15:00:37 <pknirsch> #startmeeting Fedora Base Design Working Group (2013-11-29)
15:00:37 <zodbot> Meeting started Fri Nov 29 15:00:37 2013 UTC.  The chair is pknirsch. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:00:37 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
15:00:50 <pknirsch> #meetingname  Fedora Base Design Working Group
15:00:50 <zodbot> The meeting name has been set to 'fedora_base_design_working_group'
15:01:04 <pknirsch> Hello and good day everyone!
15:01:18 <pknirsch> Not sure how many we have for todays meeting due to Thanksgiving in the US
15:01:24 * masta is here
15:01:29 <pknirsch> heya masta !
15:01:34 <pknirsch> #chair masta
15:01:34 <zodbot> Current chairs: masta pknirsch
15:02:41 <pknirsch> jreznik already told me he's on PTO today as well and won't be able to join.
15:03:08 <pknirsch> so that leaves probably only haraldh and dgilmore i think
15:03:18 <haraldh> <-
15:03:24 <pknirsch> heya haraldh :)
15:03:29 <pknirsch> #chair haraldh
15:03:29 <zodbot> Current chairs: haraldh masta pknirsch
15:05:06 <pknirsch> So we can do a quick and mainly informative meeting today if you like. I've put together a small table on https://fedoraproject.org/wiki/Base#Details with a few repoqueries to find out a bit more about what packages we would need for different scenarios
15:05:22 <pknirsch> Mainly looking at only kernel, yum and rpm
15:05:26 <pknirsch> then adding gcc
15:05:32 <pknirsch> then adding pungi resp. anaconda
15:05:43 <pknirsch> the numbers were quite interesting i think
15:05:48 <haraldh> systemd?
15:06:22 <pknirsch> is already pulled in via requirements by kernel/rpm/yum
15:06:27 <haraldh> hmm, ok, kernel requires systemd anyway, IIRC
15:06:29 <haraldh> yes
15:06:30 <pknirsch> syep
15:06:49 <haraldh> ~$ rpm -q --requires kernel
15:06:49 <haraldh> fileutils
15:06:49 <haraldh> systemd >= 203-2
15:06:49 <haraldh> dracut >= 027
15:06:51 <haraldh> ...
15:06:52 <haraldh> ok
15:07:07 <pknirsch> should that be something else?
15:07:12 <pknirsch> i mean
15:07:14 <haraldh> nono
15:07:30 <haraldh> kernel rpm yum gcc anaconda 	713
15:07:36 <haraldh> sounds pretty good
15:07:47 <pknirsch> i see one of our tasks as well to go through those lists in detail and figure our if there are any things we could prune or correct
15:07:56 <haraldh> but, is it enough to build itsselves?
15:08:06 <pknirsch> i'd need to double check that
15:08:11 <pknirsch> probably not
15:08:25 <haraldh> do we want mock etc.?
15:08:32 <pknirsch> i'll send out an email to bill to ask how he determined the 2000+ packages needed for self hosting
15:08:37 <pknirsch> hm, true
15:08:44 <pknirsch> let me run that quickly
15:09:28 <pknirsch> [root@power03 ~]# repoquery --arch=ppc64,noarch --requires --resolve --recursive kernel yum rpm mock | sort -u | wc
15:09:28 <pknirsch> 138     138    4349
15:09:31 <pknirsch> not bad
15:10:27 <pknirsch> what i found really nice was that without anaconda but with pungi for creating installer images we're only at about 230 packages
15:10:52 <masta> better than expected
15:11:11 <pknirsch> #action pknirsch to ping notting about how he queried self hosting package set
15:11:16 <pknirsch> yep
15:11:32 <pknirsch> but i fear that the self hosting requirement will pull in quite a few more packages and nasty chains
15:11:40 <pknirsch> which is what bill last week mentioned
15:11:52 <pknirsch> most likely the whole documentation chain itself with texlive etc
15:12:44 <haraldh> ah, yes
15:13:25 <haraldh> shouldn't documentation be shipped kind of precompiled?
15:14:00 <pknirsch> it does for some projects while others do it during their builds
15:14:15 <haraldh> but yeah, we should also declare that as our goal
15:14:27 <haraldh> -> work to minimize buildreqs
15:14:31 <pknirsch> mhm
15:15:01 <pknirsch> as a small and build complete base will make things a lot easier for other people to use it
15:15:01 <haraldh> start with low hanging fruits
15:15:45 <haraldh> maybe even ship a simple copy/installer script with base :)
15:15:47 <pknirsch> it's kinda boring janitorial work, but i don't think we've done a lot of that in the psat years
15:16:16 <haraldh> base "clone" installer :)
15:16:34 <pknirsch> might just use pungi for that?
