17:00:29 <abadger1999> #startmeeting fpc
17:00:29 <zodbot> Meeting started Thu Nov 14 17:00:29 2013 UTC.  The chair is abadger1999. Information about MeetBot at http://wiki.debian.org/MeetBot.
17:00:29 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
17:00:41 <abadger1999> #topic Roll Call
17:00:53 * kkeithley_ is here
17:01:00 <abadger1999> Who's here?
17:01:30 * RemiFedora is (with his headache)
17:02:05 <abadger1999> tibbs|w, spot, geppetto, Smoother1rOgZ, limburgher: FPC meeting time -- ya'll here?
17:02:10 <abadger1999> #chair RemiFedora
17:02:10 <zodbot> Current chairs: RemiFedora abadger1999
17:02:15 * limburgher yeah.  Ish.
17:02:20 <abadger1999> #chair limburgher
17:02:20 <zodbot> Current chairs: RemiFedora abadger1999 limburgher
17:02:21 * geppetto is here
17:02:25 <abadger1999> #chair geppetto
17:02:25 <zodbot> Current chairs: RemiFedora abadger1999 geppetto limburgher
17:02:34 * spot is here, just needs a few more minutes.
17:02:36 <tibbs|w> Hody.
17:02:40 <abadger1999> #chair spot tibbs|w
17:02:40 <zodbot> Current chairs: RemiFedora abadger1999 geppetto limburgher spot tibbs|w
17:02:41 <tibbs|w> Or whatever.
17:02:56 <geppetto> spot: long time … enjoying NC weather? :)
17:03:38 <abadger1999> That's quorum :-)
17:03:42 <limburgher> Bam.
17:03:44 <abadger1999> #topic #339     software collections in Fedora
17:03:49 <abadger1999> https://fedorahosted.org/fpc/ticket/339
17:03:53 <abadger1999> #chair racor
17:03:53 <zodbot> Current chairs: RemiFedora abadger1999 geppetto limburgher racor spot tibbs|w
17:04:14 <abadger1999> As noed on the list -- I've been busy figuring out the FHS potential changes.
17:04:26 <limburgher> Right.
17:04:27 <abadger1999> Does anyone have anything they'd like to mention on SCLs before I move on
17:04:29 <abadger1999> ?
17:04:48 <RemiFedora> abadger1999, have you see my message to add in the draft
17:04:49 <abadger1999> I'll skip BLAS until spot finishes up his other commitment
17:05:02 <limburgher> Not here.
17:05:21 <abadger1999> RemiFedora: ah -- yeah, the Do-not-extend-other-vendors; instead-create-a-new-scl-with-interscl-deps?
17:05:25 <abadger1999> RemiFedora: made sense.
17:05:32 <abadger1999> Just haven't put it in anywhere.
17:05:37 <RemiFedora> "All SCL packages MUST be part of en existing collection in Fedora/EPEL (cannot extend a collection from another vendor, such as RHSCL or RHDTS)"
17:06:02 <abadger1999> RemiFedora: If you'd like to put it in the draft in an {{admon/question||}} box where you think it makes sense I'll merge it when I get to that section.
17:06:07 <RemiFedora> as Slavek admit it will not work
17:06:31 <abadger1999> RemiFedora: Oh -- I had a conversation that ties into this actually....
17:06:57 * spot is here now.
17:07:07 <abadger1999> Talking with someone who is thinking of using scls, they outlines how they were thinking that sites would want to extend a vendor's scls.
17:07:33 <abadger1999> and then build their app on that hybrid platform.
17:07:40 <abadger1999> And then deploy on either RHEL or fedora.
17:07:44 <RemiFedora> abadger1999, yes, this is a "real" need, which is not covered
17:08:11 <RemiFedora> (and drop nearly all my interest on SCL...)
17:08:24 <abadger1999> To me if we want that to happen, we have to be able to have scls from other vendors in fedora.
17:09:00 <abadger1999> ie: some scl's we build might be fdr-ruby1.9.3  but others might be rht-ruby1.9.3.
17:09:23 <abadger1999> #chair Rathann
17:09:23 <zodbot> Current chairs: Rathann RemiFedora abadger1999 geppetto limburgher racor spot tibbs|w
17:09:28 <Rathann> here, sorry for being late
17:10:00 <abadger1999> RemiFedora: Just something to think about this week.
17:10:06 <abadger1999> Okay, since spot is here,
17:10:14 <abadger1999> #topic #352     BLAS and LAPACK packaging
17:10:19 <abadger1999> https://fedorahosted.org/fpc/ticket/352
17:10:28 <spot> fwiw, that whole ticket gives me a headache.
17:10:59 <spot> But I have no issue with using the mpi approach in the blas/lapack case as well.
17:11:08 <tibbs|w> Ugh, sorry, dealing with some phished accounts.  "I didn't give my password to anyone."  "Sure you didn't."
17:11:46 <limburgher> Heh.
17:12:45 <tibbs|w> So I've always sort of thought the MPI solution was overly complex but the best way to deal with a difficult situation.
17:13:13 <limburgher> Agreed.
17:13:14 <abadger1999> spot: following MPI would mean the libraries remain inerchangeable but it would also mean that the atlas packagers would need to carry local patches, correct?
17:13:45 <abadger1999> (local patches which are apparently being carried by debian already)
17:14:01 <spot> i suppose so.
17:14:10 <tibbs|w> I don't really understand why "let the maintainer choose what to link against" doesn't work here.
17:14:31 <Rathann> FWIW openblas upstream was receptive to the idea of adding lapack/blas compatibility sonames
17:15:14 <tibbs|w> And at some point as a distro it may be just smarter to pick the most receptive implementation and work with it.
17:15:24 <tibbs|w> We don't have switchable kerberos, for example.
