to be some common parts to keep things sane.
15:16:20 <nils> sct: I was more thinking in terms of libraries
15:16:32 <nils> yeah, common parts, not leafs
15:16:36 <sct> nils: Right
15:16:36 <sct> but
15:16:39 <sct> the point still stands
15:16:46 <sct> the implication for libraries depends on what's using them
15:16:48 <langdon> sct, right.. we can start with "everything" being common.. then, over time, make more things unique
15:17:06 * nils likes the idea of an "Everything" module
15:17:18 <sct> something that's a core part of Gnome GUI might have different properties than something like libxml2 that is used by all sorts of apps all over the system.
15:17:26 <nils> yeah
15:17:42 <nils> sct, langdon: anything to #info, the TL;DR from above?
15:17:55 <sct> And it's precisely because the options aren't clear cut --- it's always a balance --- that it gets hard to map out the solutions. :)
15:18:04 <nils> oh yeah
15:19:17 <langdon> i assume sct is typing a #info?
15:19:30 <sct> Can do, sure
15:20:14 <sct> #info Langdon and I are still mapping out all the options for parallel installation of dependencies, the costs and benefits of each; we'll come back when we've got a more usable map of the options.
15:20:36 <langdon> #info the map is hard.. hence not really ready yet
15:21:45 <nils> sct, langdon: good to move on?
15:22:25 <sct> I'm done on that topic, yes
15:22:28 <langdon> nils, yeah
15:22:35 <nils> ok
15:22:48 <nils> #topic Badge
15:23:08 * langdon was supposed to send out an email.. but failed to do so
15:23:36 <nils> langdon: the designers aren't rioting yet ;)
15:24:15 <nils> I participated in a little bike-shedding over on the ticket
15:24:25 <langdon> nils, well done! :)
15:24:27 <nils> #info https://fedorahosted.org/fedora-badges/ticket/466
15:24:44 <nils> langdon: nothing terribly important, just colors :)
15:25:06 <langdon> isnt that pretty much the definition of bike shedding?
15:25:22 <nils> langdon: yes :)
15:25:24 <nils> I think the final suggestion by maryshak1996 is a good one.
15:26:04 <nils> langdon e.a., do you want to put this to a formal vote?
15:26:38 <langdon> nils, i think i was supposed to do that on the ML.. but forgot
15:27:11 <nils> langdon: no biggie I guess. Vote next week=?
15:27:24 <langdon> re-action for me to send to the ML
15:27:59 <langdon> #action langdon to request a vote on the badge on the ML
15:28:12 <nils> good, next
15:28:20 <langdon> +1
15:28:24 <nils> #topic Update on Fedora stg infra (lkocman, threebean)
15:28:28 * threebean waves
15:28:39 <threebean> langdon: can you give a prompt about what you'd like us to talk about here?
15:28:50 <nils> I don't see lkocman, so threebean?
15:29:38 <threebean> I haven't paid much close attention to our stg infra pieces since flock, so no real updates from me.  if there are questions or concerns, I could try to address them.
15:30:07 <langdon> threebean, ohh.. sorry.. basically.. per sgallagh's request in channel yesterday.. can we give some guidance on how someone can make a module yet?
15:30:13 * threebean nods
15:30:35 <threebean> the requirement we have in place currently is still that you need access to stg dist-git in order to commit, push, and build things.
15:30:54 <threebean> luckily (exceptionally) sgallagh is in the provenpackager group, so that just worked without lifting any fingers.
15:31:00 <langdon> ha
15:31:20 <nils> threebean: provenpackagers have access to stg dist-git automatically?
15:31:26 * threebean nods
15:31:30 <threebean> the perms are copied from prod.
15:31:33 * nils takes note
15:31:38 <langdon> ok.. maybe we start with a #info for the first step for anyone who wants to build a module? e.g. #info come ping threebean in #f-m to get started?
15:31:43 <threebean> so since provenpackagers have access everywhere in prod, they also get access everywhere in stg.
15:31:46 <langdon> and a doc on the wiki?
15:32:11 <moto-timo> copr?
15:32:34 <asamalik> moto-timo: yeah, copr team is working on supporting modules
15:32:41 <langdon> moto-timo, the copr team is working on backend integration with our orchestrator (rida) .. asamalik status?
15:32:41 <moto-timo> excellent
15:33:03 <asamalik> moto-timo: there will be a demo of copr building module for fedora 24 and then installing it next week on youtube
15:33:22 <langdon> asamalik, do we have a plan for "source not in dist-git" as well for the "private module build" (read: copr backend)
15:33:29 * moto-timo waits with bated breath
15:33:41 <threebean> #action Develop instructions for how to build a module in the staging infrastructure.
15:33:46 <threebean> #link https://fedoraproject.org/wiki/Modularity/HowToBuildAModuleInStaging
15:33:53 <asamalik> moto-timo: but it will do it a bit differently - the module will be "handcrafted" - built the same way as a copr project now - and the modulemd yaml will be autogenerated by copr - so no need to write it manually
15:33:57 <threebean> I'll fill that out asap.
15:34:05 <asamalik> next step for copr will be support of builds from modulemd yaml
15:34:28 <moto-timo> logic progression, great
15:34:54 <langdon> asamalik, so links to source rpms essentially?
15:35:04 <langdon> assumes all rpms are part of the external api?
