15:00:56 <pknirsch> #startmeeting Fedora Base Design Working Group (2014-01-17)
15:00:56 <zodbot> Meeting started Fri Jan 17 15:00:56 2014 UTC.  The chair is pknirsch. Information about MeetBot at http://wiki.debian.org/MeetBot.
15:00:56 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
15:01:05 <pknirsch> #meetingname  Fedora Base Design Working Group
15:01:05 <zodbot> The meeting name has been set to 'fedora_base_design_working_group'
15:01:14 <pknirsch> hello everyone :)
15:01:43 <jwb> hi
15:02:09 <dgilmore> guten tag
15:02:31 <pknirsch> :)
15:02:41 * notting is here
15:03:40 <haraldh> <-
15:04:37 <pknirsch> ok, lets get started then.
15:05:05 <pknirsch> #topic Self hosting detailed review & discussion & next steps: http://www.harald-hoyer.de/2014/01/14/self-hosting-fedora-base/
15:05:35 <haraldh> yeah, basically read my blog post :)
15:05:35 <dgilmore> its not as bad as i was expecting
15:05:43 <dgilmore> but it is a mess
15:05:53 <pknirsch> mhm, but thanks a lot haraldh for doing this
15:06:09 <jwb> haraldh, which kernel do you analyze?  just curious
15:06:10 <dgilmore> we have done 2 arm bootstraps in fedora in the last couple of years, and another one is about to start
15:06:23 <dgilmore> s/arm/arch/
15:06:23 <haraldh> jwb, current F20 kernel
15:06:28 <haraldh> jwb, current F20 kernel x86_64
15:06:53 <jwb> haraldh, ah, ok.  because i fixed the docs stuff in rawhide
15:07:46 <haraldh> with modifying approx. 65 specs we could get a reduced buildroot set of 188 + perl rpms , with which we can build the kernel
15:08:10 <haraldh> and all of the 188+ packages also
15:08:21 <haraldh> so, basically the base buildroot
15:08:35 <haraldh> with which we can build the whole distro step by step
15:09:10 <pknirsch> which would be one of the benefits, right?
15:09:37 <haraldh> 1. defined self contained build root
15:09:46 <haraldh> 2. easier for new archs
15:09:53 <haraldh> 3. provable rebuilds
15:10:03 <notting> what are the modifications required? just prebuild some docs?
15:10:06 <dgilmore> haraldh: we always start new arches with gcc.
15:10:29 <dgilmore> but yeah
15:10:34 <haraldh> notting, not only
15:10:49 <haraldh> notting, e.g. gobject-introspection
15:11:01 <haraldh> you need a package with reduced functionality
15:11:36 <haraldh> anyway, we would never "ship" those packages with the reduced functionality
15:11:54 <haraldh> otherwise we would have variants of the same src.rpm
15:12:02 <haraldh> which would be a support nightmare
15:12:04 <pknirsch> so you'd propose something like a flag in specfiles to do this "reduced" set?
15:12:14 <pknirsch> like the bootstrap flags for perl?
15:12:15 <haraldh> pknirsch, exactly
15:12:47 <dgilmore> haraldh: not shipping them would require changes in koji or workflow
15:13:03 <dgilmore> we have no mechanism to not ship them
15:13:08 <jwb> technically, not true
15:13:11 <jwb> we have glibc32
15:13:16 <jwb> it isn't shipped
15:13:30 <dgilmore> jwb: right but its a different workflow
15:13:34 <haraldh> well, these packages are only needed to build themselves later on with the full functionality
15:13:35 <dgilmore> and hasnt been updated in years
15:14:02 <dgilmore> we would need to use different tags/targets to build them
15:14:31 <dgilmore> likely different branches in git or people would need to make sure to use --target
15:14:49 <haraldh> of course, if you say, that we should ship "gobject-introspection-reduced" with the same binaries...