17:15:48 <Rathann> for now openblas is exclusively x86 and mips
17:16:11 <Rathann> true, but we do have interchangeable mpi
17:16:31 <Rathann> hm, actually that statement is wrong
17:17:10 <Rathann> but then again alternative blas and lapack implementations ARE ABI compatible
17:17:25 <abadger1999> So... I think I'd be okay with either following MPI's example or maintainer chooses at buildtime.
17:17:28 <Rathann> as long as they provide the same SONAMEs
17:17:35 <abadger1999> But I'm not really sure which is the better choice.
17:17:55 <abadger1999> Anyone understand this well enough to make a case for one or the other?
17:18:20 <limburgher> Which one has the shinier Powerpoint deck and more attractive salespeople?
17:18:30 <Rathann> the latter could theoretically lead to cases where an application could link to one blas implementation directly and a dependent library to another
17:18:32 <limburgher> Oh, wait, sorry, this is FLOSS.
17:18:53 <abadger1999> limburgher: right -- here it would be "shortest (irc) name wins" :-)
17:18:54 <limburgher> Rathann:  With or possibly without catastrophic results.
17:19:29 <limburgher> abadger1999: But spot isn't a blas or lapack implementation.  Or is he?  (eyebrow)
17:19:41 <spot> i sure hope not.
17:19:51 * limburgher prepares Turing test for spot
17:20:02 <spot> I AM NOT A REPLICANT
17:20:06 * spot leaps out window
17:20:15 <geppetto> abadger1999: There's a bit more to it now than just shortest name :p
17:20:17 <limburgher> That's exactly what a replicant would say.
17:20:46 <abadger1999> Rathann: okay.  So the safer path is to use environment modules?
17:21:01 <limburgher> geppetto:  And I'm not sure any of us are in a position to fully evaluate.
17:22:02 <geppetto> yeh, my assumptions would be that tibbs|w is right and we should stop shipping N things.
17:22:09 <Rathann> abadger1999: I would say yes
17:22:30 <spot> abadger1999: yeah, i think env modules makes the most sense here
17:23:02 <geppetto> ok, not to put undue stress on spot, but I'm happy to +1 whatever he thinks :)
17:23:22 <abadger1999> Proposal: FPC favors using environment modules for this as it seems to lead to less problems for applications which might want to link to one blas implementation but a dependent library is linked to another.  If someone would present us with a draft based on that we'll be happy to review and voe on it.
17:23:49 <abadger1999> s/seems to lead to less problems for/seems to avoid the problem of/
17:24:08 <spot> +1
17:24:10 <abadger1999> +1
17:24:12 <limburgher> +1
17:24:14 <RemiFedora> +1
17:24:15 <geppetto> +1
17:24:35 <Rathann> +1
17:25:01 <abadger1999> #info Approved (+1:6, 0:0, -1:0) FPC favors using environment modules for this as it seems to avoid the problem of applications which might want to link to one blas implementation but a dependent library is linked to another.  If someone would present us with a draft based on that we'll be happy to review and  vote on it.
17:25:24 <abadger1999> #topic #355     How to package noarch packages which require a binary dependency which doesn't build on all archs?
17:25:29 <abadger1999> https://fedorahosted.org/fpc/ticket/355
17:26:09 <geppetto> https://fedoraproject.org/wiki/User:Toshio/Noarch_with_unported_dependencies is the draft
17:26:40 <tibbs|w> Ugh, "just submit the builds" is so hacky.
17:26:56 <tibbs|w> Koji does have a way to deal with this, but it isn't appropriate to even talk about that in the guidelines.
17:26:56 <abadger1999> I think I captured everything we talked about in last week's meetings but I'm definitely open to  changes.
17:27:03 <abadger1999> tibbs|w: yeah, that bothered me immensely.
17:27:41 <abadger1999> tibbs|w: oh, what is that way?
17:27:46 <spot> it seems like this might be a place where arch hinting to koji would be advantageous
17:27:48 * abadger1999 is almost afraid to ask...
17:28:18 <spot> if we could somehow tell koji "build me only on this arch" or "never try to build me on this arch"
17:28:23 <limburgher> What, pray tell, is arch hinting?
17:28:26 <tibbs|w> I'm pretty sure it has a way of constraining building of a package to certain arches.
17:28:29 <spot> it doesn't exist. :)
17:28:31 <spot> afaik
17:28:40 <abadger1999> <nod>
17:28:42 <tibbs|w> But it has been years since I looked at it.
17:28:43 <spot> but it seems like we need it to.
17:28:43 <geppetto> limburgher: some way to hint on what arch a noarch package should be built, I assume.
17:28:44 <limburgher> Or you could just arch the package.
17:29:00 <tibbs|w> But arching the package is the cleanest way without getting involved in that, I think.
17:29:02 <limburgher> Seems like a legit reason to do so, and less hacky.
17:29:04 <geppetto> tibbs|w: Some way that would work with a mass rebuild?
17:29:15 <tibbs|w> I have forgotten most of what I knew about koji.
17:29:34 <spot> I don't think we should advocate trying a hundred rebuilds until it lands on a compatible architecture.
17:29:47 <spot> the other option would be a metapackage hack
17:29:51 <abadger1999> the other workaround would be to tell people to create stub packages on the arches that lack something important.
17:30:00 <limburgher> spot: god no.
17:30:05 <spot> e.g. arm-only.arm.rpm
17:30:08 <geppetto> spot: Still doesn't help the build case, right?
17:30:13 <racor> Buildarch: <arch>  + ExclusiveArch: <subset of arch> + Arch: noarch?
17:30:17 <spot> oh, right.
17:30:30 <abadger1999> spot: with the new builder setup, as long as it builds on arm we're okay ;-)
17:31:03 <spot> i think the second paragraph under BuildRequires in the draft needs to be removed.
17:31:29 <spot> and we should just present the latter "choice".
