2007 May 17 FESCo Meeting
- Brian Pepple (bpepple)
- Jason Tibbitts (tibbs)
- Jesse Keating (f13)
- Toshio Kuratomi (abadger1999)
- Bill Nottingham (notting)
- Kevin Fenzi (nirik)
- Dennis Gilmore (dgilmore)
- Josh Boyer (jwb)
- Warren Togami (warren)
- Jeremy Katz (jeremy)
- Tom Callaway (spot)
- Rex Dieter (rdieter)
- Christian Iseli (c4chris)
- It's expected that the Package DB should go into production shortly after F7 is released.
Enforcement of Packaging Guidelines
Compat policy for future
- tibbs is going to send an e-mail to mailing lists regarding packages that download non-redistributable content, to get feedback from the community.
<bpepple> FESCo meeting ping -- abadger1999, bpepple, c4chris, dgilmore, f13, jeremy, jwb, notting, rdieter, spot, nirik, tibbs, warren
* rdieter here
cweyl|work grabs his rabble seat
wwoods heads into cheap seats and quiets down
jeremy is here
nirik is here
--- bpepple has changed the topic to: FESCO-Meeting -- MISC -- Package Database - abadger1999
<bpepple> figure we can start with this while people show up.
* notting is here
<bpepple> anything to report on this abadger1999?
* warren here
<abadger1999> I've started working on the new sync.
From the current pkg.acl and owners.(epel)?.list
Should be done by tomorrow.
<abadger1999> I've shed my outside obligations until I start at RH in June so things should progress much faster now.
<bpepple> abadger1999: cool. do we know when the package db will be ready for production?
<abadger1999> I think the consensus is it's too much change to introduce before the release so sometime shortly after F7 is released.
<skvidal> abadger1999: are you a remotee, too or are you working from somewhere else?
<abadger1999> skvidal: I'll be remote as well.
<skvidal> abadger1999: when's your first day?
<bpepple> anything else regarding the package db, or should we move on?
* dgilmore is here
<abadger1999> Let's move on.
skvidal: Orientation on June 4.
--- bpepple has changed the topic to: FESCO-Meeting -- MISC -- how to handle enforcement of guidelines? -- https://www.redhat.com/archives/fedora-maintainers/2007-May/msg00563.html
<bpepple> is spot about?
* skvidal leaves everyone alone now, sorry for the interruption
<bpepple> this was being discussed on the mailing list yesterday, and a few of the FESCo members wanted to discuss this today.
though I don't see spot or jwb.
<tibbs> I think it needs to be discussed at some point, although I recall that spot also indicated that he thought it should go to the board instead.
<abadger1999> Board is fine with me but we should register an opinion.
<bpepple> abadger1999: I agree.
<rdieter> Is it a question of who's job/domain it is for enforcement?
<abadger1999> Hey pjones, glad you could make it.
<tibbs> The latest ML discussion seems to be driven by to RH folks who are against the idea of guidelines as anything other than suggestions.
<bpepple> how do folks think we should handle maintainers that refuse to follow the package guidelines?
<notting> tibbs: patrice works for rh?
<nirik> well, I think the first step should be: "hey, don't do that" and see if they come around...
<pjones> bpepple: raise the issue to fesco, have fesco appoint somebody to arbitrate between the maintainer and other parties, keeping the guidelines in mind.
<rdieter> we need to clarify the difference between refusal to comply, and a maintainer who thinks they have a case for exceptions.
<tibbs> notting: No, I'm referring to two other participants in the discussion.
<abadger1999> notting: I think tibbs is referring to the replies to the "Plan for tomorrow's meeting" thread.
<pjones> nirik: absolutely; the fundamental question here is "what happens when I find a broken package". The answer is "talk to the maintainer"
<tibbs> Patrice doesn't seem to care what the prevailing interpretation of the guidelines is; he intends to ignore them.
<bpepple> tibbs: right, his reply was after abadger1999 told him not to do that.
<nirik> well, I don't speak for him, but It sounded like at first he thought it was ok to do that to not "bother" fesco/packaging with it...