15:15:02 <dgilmore> if it was to be common we would need to define a workflow and process to do the building
15:15:12 <dgilmore> haraldh: and if we dont ship them how can base be self hosting
15:15:33 <haraldh> long story .. short... we can't be self hosting :)
15:15:52 <dgilmore> haraldh: we can, but not in your view of the world
15:15:57 <haraldh> at least not with 188+ packages :)
15:16:14 <haraldh> we can, with 1806 packages, though
15:16:34 <haraldh> as you can see in the first graph of my blog post
15:17:03 <dgilmore> id rather the 1806 packages
15:17:15 <dgilmore> and be self hosting in what we ship
15:17:19 <haraldh> but... I don't want to have qt, kdelibs, gtk2 .. etc. in BASE
15:17:49 <haraldh> but that is only my personal opinion
15:17:51 <sghosh> this is an important distiction to make - do we want the distribution as shipped t be self hosting - or do we want to ship the capability for rebuild from scratch
15:18:13 <haraldh> sghosh, well the distribution _is_ self hosting
15:18:20 <sghosh> assumption that the two are the same is not correct any more
15:18:25 <haraldh> the question is, if "base" should be self hosting
15:18:41 <dgilmore> i think strongly base should be self hosting
15:18:43 <sghosh> sorry - talikng about the base - same scope as you
15:19:09 <dgilmore> I feel the definition we had at the start for base was right
15:19:43 <notting> having a small-ish set that's used for bootstrapping seems worthwhile even without deep changes to what we ship
15:19:44 <haraldh> well, some packages lose functionality, if you don't have e.g. "kdelibs-devel" in the build root
15:19:52 <dgilmore> anaconda, base build tools, compose tools, minimal install should all be in base
15:20:13 <dgilmore> gtk3 has to be in base unless we drop out anaconda
15:20:29 <jwb> notting, so <package>-bootstrap?
15:20:38 <pknirsch> right notting, especially if we "only" would have to modify a small set of packages to achieve that and, in case of a bootstrap can then build them like that
15:20:53 <dgilmore> documenting how to bootstrap the distro holds value
15:21:05 <pknirsch> which wouldn't change the normal workflow for general builds
15:21:55 <pknirsch> and we've discussed the idea of split srpms for docs and other stuff already. i know SuSE does it like that, but it's a significant burden i think for our maintainers
15:22:04 <pknirsch> and we rejected that idea
15:22:58 <notting> jwb: something like that. whether that's a separate source rpm, or a macro-ized version of the same spec
15:23:06 <jwb> yeah
15:23:12 <pknirsch> though i think it would be worthwhile to investigate if any of the changes to those 65 spec files could be general changes (so not bootstrap only)
15:23:16 <haraldh> I would vote for a  macro-ized version
15:23:27 <haraldh> pknirsch, yep
15:23:48 <pknirsch> which would very likely already result in a smaller footprint for self hosting base
15:24:15 <haraldh> or in my list, e.g. there is valgrind, which you could get rid of, if you throw out valgrind in those specs BR it
15:24:17 <dgilmore> being able to "%define bootstrap 1" and build seems like the best way
15:24:17 <pknirsch> and i do agree with dgilmore, self hosting feels right imho
15:24:24 * pknirsch nods
15:24:32 <dgilmore> though we would need to reguallry do the bootstrap to make sure its right
15:25:11 <pknirsch> yep, and thankfully we have scripts now to verify that, too and produce the dep graphs now
15:25:28 <haraldh> so, we have 1806 packages in our first "Fedora Base" set?
15:25:36 <haraldh> and we will try to reduce them
15:25:49 <haraldh> and introduce a macro for bootstrapping
15:27:03 <pknirsch> right, but we'll have to talk with the maintainers first to see what they think about the idea. any volunteers?
15:27:48 <haraldh> even looking at the first monster graph helps for reducing the package set. e.g. the upper left part is dominated by fonts, which it seems only perl-Tk-Pod requires...