17:31:37 <geppetto> Hmm … if they have an arched top level pacakge, with "all" of the real package behind a noarch sub-package … that would work, right? In that the noarch part would only be built on the arched machines?
17:32:19 <abadger1999> hmm... but then if it doesn't build on ppc... you have fedora packages that will have to be arched, won't you?  Simply because we keep ppc builders for epel building?
17:32:31 <abadger1999> actually nevermind.
17:32:52 <abadger1999> koji would have to keep ppc out because we don't have a ppc fedora tree.
17:33:35 <abadger1999> geppetto: yep, that would work.  Would that mean a different package name though?
17:33:43 <tibbs|w> I think the koji thing I'm thinking about is "koji set-pkg-arches" but I think that only does something related, not entirely useful for this case.
17:34:02 <geppetto> abadger1999: No, you could put all the normal package data in the top level metapackage, and just name the bit containing the real stuff foo-data.
17:34:48 <abadger1999> geppetto: ah... so on all architectures, you'd have an empty, arch'ed metapackage?
17:34:49 <tibbs|w> But it's true that koji could eventually learn how to do this.  In the meantime, as is usual for us when the tools don't do exactly what packagers want, is to document the workaround.
17:34:50 <geppetto> abadger1999: Nevermind … that wouldn't work as the bad arches can install the result, so meh.
17:35:10 <geppetto> abadger1999: So you need the noarch part to have the findable name :(
17:35:23 <abadger1999> geppetto: ExclusiveArch would mean the packages don't show up on the bad arches, though, right?
17:35:26 <tibbs|w> And I think the cleanest workaround here is to say "just use arch".  Yum can handle going noarch to arch-full and back these days, right?
17:35:32 <abadger1999> (we won't put them in the repo)
17:35:46 <geppetto> abadger1999: Yeh
17:36:23 <geppetto> abadger1999: The only problem is the name … so instead of having foo you'd have the noarch bit be "foo-something"
17:36:34 * abadger1999 notes that all nodejs packages are currently using the try and resubmit if failed option.
17:36:52 <geppetto> abadger1999: For most cases I'm not sure that's worth the cost of just saying "use arches".
17:37:03 <abadger1999> and there's probably other classes of packages that do the same.
17:37:11 <tibbs|w> Has anyone actually asked infra or the koji devs about the possibility of adding the needed functionality?
17:37:13 <spot> dgilmore: any interest in trying to add a hack to koji to allow us to hint to koji where a package can be built? this could live in git or in the spec...
17:37:15 <geppetto> Failing mass rebuilds randomly seems like a really bad idea.
17:37:24 <tibbs|w> I guess that answers my question.
17:37:35 <geppetto> And having N copies of the nodejs stuff doesn't seem like a big enough hit to do anything about it.
17:37:51 <abadger1999> geppetto: +1  yeah, that's why I dislike this but... it is current practice.
17:37:57 * abadger1999 digs up the nodejs rationale on list
17:39:35 <geppetto> spot: I think the main problem is that you'd need something in the specfiles … which means you need something added to rpm/rpmbuild, and then something added to koji to make use of it.
17:40:00 <spot> geppetto: well, if we accept that all builds should be done through something like fedpkg
17:40:13 <spot> or the "koji" cli
17:40:26 <spot> then it should be possible to have a file in git that controls the hinting
17:40:30 <abadger1999> actually the rationale is here: https://fedorahosted.org/fpc/ticket/355#comment:4  The para starting "ARCH ALL THE THINGS"
17:41:08 <abadger1999> <nod>  ExcludeArch/ExclusiveArch could be the hinting too if it needed to be in the spec file, right?
17:41:16 <spot> abadger1999: in theory.
17:41:17 <tibbs|w> koji knows other things about packages.  These get set by koji admins.  It's not the cleanest way, but it would be the simplest.
17:41:20 <abadger1999> It's just that koji would have to be able to get that information out.
17:41:28 <geppetto> abadger1999: Both arguments are known, and meh.
17:41:57 <geppetto> abadger1999: Who cares if there are 3 copies of some small amount of JS … and users being a little confused about why it's not noarch is meh.
17:42:13 <abadger1999> <nod>
17:42:56 <geppetto> So I'm currently +1 on spots proposal (abadger1999 proposal but remove the "just resubmit" bits)
17:43:06 <tibbs|w> I agree.
17:43:19 <spot> +1
17:43:19 <geppetto> If someone want to propose something that uses metapackages, I might +1 that depending on how ugly it is.
17:43:20 <tibbs|w> Maybe koji can grow some extra functionality later, and we can revisit.
17:43:46 <geppetto> And, yeh, if someone can change koji we can change it.
17:44:00 <abadger1999> I can be +1 and I'll ask the people on the ticket to revisit/attempt to persuade us of a better option.
17:44:33 <abadger1999> (mentioning that failing mas rebuilds randomly is one of the primary problems we see with fail-and-resubmit)
17:44:42 <Rathann> nodejs is currently arch'd from what I see in the spec
17:44:50 <tibbs|w> I guess as long as it's explicit that we're documenting a workaround, so nobody gets the idea that we actually want it that way.
17:45:04 <abadger1999> Counting tibbs|w's I agree we're at +4
17:45:17 <tibbs|w> Yes, that's a +1 from me.
17:45:32 <abadger1999> Rathann: nodejs itself or the packages that depend on nodejs?  (it's the packages that depend on nodejs that are noarch but won't build on X)
17:45:46 <abadger1999> note... I think with nodejs the problems are all in EPEL, not Fedora.
17:45:50 <abadger1999> iie: ppc64.
17:45:55 <Rathann> nodejs itself
17:46:04 <limburgher> +1
17:46:32 <abadger1999> That's +5.  Rathann, care to voe for the record?