<|DrJef|> bpepple, enforcement for volunteers is an extremely difficult issue, we don't have any performance review metrics to encourage volunteers to play well with guidance
<pjones> nirik: if the maintainer refuses to fix it, doesn't respond, or whatever, bug FESCo, and then FESCo can get somebody to arbitrate. Maybe there's a solution in the package, or maybe there's a reason to change the guidelines, or maybe we should just note "this package is weird in this regard", but arbitration is the way to decide.
<pjones> |DrJef|: right, which is why the question I stated above is really the driver of the process.
<rdieter> pjones: +1
<nirik> pjones: +1
<f13> I like pjones's ideas.
<bpepple> +1 to pjones idea here also.
<abadger1999> pjones: +1.
<warren> pjones, +1
<notting> we can have it in the CLA that by signing this you agree to binding arbitration with fesco in the case of any disputes...
* f13 glares at notting
<tibbs> +1, but that's not really a change from the way it's been so far.
<pjones> notting: well, we can just appoint somebody else maintainer or dump the package if somebody is truly a PITA.
tibbs: yeah, really it's just that nobody *knows* what the process is right now.
<abadger1999> It's a more efficient form of what we want. The goal is to give packagers the ability to break a guideline when it makes sense and not otherwise. Whether we get there from FESCo deciding or a representative of FESCo deciding makes no difference to me.
<tibbs> Well, spot's not here; let's appoint him.
<nirik> he maintains 60 packages, FYI
<|DrJef|> notting, other than the CLA... having an explicit rules of conduct manifesto that is not strictly legally binding wouldn't be a horrible idea
<warren> notting, but RH engineers have a separate CLA
<pjones> abadger1999: well, point being, a packager shouldn't have to ask "can I violate the guidelines". Almost any time somebody *considers* asking that, they've got a good enough reason.
<bpepple> probably should appoint someone that isn't on the packaging committee.
<pjones> warren: he was being facetious, I think.
<abadger1999> pjones: Well... That I don't agree with.
* jwb is here now
<cweyl|work> So what would be an appropriate first step on noticing a packaging issue? I typically file a bug, but that might be something good to be made explicit
<abadger1999> At least not in the current situation.
<pjones> abadger1999: I don't think people generally recognize when they're violating the guidelines.
<abadger1999> That is fine.
<pjones> abadger1999: thus, they don't consider asking.
<bpepple> cweyl: I would think notifying FESCo.
<|DrJef|> pjones, not in this case
<nirik> I think this would fall under the maintainers responsabilities thing that we never finished putting together...
<pjones> cweyl|work: filing a bug is an adequate way of contacting a maintainer, sure.
<notting> warren: is joke, read your credit card agreements
<cweyl|work> bpepple: well, that's on a "I refuse to comply", right?
<jwb> bpepple, notifying FESCo with a bug is better
<pjones> |DrJef|: not in *this* case, but in the regular case, yes?
<cweyl|work> pjones: cool.
<warren> notting, nod
<tibbs> Packaging bugs are stil bugs, so I see no reason not to file one but of course anyone can contact anyone via email or IRC or whatever.
<|DrJef|> pjones, i don't know what a regular case means
<pjones> |DrJef|: I'm sure there are plenty of places where packages violate the guidelines and nobody has noticed. In general, we don't need to worry about those.
(I suspect I maintain more than one of such violatons...)
<f13> there are many unnoticed violations too, like not preserving timestamps when copying files from sourcetarballs
<nirik> I think we don't want to get to formal or soon we will have lawyers involved. ;)
<f13> which can lead to multilib issues
and then there are some silly ones, like having a period at the end of a summary
<abadger1999> pjones: And in the regular case what happens is that reviewer and packager notice issue. Issue is taken to f-devel or f-maintainers. People discuss and come to some conclusions. if guidelines must be changed to accomodate someone takes it to the PC while the review goes forward (sometimes with the review finishing othertimes with the reviewer refusing to finish until the guideline is approved.)
<pjones> f13: right. but when those cases result in bugs, somebody will be affected, and they'll file a bug, and in general the maintainers will do the right thing.