15:28:11 <haraldh> Why it requires those for building the rpm is unknown to me..
15:29:03 <haraldh> ah, no.. it requires them at install time
15:29:05 <jsmith> Maybe just sending out periodic emails to the list saying "Do we really need X?"
15:29:16 * jsmith throws out random ideas
15:29:27 <haraldh> jsmith, well there is wayland coming *joking*
15:29:35 <jsmith> :-)
15:29:36 <pknirsch> bad pun :P
15:30:00 <pknirsch> but ye, if no volunteers i can do that, i should have a few cycles next week to do that
15:30:20 <haraldh> go! :)
15:30:27 <haraldh> make it so
15:30:32 <pknirsch> #action pknirsch contacting the 65 maintainers of the reduced set packages with the %bootstrap flag idea
15:31:54 <pknirsch> we probably won't need any changes to the rpm-config package for that i suspect, but if we do i'll poke Panu and the rpm team about it as well
15:33:00 <pknirsch> alright, so we got some good info and some next steps for this initiative, good!
15:34:38 <pknirsch> next i wanted to quickly bring up the topic of the inter WG coordination.
15:35:04 <pknirsch> #topic Inter WG coordination -> FOSDEM, DevConf?  + https://fedorahosted.org/fesco/ticket/1220 and https://fedorahosted.org/fesco/ticket/1221
15:35:44 <dgilmore> how did you see that working?
15:35:46 <pknirsch> It basically reflects what we've talked about last week already
15:37:07 <pknirsch> well, there are several approaches, but i would like to avoid huge hour long meetings of all committees as those kind of meetings tend to not lead to anything
15:37:40 <pknirsch> so what FESCO is at least requiring now is a brief status update of all WG representatives during the Wednesdays FESCO meeting
15:37:45 <dgilmore> i dont really think it will be effective but its better than nothing
15:37:53 <pknirsch> which i think is a good start, but i'm not sure if it's going to be sufficient
15:38:02 <pknirsch> right
15:38:24 <pknirsch> the question is, how can we improve that? any ideas?
15:39:14 <dgilmore> i think having one person from base liase with the working groups and bring the feedback back here would work well
15:39:23 <pknirsch> one possibility is to have one overall coordinator for the whole effort, but yea, not sure if that would work and who that would be
15:39:31 <dgilmore> sorry i seem to be having internet conectivity issues
15:39:36 <pknirsch> np
15:40:27 <dgilmore> pknirsch: well one per wg
15:40:56 <dgilmore> one to work with  Workstation one with Server one with Cloud and one with env and stacks
15:41:18 <dgilmore> but if we add too many products.
15:42:16 <dgilmore> e will hit scalability issues
15:42:18 <pknirsch> ye, but thats going to be hard to scale and coordinate no matter what i believe
15:42:20 <dgilmore> we
15:42:32 <dgilmore> yeah
15:43:08 <pknirsch> but if we can't get it even for "only" 4 teams, then we have much bigger other issues than scalability
15:43:35 <pknirsch> so i like jrezniks idea of an inter-wg coordinator
15:44:15 <pknirsch> who's main role would be to attend and follow all WGs activities and then report back, probably at the FESCO meeting
15:44:39 <dgilmore> something to keep in mind
15:45:55 <dgilmore> i need to follow the working groups and work with them all. just to make sure we produce whats needed
15:46:13 <dgilmore> but I think somoen else should be the liason
15:46:31 <dgilmore> that way we have two sets of eyes on it
15:47:36 <pknirsch> right, and i think jreznik sort of volunteered to do so
15:47:46 <pknirsch> at least according to the 1220 ticket
15:48:34 <dgilmore> kinda reads that way
15:48:50 <notting> i'm ok with that. would still recommend everyone read the PRDs as they come in just so people are aware of what people are trying to do