17:46:35 <limburgher> Sorry, in flaky mode again today.  Busy.
17:47:25 <Rathann> +1
17:47:32 <RemiFedora> +1  (sorry 2 meeting at the same time)
17:47:53 <abadger1999> #info Noarch_with_unported_dependencies minus the paragraph specifying the fail-and-resubmit option passed (+1:7, 0:0, -1:0)
17:48:08 <abadger1999> #topic #357     time-api prior to openJDK8
17:48:13 <abadger1999> This one passed in ticket.
17:48:20 <abadger1999> I've updated the guidelines.
17:48:34 <abadger1999> for the virtual provide for timeapi...
17:48:51 <abadger1999> something like bundled(openjdk8-time-api) ?
17:49:23 <abadger1999> (ie: PackageItComesFrom-NameForAPI)
17:49:43 <Rathann> makes sense
17:50:03 <geppetto> sure
17:50:31 <abadger1999> And -- do we want to be explicit about approving the tickets that fall under this?
17:50:39 <tibbs|w> I think so.
17:51:04 <tibbs|w> It should be pretty rare.
17:51:09 <abadger1999> <nod>
17:51:14 <abadger1999> okay, then real quick let's do:
17:51:31 <abadger1999> #topic 365 New virtual Provides for javax time-api
17:51:36 <abadger1999> https://fedorahosted.org/fpc/ticket/365
17:51:51 <spot> +1
17:51:55 <abadger1999> This is for javax-timeapi itself.
17:51:57 <abadger1999> +1
17:52:04 <RemiFedora> +1
17:52:06 <geppetto> +1
17:52:07 <tibbs|w> That's a lot of virtual provides.
17:52:08 <Rathann> +1
17:52:09 <limburgher> +1
17:52:14 <tibbs|w> +1
17:53:09 <abadger1999> #info virtual provides for java8's time api granted (+1:7, 0:0, -1:0)
17:53:59 <abadger1999> #topic 366 Should we hand enabling third party repositories back to fesco?
17:54:03 <abadger1999> https://fedorahosted.org/fpc/ticket/366
17:54:46 <tibbs|w> I'm all confused about why this is coming up now.
17:55:06 <RemiFedora> I don't understand the need
17:55:07 <abadger1999> So the three products are trying to figure out their boundaries.
17:55:21 <tibbs|w> FESCo asked us to write guidelines preventing this thing; we did.
17:55:33 <abadger1999> and one of the things they've asked is "Can we enable third party repos?"
17:55:46 <tibbs|w> If now they want us to remove those guidelines, they need only ask us to change the guidelines.
17:55:47 <sgallagh> It's not just about the products, it's also about COPRs
17:56:02 <tibbs|w> Anyone remember the initial issue?
17:56:13 <abadger1999> sgallagh: only if we ship packages in fedora that point to coprs.
17:56:25 <abadger1999> tibbs|w: Yeah, the original fesco ticket has one example
17:56:32 <abadger1999> (well, two examples)
17:56:37 <abadger1999> https://fedorahosted.org/fesco/ticket/671
17:56:43 <geppetto> tibbs|w: Legal I think was the main one.
17:56:56 <tibbs|w> That's not actually the original issue.
17:57:01 <tibbs|w> It's much older than that.
17:57:03 <spot> the legal issue can't be waived, not unless US law has changed when I wasn't looking.
17:57:05 <abadger1999> tibbs|w: that's true.
17:57:13 <abadger1999> I think there was one from 2006?
17:57:19 <spot> but i don't think that's the issue here.
17:57:36 <racor> tibbs|w: In the early days there were people who wanted to ship binaries and/or to packages adding *repos.
17:57:40 <abadger1999> https://www.redhat.com/archives/fedora-packaging/2006-October/msg00132.html
17:57:44 <tibbs|w> Basically someone submitted a package that instead of actually containing the software for which it was named, it just contained a repo definition, and the guy pushed the software via the repo.
17:57:55 <limburgher> Oh, like Google Chrome?
17:58:19 <racor> tibbs|w: This matches my memory.
17:58:37 <tibbs|w> But this was back in the early days, maybe even before the Core-Extras merge.
17:58:45 <racor> limburgher: No, more like rpmfusion-release.
17:58:58 <limburgher> racor:  Potato, Kartoffel.
17:59:25 <tibbs|w> In any case, Legal will always override and we obviously can't allow the addition of repos we don't "control" in some sense.
17:59:37 <limburgher> This------^
17:59:55 <geppetto> tibbs|w: Right, hopefully that's obvious to people that we can't allow what the clement thing did.
17:59:56 <tibbs|w> But if Fedora is going to have multiple official repos, then Fedora is going to need to ship repo definitions.
18:00:04 <tibbs|w> Yes, clement.  That was it.
18:00:05 <racor> limburgher: rpmfusion-release as part of Fedora package, not as external package
18:00:11 <limburgher> So I don't see how we can allow it, it could be an end run around Legal and the guidelines.
18:01:02 <geppetto> limburgher: Well on the other side there's the idea that not everything will be in "fedora", and not all of those repos. will be installed/enabled by default.
18:01:19 <abadger1999> limburgher: I think the (partial) idea is no, it can't be an end run around legal.  yes, it could be an end run around the guidelines.
18:01:24 <tibbs|w> http://www.spinics.net/lists/fedora-extras/msg30902.html
18:01:26 <tibbs|w> 2006.  Wow.
18:01:28 <geppetto> So I can see how we'd be ok with installing/enabling one of those Fedora product repos. via. package install.
18:01:58 <geppetto> tibbs|w: yeh, abadger1999 pointed to the fedora-packaging email above.
18:02:24 <racor> limburgher: Legal issue were only part of the issue, others were "rpmdb consistency" and installers not using rpm to install system-wide packages
18:03:42 <limburgher> racor:  Right, that could get very, very ugly.