15:16:36 <pknirsch> that does it
15:17:02 <pknirsch> though hm, that probably pulls in anaconda into the image
15:17:08 <pknirsch> but could write other templates i guess
15:17:09 <haraldh> yeah
15:17:18 <haraldh> anywayx
15:18:14 <pknirsch> #info One potential goal for the committee should be janitorial work on build require cleanup. Mark for topic for one of the next meetings
15:20:20 <haraldh> shouldn't systemd&dracut be part of the Base ?
15:20:28 <haraldh> key packages?
15:20:40 <haraldh> I mean PID 1 ...
15:21:14 <masta> I would guess they are in Base
15:21:23 <masta> assume even
15:21:25 <haraldh> https://fedoraproject.org/wiki/Base#Details
15:21:59 <pknirsch> ah, sure, please add them
15:22:07 <pknirsch> so it's officially documented
15:22:20 <haraldh> done
15:22:23 <pknirsch> cheers
15:22:43 <haraldh> now we can bootup :)
15:22:46 <pknirsch> heh
15:22:47 <pknirsch> ye
15:22:57 <haraldh> and run rpm :)
15:23:01 <pknirsch> jup
15:23:12 <haraldh> add a shell?
15:23:20 <pknirsch> who needs a shell? ;)
15:23:34 <haraldh> right
15:23:38 <pknirsch> e.g. take a really pure desktop system
15:23:44 <pknirsch> that might not ever need a shell
15:23:44 <haraldh> right
15:23:50 <pknirsch> e.g. android
15:24:05 <pknirsch> (though i suspect they have one)
15:24:10 <haraldh> they do
15:24:13 <pknirsch> mhm
15:24:29 <haraldh> selinux?
15:24:46 <pknirsch> hm
15:25:43 <pknirsch> from a broad perspective, no. e.g. embedded systems might not need them and it would have a relatively big performance impact there
15:25:49 <haraldh> I think most of the other WG define products which need it
15:25:50 <pknirsch> does systemd work without selinux?
15:25:55 <pknirsch> yea
15:26:10 <haraldh> systemd is linked ofc with the libraries
15:26:15 <haraldh> but works fine without
15:26:25 <haraldh> even if linked with the libs
15:26:25 <pknirsch> if most or all WGs need selinux i think it should be in Base
15:27:47 <haraldh> I think, we should make a list with "minimal" and "base"
15:27:52 <pknirsch> at least thats how i see it: Base should provide those fundamental functionalities that are being used by all or almost all WGs
15:27:54 <haraldh> whereas selinux is in "base"
15:27:58 <pknirsch> mhm
15:28:52 <haraldh> hmm
15:29:06 <haraldh> "minimal" for containers is different
15:29:28 <haraldh> perhaps only "glibc" :)
15:30:34 <masta> haraldh: that is a good idea. by the way, do we also control the idea of minimal or just base?
15:30:35 <pknirsch> let me see what that produces actually
15:30:50 <pknirsch> 15 packages, no kernel :)
15:31:25 <pknirsch> well, any WG can define their "minimal" set of packages
15:31:32 <haraldh> masta, well, I see it as my goal to define "minimal"
15:31:37 <pknirsch> but for Base i think it would be really good to do so
15:31:44 <haraldh> and keep it as "minimal" as it can be
15:31:50 <pknirsch> for different scenarios that we see Base being useful for
15:31:55 <pknirsch> e.g. the glibc case
15:31:59 <pknirsch> or the kernel only one
15:32:29 <pknirsch> not that i'd encourage it, but i could see someone wanting to use a very minimalistic fedora and then use apt or debpkg on top of it
15:32:37 <haraldh> hmm, kernel only without modules is not so much working
15:33:11 <pknirsch> doesn't the kernel package include all the modules?
15:33:19 * pknirsch may be a bit behind on that
15:33:27 <haraldh> yeah, but something has to load them
15:33:31 <pknirsch> ohhh
15:33:33 <pknirsch> hehe
15:34:00 <haraldh> ofc, one could go with kernel and kmod alone
15:34:10 <haraldh> and his own PID1
15:34:27 <haraldh> udev is not needed even
15:34:54 <pknirsch> that would be rather minimalistic
15:35:02 <haraldh> as minimal as it can be
15:35:10 <haraldh> booting a machine
15:35:36 <pknirsch> but ye, lets collect those different things and provide info about them. and maybe one or the other is useful for someone
15:36:03 <pknirsch> e.g. imagine a hpc cluster with thousands of machines. a very lightweight minimal system for them would be really good, too
15:36:29 <haraldh> I'll make graph, what is needed for what minimal stage, until next week
15:36:35 <haraldh> ^ a
15:36:36 <pknirsch> thanks!