15:48:59 <pknirsch> yep
15:49:52 <pknirsch> also, quite a few of the different WGs will be either at FOSDEM late January or at the DevConf in Brno, so i think a get-together there would be really good and iirc Matt is trying to organize that
15:49:55 <dgilmore> notting: indeed
15:50:08 <dgilmore> pknirsch: ill be at both :)
15:50:13 <pknirsch> :)
15:50:44 <dgilmore> definetly having an inperson meeting would be good
15:50:50 <pknirsch> yep
15:51:21 <dgilmore> pknirsch: when we nt thave time, i want to talk about how we will deliver the base product
15:52:31 <notting> pknirsch: i'll be at devconf, not fosdem
15:52:59 <jwb> same
15:53:45 <dgilmore> seems like devconf then
15:53:47 <pknirsch> alright, then lets all go for a beer or two :)
15:53:51 <pknirsch> at devconf
15:55:15 <pknirsch> #action Plan for meeting at DevConf in Brno in Feburary
15:55:40 <pknirsch> #action homework for everyone: Read the PRDs of all other WGs
15:56:35 <pknirsch> oki, thats about it for that topic i think, so lets open up the floor
15:56:39 <pknirsch> #topic Open Floor
15:57:01 <dgilmore> how we ship base
15:58:23 <dgilmore> we have a kickstart for each of workstation, server, and cloud
15:58:29 <dgilmore> I think we need one for base
15:58:45 <dgilmore> that buiolds us a package tree of everything contained in the base set
15:58:48 <pknirsch> hm, yea, i think that would be good
15:58:50 <dgilmore> along with a boot.iso
15:59:40 <notting> why would base be a first-level installable thing, rather than just a repo the other products can pull from?
16:01:00 <haraldh> repo/comps should be enough
16:01:06 <pknirsch> hm, i distinctly remember we had the discussion about installation before
16:01:43 <haraldh> base + anaconda should be installable
16:02:05 <dgilmore> it doesnt need to be installable
16:02:15 <haraldh> or a yum install into a chroot should be bootable .)
16:02:21 <dgilmore> but we need to test that the installer works right
16:02:45 <haraldh> and that we can test easily without anaconda
16:02:46 <dgilmore> and anaconda is in base
16:03:07 <dgilmore> haraldh: we cant when anaconda is the installer
16:03:18 <notting> right, you should be able to *make* it installable for testing ... i was just reading 'deliver/ship' as having a "fedora base" or similar deliverable for end users, which seems overkill
16:03:31 <jwb> agreed
16:04:23 * haraldh has to reread the old logs about anaconda in base...
16:04:45 * masta looks in
16:04:52 <masta> I see I'm late
16:05:53 <haraldh> 15:56:38 <haraldh> Hmm, ok, in the term of "all installation options we support", adding anaconda makes sense
16:05:57 <haraldh> ... meh
16:06:10 <dgilmore> notting: i think we need to make it, dont have to ship it
16:08:37 <jwb> i need to drop off.  have an appt i'm late for
16:08:48 <jwb> i'll review logs when i get back
16:08:49 <pknirsch> np, we're 8 minutes over anyway :)
16:08:52 <pknirsch> thanks jwb !
16:09:56 <pknirsch> dgilmore: would you be willing to do the kickstart for base then?
16:12:20 <dgilmore> pknirsch: sure
16:12:24 <pknirsch> cool
16:12:27 <dgilmore> ive already started on it
16:12:41 <pknirsch> #action dgilmore putting together a kickstart for Base installation testing
16:14:16 <dgilmore> thats all i had
16:14:20 <pknirsch> oki
16:14:24 <pknirsch> thanks dgilmore
16:14:49 <pknirsch> anything else for today? if not, i'll try to recover from my cold i caught the past 3 days on my business trip :/
16:15:11 <haraldh> gute Besserung
16:15:14 <pknirsch> thanks :)
16:15:36 <pknirsch> alright, then thanks everyone for joining today, and have a great weekend!
16:15:39 <pknirsch> #endmeeting