18:04:17 <tibbs|w> In any case, anything that isn't an absolute prohibition is going to have to be written carefully.
18:04:20 <geppetto> abadger1999: I would say in general to vote on any proposal here we'd need to know what the repos. will be used for, what their rules will be etc. etc.
18:05:14 <RemiFedora> I would be against any "external" repo provided in fedora configuration/package.
18:05:18 <abadger1999> geppetto: <nod>  Do we want to be in charge of seeing that repos are vetted or see if fesco wants to do so?
18:05:38 <geppetto> abadger1999: I assume we'd be much better equipped than fesco to do that.
18:06:14 <limburgher> But how often do we vet the repos?  Does this include vetting the procedures by which things enter those repos?
18:06:21 <racor> RemiFedora: +1, I am opposed to any exception.
18:06:24 <geppetto> limburgher: once, and yes.
18:06:51 <RemiFedora> If stuff in those repo is suitable to fedora, it have to be included in fedora
18:07:03 <abadger1999> spot: Do you recall enough of this from before that you can paraphrase what Legal's position might be?
18:07:10 <geppetto> abadger1999: But I guess I'm happy to have fesco to do the final ACK, if they'd prefer to get input from FPC, security, legal, etc.
18:07:30 <spot> We cannot point via .repo file to any repository which contains packages not vetted by Fedora Legal.
18:07:32 <abadger1999> like, fedorapeople would be allowed since we already have to vette them but jpackage/other repos are too risky?
18:08:02 <geppetto> abadger1999: Pretty sure nobody has legally vetted fedorapeople
18:08:08 <abadger1999> spot: k.  and would that be an ongoing vetting?
18:08:25 <spot> abadger1999: as much as the prospect thrills me, yes.
18:08:41 <limburgher> Nothing stopping me from pushing adobe reader RPMS to my fp.o repo, though obviously I'd have to take it down if discovered.
18:08:53 <abadger1999> spot: and... where would fedorapeople fall in that?  we'd have to do ongoing, additional vetting?
18:08:54 <limburgher> Nothing but revulsion, that is.
18:09:08 <geppetto> limburgher: right, I guess you might get banned from fp.o if you did get found out too
18:09:22 <spot> i really really do not want official packages pointing to fedorapeople.
18:09:33 <spot> that is a never ending legal review process for me.
18:10:03 <geppetto> spot: I think the idea is less about fp.o … and more about specific copr repos.
18:10:17 <spot> is this all an end-run around the guidelines?
18:10:26 <limburgher> geppetto: I would hope so.
18:10:27 <spot> if so, i'd rather work on adding exceptions.
18:10:35 <geppetto> spot: I'd assume not.
18:10:49 * spot isn't sure why we want to point to copr repos from Fedora packages.
18:10:58 * RemiFedora too
18:11:06 <abadger1999> jwb, sgallagh: I was kinda assuming this was a prtial end run around the guidelines. Could you tell us if there's other aspects to this?
18:11:30 <tibbs|w> I think the bottom line is that if the packages in the repo didn't go all the way through the standard review process, we don't want a repo config file in the distro at all.
18:11:47 * abadger1999 notes that he was thinking repos.fedorapeople.org when talking about fedorapeople... and copr repos would be he next generation of that.
18:12:18 <jwb> abadger1999, i haven't been paying attention to whatever is going on in this channel.  small summary?
18:12:42 <abadger1999> jwb: <nod> we're talking about the third party repos enabling
18:12:43 <tibbs|w> Also, obligatory note that "copr" is pretty close to one of the most cringe-inducing names ever.
18:13:06 <abadger1999> tibbs|w: yes, yes it is.
18:13:29 <limburgher> Oh, as if "plague" was awesome. :)
18:13:42 <abadger1999> jwb: there's a bunch of legal issues which may make it unresolvable but we're also trying to understand what benefit we get from it.
18:13:52 <spot> tibbs|w: i wouldn't mind if it was renamed to "vidal", personally.
18:14:31 <jwb> abadger1999, i don't think anyone is trying to do an end run around any guidelines.  the WG just wants to know if it can either make 3rd party repos discoverable or enabled by default so if someone searches for e.g. chromium, it either installs it or pops something up that says "this is available in this 3rd party repo.  do you want to enable that repo?"
18:14:37 <limburgher> spot: <sigh>
18:14:49 <jwb> repeat for other examples
18:15:03 <jwb> i think "steam" was the other one cschalle used in his example
18:15:13 <abadger1999> jwb: well -- the question is -- if it's not an end run around the guidelines, then chromium should just be packaged for fedora.
18:15:38 <jwb> abadger1999, it would help me if you would be clear on what you mean by "guidelines"
18:15:43 <jwb> which guideline?
18:15:48 <jwb> the %docdir one?
18:15:48 <abadger1999> Packaging guidelines.
18:15:52 <abadger1999> in general
18:15:57 <abadger1999> no.
18:16:15 <geppetto> wow … chomium and steam … seems a pretty simple -1000, next.
18:16:29 <spot> i think if we had a repo that had only things which were not legally questionable (but not okay for Fedora), it would not be legally forbidden to query that repo, but once you're doing that, you're better off just taking those distasteful things into Fedora proper.
18:16:33 <abadger1999> jwb: We're more talking about the packages in the external repo.
18:17:01 * spot is actually seriously considering asking for chromium to get bundling exceptions under the same logic as firefox.
18:17:10 <limburgher> *cough*rpmfusionandbedonewithit*cough
18:17:37 <jwb> what meeting is this?
18:17:37 <tibbs|w> So, would it be acceptable for someone to package an app that knows how to browse the available coprs and simply generates repo files for the ones the user selects?
18:17:46 <jwb> oh FPC.