15:38:02 <masta> for minimal things I always thing maybe util linux, core utils, systemd/udev, selinux, auditing, rpm
15:38:08 <haraldh> grub2/gummiboot/syslinux .. any thoughts?
15:38:12 <pknirsch> doh, i forgot the #topic again at the start :)
15:38:59 <pknirsch> masta: would that reflect your embedded systems?
15:39:05 <pknirsch> even with selinux?
15:39:10 <masta> oh yes
15:39:14 <pknirsch> hm, ok
15:39:19 <haraldh> auditd?
15:39:45 <masta> I think I discovered audit stuff is required at some point
15:40:18 <masta> but we don't have to say audit stuff is required here
15:40:33 <pknirsch> mhm
15:40:44 <haraldh> I don't think auditd is technically required
15:40:48 <pknirsch> but still that would be interesting info where your requirement for audit came from?
15:41:07 <masta> actually scratch that, auditd was not required actually, just the AUDITSYSCALL kernel parameter was required for systemd
15:41:14 <haraldh> ah
15:41:15 <haraldh> ok
15:41:17 <pknirsch> ah
15:41:17 <masta> sry about that
15:41:18 <pknirsch> ye
15:41:21 <pknirsch> no worries :)
15:42:54 <pknirsch> and as mentioned, feel free to add your info to the wiki, either with a separate use case or under the details or where you see the fit
15:43:43 <pknirsch> having more review of the base system is definitively a precursor to how everything will look like later
15:43:50 <pknirsch> and what we can share with the other WGs
15:44:34 <masta> well I like the idea we put some of these down in the minimal, and we just bring in minimal to make Base... so we define both. I'm surprised all the other WGs get to redefine minimal
15:45:26 <pknirsch> well, they can if they think it's useful for their specific variant
15:45:32 <pknirsch> imagine a Workstation
15:45:45 <pknirsch> you could think of a minimal workstation as well
15:45:49 <haraldh> X would be part of their minimal
15:45:52 <pknirsch> mhm
15:46:00 <pknirsch> maybe with a login and a desktop manager
15:46:20 <pknirsch> and anything after that is not really minimal anymore for them
15:46:41 <pknirsch> maybe i expressed myself a bit wrong there, those minimal sets would be for their variant, not for Base
15:46:59 <pknirsch> same for Server
15:47:05 <masta> hmm
15:47:29 <pknirsch> they might decide a Server has to have some functionality, e.g. ssh, config mgmt, IPA
15:48:44 <pknirsch> and maybe (and thats just an crazy idea here now) the minimal set for Base would then be the set of packages shared by all variants/products in their minimal representations
15:49:08 <pknirsch> it shouldn't be much different from what we're looking at right now i suspect
15:49:17 <masta> ok
15:49:34 <pknirsch> but again, thats just my view
15:49:38 <pknirsch> :)
15:51:44 <pknirsch> we might end up defining Base differently, depending on what and how other WGs see us. I'll get more involved over the next week with them to see a bit more about their concrete requirements and views of Base
15:52:01 <haraldh> right
15:52:17 <haraldh> but that does not stop us from working on the real minimal sets
15:52:22 <pknirsch> exactly
15:52:25 <haraldh> right now
15:52:28 <pknirsch> as thats something we will definitely need
15:52:39 <pknirsch> no matter what any of the WGs decide to need in addition to that
15:52:46 <haraldh> correct
15:52:57 <masta> seems right
15:53:05 <pknirsch> very well put haraldh :)
15:53:22 <haraldh> Btw, what are our super powers right now?
15:53:31 <Viking-Ice> why dont you just keep the core to something as small as what makes up mir or tizen?
15:53:32 <pknirsch> LAZOR BEAMS!
15:53:37 <haraldh> Do we have to do the work? Do we have to file BZs?
15:54:04 <pknirsch> Viking-Ice: so how small is mir or tizen then?