15:35:23 <langdon> thanks threebean
15:35:27 <langdon> threebean++
15:36:02 <langdon> any more questions/status/et al on fedora-stg? i mostly wanted to point out we hadn't told anyone how to do it yet except verbally
15:36:17 <asamalik> langdon: in the first demo, you will have the same way of providing rpm sources as you have now - link, direct upload, git with tito, PyPI, Rubygems
15:36:20 <langdon> #action threebean to write up how to get started in fedora stage
15:36:29 <asamalik> langdon: is that what you mean?
15:36:44 <langdon> threebean, can i also ask you to send a note to the ML when you are happy with the wiki content?
15:36:48 * threebean nods
15:36:59 <threebean> will add that to the acceptance criteria ;)
15:37:14 <asamalik> langdon: then, all the packages get saved into Copr's dist-git and built
15:37:29 <langdon> asamalik, yep.. curious about some the "module feature" .. aka filtering binaries, install profiles, private and public binaries
15:37:35 <langdon> threebean, :)
15:37:38 <asamalik> langdon: UI
15:38:03 <asamalik> langdon: basically, you click on "release module" and will need to define API, install profiles, and filters by clicking checkboxes
15:38:12 <asamalik> at least at the first demo
15:38:17 <asamalik> later that would be one of the ways
15:38:22 <langdon> asamalik, ohh cool.. i didn't even expect that much
15:38:39 <asamalik> langdon: copr is cool, you should expect only cool stuff from copr!
15:38:45 <moto-timo> lol
15:39:06 <langdon> asamalik, can we ask them to expose the modulemd.yaml to the user? so they can "build it in copr" then "submit it to fedora-stg" ?
15:39:14 <moto-timo> +1
15:39:18 <asamalik> langdon: yes, no problem
15:39:29 <langdon> asamalik, sweet..
15:39:46 <langdon> i think our example set could grow quite easily with that
15:40:02 * moto-timo drools a little bit
15:40:12 <asamalik> I will publish a blog post soon about Copr in the modularity world
15:40:12 <langdon> as somewhat of an aside.. have we thought about how a modulemd represents a container at all yet?
15:40:17 <asamalik> I already started writing it
15:40:31 <langdon> asamalik, yeah.. just make sure to coordinate it through nardasev ..
15:40:42 <asamalik> langdon: yap
15:41:41 <asamalik> langdon: I gues you just build it as normally and put the result + deps into a container?
15:41:53 <langdon> geppetto, putting you on the spot.. "<langdon> as somewhat of an aside.. have we thought about how a modulemd represents a container at all yet?"
15:42:12 <asamalik> langdon: and as the deps also include the "base runtime", there is everything we need
15:42:21 <geppetto> langdon: As far as I know we haven't
15:42:32 <geppetto> langdon: Maybe sct has?
15:43:22 <asamalik> langdon: ha! except the RUN command :) and ETRYPOINT, and other stuff... :)
15:43:30 <langdon> geppetto, i don't think so.. not even sure petr has.. we should probably get it in the backlog
15:43:34 <geppetto> I'd also guess that we'd need to represent each container type differently, but I could be wrong
15:43:42 <langdon> i have to talk to some container people next week and i think they are gonna ask me
15:44:20 <langdon> geppetto, yeah.. we need something to say "format" .. even just to differentiate rpm from container
15:44:36 <geppetto> Sure, that's the easy bit :)
15:44:59 <nils> langdon: container built from RPMs... *wanders off*
15:44:59 <langdon> geppetto, no lie.. i wonder how we are even gonna carry it
15:45:07 <langdon> nils, :P
15:45:19 <langdon> geppetto, can i ask you to make a card?
15:45:24 <nils> langdon: we could try it the other way round :D
15:45:32 <geppetto> For a spike?
15:45:56 <langdon> geppetto, yes? i think so? we probably need some experimentation
15:46:17 * geppetto nods
15:47:49 <langdon> ok.. i think we have exhausted that subject.. any other thoughts/comments/concerns on stg infra?
15:48:44 <threebean> :)
15:48:55 <langdon> nils, open floor?
15:49:03 <nils> langdon: I was about to say it :)
15:49:07 <nils> #topic Open Floor
15:49:10 <jkurik> just for an info: The server-wg meeting has moved 5 hours later: https://lists.fedoraproject.org/archives/list/server@lists.fedoraproject.org/thread/FJM2IU4UDEIXHSDGOMQMEV64GUZ2ZK4W/
15:49:20 <threebean> jkurik: thanks :)
15:49:42 * moto-timo finally watched the Flock demos and Sprint 10/11 videos
15:50:21 <langdon> moto-timo, comments?
15:50:44 <moto-timo> I like what I see. I like the direction things are taking.
15:51:00 <langdon> cool
15:51:03 <moto-timo> I need to play with it a little bit
15:51:21 <langdon> we would like others doing that.. we need the feedback
15:51:29 <moto-timo> My team uses (abuses?) containers to no end, so we are interested in that aspect
15:51:35 <langdon> ha
15:52:01 <moto-timo> and as I mentioned before, golang is hosed without modularity
15:52:03 * langdon needs to put server-wg on his calendar
15:52:58 <langdon> ok.. anybody got anything else?
15:53:12 * asamalik does not have anything else
15:53:17 * threebean moves to close
15:53:36 <nils> sounds good to me
15:53:49 * langdon read that as "threebean moves closer" which he thought was a weird remark to make
15:54:09 <nils> haha
15:54:18 <moto-timo> lol
15:54:20 <nils> okay
15:54:28 <nils> thanks for coming everybody
15:54:30 <nils> #endmeeting