18:17:47 <spot> jwb: Packaging Committee
18:18:06 <abadger1999> jwb: and whether the idea is to get those accessible to fedora users without making the packages themselves pass the packaging guidelines.
18:18:10 <limburgher> tibbs|w:  Oh.  God.
18:18:15 <geppetto> tibbs|w: coprs automatically generate the .repo files … the app. would just need to download/install them.
18:18:25 <tibbs|w> Either way.
18:18:25 <limburgher> Lacrimosa.
18:18:27 <jwb> abadger1999, people, for whatever reason, don't always want to put their packages in fedora.  that doesn't mean they aren't legally allowable
18:19:20 <abadger1999> jwb: <nod>  I'm not looking at the legal problems right now... Those are pretty steep but separate.
18:19:47 <abadger1999> jwb: I'm currently just trying to get answered, what the benefit is.
18:19:54 <jwb> ok, then if you want to call "packager desire to not expend the effort to get something into fedora proper" an end-around, then yeah
18:20:09 <jwb> the benefit would be to allow people to install stuff not in fedora.
18:20:11 <jwb> clearly.
18:20:35 <jwb> or, more easily and discoverably install taht
18:20:36 <geppetto> yeh, I mean that's pretty much the definition of end-run-around
18:20:41 <abadger1999> <nod>
18:20:50 <abadger1999> jwb: thanks.
18:21:30 <jwb> geppetto, eh.  "choosing not to play" isn't really an end-run-around.  it's not like non-compliant packages are getting shoved into fedora
18:22:15 <jwb> however, i don't care what you call it.  the benefit sought is more easily installed packages in 3rd party repos
18:22:23 <geppetto> jwb: It will result in that though: Fedora packaging is hard, let's just dump random crap elsewhere and shove a repo. file onto the "Fedora users"
18:22:32 <RemiFedora> "packager desire to not expend the effort to get something into fedora proper" => they don't want to be in fedora... so why should we care of them ?
18:23:02 <jwb> geppetto, that's possible.  whether this is allowed or not doesn't make that problem any less possible.
18:23:36 * spot thinks that 99% of the stuff in this category is legally impossible to point to, and the two items which might be could perhaps be special cased.
18:24:01 <spot> if this boils down to "chromium" and "steam", the former could request exceptions for its bundling sins
18:24:13 <spot> and the latter could exist in a fedora-nonfree repo.
18:24:21 <jwb> those were the examples.  i don't think they're the only examples
18:24:23 <geppetto> jwb: If I can install my thing with yum commands, even if it takes N commands as I have to install XYZ to get a repo. and then FOO, that makes it way more viable than telling someone to download XYZ from a random webpage.
18:24:24 <spot> which i don't think the community wants, but could be done.
18:25:28 <spot> from a legal perspective, if there was a repo that has no legal issues, we could choose to point to it/search through it.
18:25:48 <spot> i happen to think thats not a great idea, but the answer remains.
18:25:51 <abadger1999> Proposal: If FESCo would like to allow pointing to repos that don't have Official Fedora Content they can let us know and have someone propose a guideline draft that we can critique and vote on.  However, after talking with Fedora Legal, the requirements for us to be able to point to repositories outside of our control may be so costly that in practice there's very few repositories that we can actually point to.
18:25:57 <jwb> so from a technical perspective, we already allow .repo files in %docdir, right?
18:26:19 <spot> jwb: as long as nothing in that corresponding .repo has legal issues.
18:26:24 <jwb> who reviews that?
18:26:29 * spot is supposed to
18:26:36 <spot> i haven't seen any of those files though.
18:27:24 <jwb> so with those two givens, it would be possible for an application (app-installer) to take a query (say for chromium), enable repos in %docdir, search the metadata, and prompt the user if they want to enable the repo to install, correct?
18:28:11 <spot> there are no legal issues to that, assuming that the repo files meet the aforementioned legal standard.
18:28:45 <jwb> it goes against the "can't ship repo files" aspect somewhat, because they have to be enabled to get the metadata
18:29:02 <spot> yes, but i can't speak to that aspect.
18:29:11 * spot is trying to be "legal guy" only.
18:29:23 <jwb> ok, the question is posed to FPC in general
18:29:55 <abadger1999> spot: and the potential Legal issue is that the repositories have to be vetted in an ongoing manner (which we currently don't do only because there are none that we know of?  or would that be a wholly new requirement?)
18:30:06 <jwb> actually, it was posed to FESCo.  i lost how we're talking about it here, but i'm trying to roll with that
18:30:52 <abadger1999> jwb: fesco doesn't actually have the policy in question -- they sent the basic idea to FPC to write up a guideline for it.
18:30:57 <limburgher> abadger1999:  And it could be labour-intensive, and/or difficult to automate.
18:31:15 <geppetto> I would currently assume that any .repo files in %docdir have roughly the same level of oversight as random .repo files on the internet … so assume we wouldn't allow an application we ship to enable .repo files from the internet, then I'd assume we wouldn't want it to install files from %docdir either.
18:31:52 <limburgher> geppetto: <nod>
18:32:04 <RemiFedora> geppetto, we have a guidelines which says, nothing in %doc whould be required at runtime ;)
18:32:18 <geppetto> But that's a lot of assumptions … if someone proposed a policy exception for some application with a fixed list of .repo files in %docdir it wanted to install, I might be convinced to +1 it.
18:33:07 <geppetto> RemiFedora: right, that too … although I doubt people using this would also be playing with tsflags to not installing docs :)
18:35:00 * RemiFedora thinks, we already provides external repo configuration...
18:35:22 <RemiFedora> not RPM repo, but sources repo (cpan, pear, ...)
18:35:30 <abadger1999> geppetto: eh -- it's implementation, but what I'd do is have the third party repo files be placed somewhere new like %{_datadir}/extra-repos/{yum,smart}  then Feodra Legal would know what files to vette and the app installer wouldn't depend on things in %{_docdir}
18:36:37 <spot> guys, as fun as this is, i need to go get lunch.