15:54:06 <Viking-Ice> http://releases.merproject.org/~carsten/mer-core-i586.ks
15:54:46 <haraldh> @Mer Core
15:54:53 <haraldh> well, that's the point :)
15:54:55 <Viking-Ice> http://gitweb.merproject.org/gitweb?p=mer/project-core.git;a=tree;f=patterns;hb=HEAD
15:55:22 <Viking-Ice> you can see there in the raw output what exactly is in mer-core
15:55:26 <haraldh> Viking-Ice, we come automatically to that point :)
15:55:40 <haraldh> as we work our work up
15:55:51 <haraldh> as we work our way up
15:56:00 <haraldh> in the dependency foodchain
15:56:38 <Viking-Ice> yeah but it's unavoidable from a logical sense that the other WG's will have to define their own base since the needs on server ( or cloude ) != the same needs on workstation
15:56:48 <haraldh> sure
15:57:30 <Viking-Ice> hence I proposed 3 layer solution 1 core ( what you define ) one base ( which was mutual for each server role ) then each role by itself
15:58:23 <haraldh> if you call it Base-"base" or "1 core" does not matter
15:58:47 <Viking-Ice> the same can be applied to the workstation group if they stop fighting amongst themselves and come to the only logical conclusion one base ( mutual for each DE ) then each DE by itself
15:58:52 <haraldh> additional to Base-"base", I also want to have a Base-"minimal" :)
15:59:09 <pknirsch> right Viking-Ice
15:59:19 <pknirsch> thats what i meant with my minimal sets for each variant
16:00:37 <Viking-Ice> haraldh, I dont think you should split the "core" to core - minimal and or core - extras etc
16:00:58 <Viking-Ice> since as soon as you start doing that you will start more headache for us in QA
16:01:34 <Viking-Ice> then atleast I see us wanting
16:01:49 <pknirsch> as you'd then have to test all the variants in Base, right Viking-Ice ?
16:01:57 <pknirsch> or at least would feel inclined to do so
16:02:20 <pknirsch> thereby blowing up the whole test matrix
16:02:31 <masta> got a hard stop, must go now. =(
16:02:34 <masta> bye all
16:02:36 <pknirsch> cya masta !
16:02:59 <Viking-Ice> pknirsch, yeah the way forward that I will be proposing for us to do in QA is to that we try to move into grouped updates and test them as well as build an QA community layer on top of that
16:03:27 <pknirsch> mhm
16:03:30 <haraldh> "grouped updates" - yes, plz
16:03:40 <haraldh> that's what I proposed with my release cycle
16:04:31 <haraldh> grouped "base" updates
16:04:42 <pknirsch> yep
16:04:43 <haraldh> therefore "base" has to be pretty self containing
16:05:21 <Viking-Ice> the length of the release cycle in your proposal was bit off but yes we need to approach this as group in QA we already discussed all that before the the current RH team dedicate to Fedora QA
16:05:22 <haraldh> and probably even self buildable
16:05:48 * pknirsch nods
16:05:51 <Viking-Ice> yes but getting there is well way in the future
16:06:14 <haraldh> doesn't hurt to set it as a long term goal
16:06:33 <pknirsch> i might have something i can throw at this, we'll see
16:06:48 <Viking-Ice> once I have finished the community proposal I will work on a new build system proposal what David A. said made sense as an future approach to such system
16:06:49 <pknirsch> been putting together a bunch of scripts for automated rebuilds and composes
16:07:13 <pknirsch> David A?
16:08:36 <Viking-Ice> David Airlie
16:09:27 <pknirsch> ah, ok
16:09:39 <Viking-Ice> the next gen community platform I currently working on requires new bugzilla/build system ( or modification to existing ones ) thus it makes sense to design a new system and or get discussion going between distros about that
16:10:44 <pknirsch> sure
16:11:17 <Viking-Ice> in anycase that community platform might be suited for Fedora or Fedora not ready for such distro agnostic approach and change ( which was one of the reason I walked out of the serverWG )
16:11:28 <Viking-Ice> might not be I mean
16:11:36 <Viking-Ice> ( suited for )
16:12:34 <pknirsch> mhm, i know what you mean
16:14:11 <Viking-Ice> anywho just few notes to this meeting...
16:14:15 * Viking-Ice goes back to work
16:14:53 <pknirsch> alright, think we're done anyway. or haraldh, anything else from you?
16:14:57 <haraldh> so, pknirsch, let's make some decisions :) Ready to vote? :)
16:15:00 <pknirsch> hahaha
16:15:15 <haraldh> yeah.. done
16:15:44 <pknirsch> oki, thanks masta and haraldh and Viking-Ice for being here today! and have a great weekend!
16:15:51 <haraldh> u2
16:15:57 <pknirsch> #endmeeting