18:36:49 <jwb> i'm in 2 other meetings right now.  can someone ping me with whatever the FPC is deciding/recommending?
18:37:03 <abadger1999> spot: Thanks -- do you have an answer for my legal clarificatoin before you go?
18:37:14 <abadger1999> also..
18:37:38 <abadger1999> Does anybody have criticism on my proposal/an alternate proposal they'd like to make?
18:37:41 <spot> i would assume that any .repo files coming in as %doc are being cleared as pointing to a "legally okay" repo.
18:37:58 <spot> this hasn't been done to date because i am unaware of any such .repo files.
18:38:03 <geppetto> abadger1999: At that point might as well put them in /etc/yum.repos.d but with enalbed=false
18:38:35 <spot> where these .repo files live on the filesystem is legally irrelevant.
18:39:06 <abadger1999> geppetto: <nod> I've toyed with that idea as well... I seem to recall someone thought --enablerepo='*' was a reason to leave them out of yum.repos.d but I'm flexible.
18:39:09 <geppetto> abadger1999: Your proposal could be reworded to the generic: As always, if someone proposes some policy we'll look at it ;)
18:39:54 <geppetto> abadger1999: Anyone doing --enablerepo='*' will have a bad time already.
18:40:05 <geppetto> But, FWIW I'm +1 on abadger1999 proposal
18:40:06 <limburgher> abadger1999:  I'm sort of . . .meh. . .nearly +1. . .on your proposal, but I'd rather just say No, Don't Do It.
18:40:08 <abadger1999> geppetto: true.  but since this policy is made in collaboration with fesco... we really need to have confirmation from fesco that they want us to change it anyway.
18:40:17 <abadger1999> okay...
18:40:23 <abadger1999> How about an additional line
18:41:30 <abadger1999> Given the costs to benefits, FPC also recommends that third party repos not be enabled.  Alternatives like a fedora-non-free repo would be a better avenue to explore.
18:42:10 <limburgher> I don't like that.  That already exists.
18:42:24 * abadger1999 crumples that up and throws it in the trash.
18:42:30 <limburgher> Or, rather, great line, sans the second sentence.
18:42:37 <abadger1999> ah okay.
18:42:57 <limburgher> So now it's all wrinkly.  Sorry.
18:43:22 <abadger1999> So proposal:    If FESCo would like to allow pointing to repos that don't have Official Fedora Content they can let us know and have someone propose a guideline draft that we can critique and vote on.  However, after talking with Fedora Legal, the requirements for us to be able to point to repositories outside of our control may be so costly that in practice there's very few repositories that we can actually point to. Given the costs to
18:43:24 <abadger1999> benefits, FPC also recommends that third party repos not be enabled.
18:43:45 <limburgher> I can +1 that.
18:44:12 <abadger1999> +1
18:45:27 <geppetto> +1
18:45:57 <abadger1999> tibbs|w, RemiFedora, racor, Rathann: Votes?
18:46:17 <RemiFedora> 0
18:47:14 <abadger1999> RemiFedora: I'll note that we do have to tell fesco something.... is there something that we could add to this so that you'd be +1?
18:47:44 <RemiFedora> no, sorry, I still don't understand the need, and why we could point to external repo
18:47:49 <abadger1999> k
18:48:16 <RemiFedora> for me if stuff is acceptable for fedora, it should go in fedora. End.
18:50:18 <tibbs|w> Sorry, the auditor walked in and pulled me away.
18:50:28 <tibbs|w> I dislike auditors.
18:50:42 <tibbs|w> +1 to abadger1999's proposal.
18:51:07 <jwb> RemiFedora, i think that stance is going to limit future adoption at some point
18:51:22 <jwb> but that is just my opinion and doesn't need to be debated here
18:51:22 <abadger1999> Okay, we're at +4.
18:52:22 <RemiFedora> jwb unanimity is not mandatory ;)
18:52:29 <abadger1999> I think that spot would +1 that proposal in the ticket so I'll risk asking for one more vote there.
18:53:16 <abadger1999> If it doesn't get +5 by next week's fesco meeting, I'll send fesco the proposal and the vote count.
18:55:02 <RemiFedora> abadger1999, near 2h... probably we should end soon ?
18:55:04 <abadger1999> #info Proposal (See ticket https://fedorahosted.org/fpc/ticket/366#comment:1) Is currently (+1:4, 0:1, -1:0) will ask for more votes in ticket.
18:55:07 <abadger1999> Yep.
18:55:12 <abadger1999> #topic Open Floor
18:55:14 <kkeithley_> #363? I'd like to get resolution please? Unlikely today --- in the ticket, or next week? Thanks
18:55:30 * leamas sighs
18:56:22 <abadger1999> #topic #363 exception for bundled library libntirpc in nfs-ganesha
18:56:32 <abadger1999> https://fedorahosted.org/fpc/ticket/363
18:56:41 <abadger1999> We have just enough to make quorum here.
18:56:54 <geppetto> Given this bit "The long term plan is to package libntirpc once the interfaces stabilize"
18:57:11 <jwb> abadger1999, FYI i added spot on CC to the fesco ticket for the legal aspect
18:57:20 <geppetto> I'd be mostly happy to +1, if there was any kind of fixed timeline.
18:57:48 <abadger1999> So if there were a timeframe for libntirpc to be released standalone I would would definitely be able to vote +1 on this.
18:57:55 <geppetto> kkeithley_: So any known timeline?
18:57:59 <kkeithley_> I can inquire about a possible timeline.
18:58:19 <abadger1999> jwb: <nod> -- note that spot, tibbs, and I should be on the cc of all tickets (just like nirik setup for the fesco liasons at last meeting)
18:58:23 <limburgher> That would certainly be helpful.
18:58:48 <RemiFedora> I'm also +1 for a temporary exception until can be packaged as shared, with a stable API (this is obvioously a fork)
18:58:58 <nirik> abadger1999: in fesco trac? or fpc?
18:59:12 <abadger1999> kkeithley_: k.  If there's a timeline I think we could pass a temporary exception next meeting without problem.
18:59:55 <abadger1999> kkeithley_: If there's no fixed (estimated, good faith) timeline then it'll be a much longer discussion and we have a never ending chain of scl issues to sort out :-/
19:00:05 <abadger1999> nirik:  fesco
19:00:19 <nirik> abadger1999: I can add you all if you like... @fedoraproject.org ?
19:00:27 <kkeithley_> Right. FWIW, I got stuck doing the packaging "as a favor" so I have no particular knowledge about it.
19:00:36 <abadger1999> nirik: You could -- but I was saying thatI was pretty sure we already were.
19:00:46 <nirik> ah,didn't see. ok.
19:00:47 <kkeithley_> But I've made the inquiry. Will update the ticket with their response
19:00:59 <abadger1999> kkeithley_: thanks for understanding and querying :-)
19:01:01 <kkeithley_> their == developers
19:01:23 <abadger1999> #topic Open Floor
19:01:26 <abadger1999> Anything else?
19:01:38 <RemiFedora> FYI, FPC should appears soon on https://fedoraproject.org/wiki/F20_anniversary_tshirt
19:01:38 <leamas> #362: Do we all agree on that lpf-* packages should not be allowed in fedora? Seems to be the common view, (and also mine). I vote in the ticket woudl be nice...
19:02:15 <abadger1999> #topic #362 lpf should not be allowed in Fedora
19:02:20 <abadger1999> https://fedorahosted.org/fpc/ticket/362
19:02:32 <kkeithley_> That was quick. ganesha/ntiprc says about a year.
19:02:47 <kkeithley_> correction, sometime next year
19:03:14 <geppetto> leamas: Do you disagree with abadger1999 comment?
19:03:21 <abadger1999> leamas: shoot... I was kinda waiting on spot.
19:03:23 <geppetto> leamas: If so … in what way?
19:03:55 <limburgher> abadger1999:  RIght.
19:03:58 <leamas> Which commetn?
19:03:58 <abadger1999> But yeah, from everything I've read, I agree that lpf itself is valid and lpf-* should go to rpmfusion
19:04:15 <abadger1999> leamas: https://fedorahosted.org/fpc/ticket/362#comment:4  (I'm toshio)
19:04:15 <geppetto> leamas: https://fedorahosted.org/fpc/ticket/362#comment:4
19:04:35 <leamas> Sorry... No, I agree :)
19:04:36 <RemiFedora> abadger1999, I agree, but I don't know if "lpf" alone make sense...
19:04:53 <abadger1999> RemiFedora: well, it's the same sort of thing as dkms.
19:05:23 <RemiFedora> yep
19:05:30 <abadger1999> it's a framework that makes building packages easier.
19:06:02 <abadger1999> Shall we vote and let spot correct us if he sees anything wrong?
19:06:24 <geppetto> Sure, I'm happy to +1 lpf is good lpf-* is bad.
19:06:28 <abadger1999> Proposal: lpf is okay for Fedora, lpf-* packages should move to a more appropriate repo
19:06:30 <abadger1999> +1
19:06:32 <geppetto> +1
19:06:45 <RemiFedora> +1
19:07:20 <abadger1999> Everyone else might be going on to other things.  I'll call for more votes in ticket
19:07:40 <leamas> Thanks !
19:07:41 <abadger1999> #info  lpf is okay for Fedora, lpf-* packages should move to a more appropriate repo Needs more votes: (+1:3, 0:0, -1:0)
19:07:50 <geppetto> abadger1999: Want to do the same for 363?
19:07:59 <abadger1999> #topic 363 exception for bundled library libntirpc in nfs-ganesha
19:08:03 <leamas> If possible, yes
19:08:12 <geppetto> Given 2014 timeline … +1
19:08:15 <leamas> (I'm the reviewer)
19:08:24 <abadger1999> yep, let's vote to allow this with what.. a 2 release temporary exception?
19:08:27 <leamas> Sorry, wrong bug
19:08:37 <geppetto> abadger1999: sure
19:08:53 <limburgher> +1 for lpf
19:09:00 <abadger1999> Proposal Temporary bundling exception for libntirpc in nfs-ganesha until after Fedora 22
19:09:04 <geppetto> abadger1999: Maybe 3 depending on how you count.
19:09:12 <abadger1999> Proposal Temporary bundling exception for libntirpc in nfs-ganesha until after Fedora 23
19:09:15 <limburgher> +1 for nfs-ganesha
19:09:16 <geppetto> I'd allow it to be bundled in F22
19:09:16 <abadger1999> +1
19:09:29 <geppetto> meh. +1
19:09:50 <RemiFedora> +1
19:09:53 <abadger1999> the last topics of the meeting are always easy to get +1's ;-)
19:09:58 <geppetto> ha
19:10:01 <gholms> Heh
19:10:04 <abadger1999> Okay, I think that's everyone who's here.
19:10:19 <limburgher> Yeah, I have a hard stop in 15.
19:10:24 <abadger1999> #info Temporary bundling exception for libntirpc in nfs-ganesha until after Fedora 23 -- need more votes in ticket (+1:4, 0:0, -1:0)
19:10:44 <abadger1999> #topic Open Floor
19:10:57 <abadger1999> if there's anything else holler or I'll close in 15s.
19:11:22 <abadger1999> RemiFedora: Thanks for putting us on the aniversery T-Shirt page :-)
19:11:27 <abadger1999> #endmeeting