<nirik> have fesco assign an arbitrator to talk to the maintainer and try and work out something, if that fails, remove maintainer?
<f13> pjones: yep, however we're trying to prevent the bug from happening by having htem in the guidelines.
<jwb> nirik, no
<pjones> f13: what we need to have policy for is when maintainers refuse to do the right thing, or when people are asking them to do something and it /isn't/ right.
<bpepple> pjones: +1
<abadger1999> pjones: +1
jwb: alternate proposal? :)
<pjones> f13: sure, which means the guidelines are "best practices", and they should guide both how to make a package and how to resolve *real* disputes which affect people.
<jwb> nirik, oh i thought you were asking if we _had_ assigned someone. sorry, misunderstood
<nirik> I was just trying to see if we could come up with a rough plan for such things...
<pjones> To be clear, I'm not saying packagers shouldn't say "hey guys, I've got a case which seems wrong when I follow the guidelines...". I just don't think we need to institutionalize that part of package maintenance.
<abadger1999> pjones: I disagree with that terminology... The guidelines should be rules but able to be changed/bent by the arbitrator.
<pjones> abadger1999: I *think* that's largely a semantic disagreement...
<abadger1999> Probably :-)
<bpepple> Ok, so are FESCo members ok with pjones proposal?
<abadger1999> Could we spell it out?
<jwb> please spell it out
<f13> pjones: you know we're going to task you with coming up with a draft (:
* nirik would also like to see exactly what we are voting on.
<f13> because you have _sooo_ much free time :/
<pjones> f13: I figured somebody was about to say that. At least you're close enough to throw usb toys at over the cube wall.
ok, want me to type up a more formal proposal and send it to the list?
<warren> yes please.
<pjones> (anybody mind if I fix raid in the rescue images first?)
<bpepple> pjones: yeah, that would be great.
<f13> I don't mind at all (:
<jwb> can someone recap it briefly for me?
i was late... sorry
<bpepple> if the maintainer refuses to fix it, doesn't respond, or whatever, bug FESCo, and then FESCo can get somebody to arbitrate. Maybe there's a solution in the package, or maybe there's a reason to change the guidelines, or maybe we should just note "this package is weird in this regard", but arbitration is the way to decide.
<pjones> bpepple: that's the long and short of it, yeah ;)
<dgilmore> bpepple: im ok with that
<jwb> i'm ok with that
<bpepple> pjones: just copying what you initially wrote. ;)
<pjones> bpepple: I think that was the second draft, but sure ;)
<abadger1999> That would be fine.
<cweyl> so, from my rabble seat. file a bug, on dispute go to fesco, fesco assigns moderator, moderator mediates -- decision binding on all involved, guidelines can be "bent"; fesco has final power to enforce said moderators decision?
<jwb> i'd like to know what happens when arbitration fails though
* nirik was also wondering that
<pjones> cweyl: maybe file a bug and if it gets no response email the maintainer to get his attention, or whatever.
<abadger1999> yes. That needs to be the second half of this discussion.
<dgilmore> jwb: start the orphan process
<pjones> jwb: the arbitrator has to have teeth, I guess.
<cweyl> pjones: yeah. bz is just the canonical method of contacting them
<pjones> what dgilmore said.
<cweyl> jwb: isn't this why fesco funded the orbital lazer?
<jwb> k. pjones could you just add that in your draft?
<pjones> jwb: sure.
cweyl, they haven't given me the keys to the lazer yet
<bpepple> pjones: thanks appreciate your help on this.
<nirik> well, I think fesco and/or board should make the final decision if it fails?
<jwb> cweyl, i still have to rely on my lame sharks with lazers on their heads
<dgilmore> jwb: im keeping them locked up for now
we had to reallocate laser firing budget to storage
<|DrJef|> is the contributor community healthy enough for teeth?
<cweyl> nirik: fesco, appeal can be made to board?
<warren> cweyl, the orbital laser uses the binary-only nvidia driver and it crashed. We can't debug it.
<bpepple> |DrJef|: Do we really want to keep maintainers that refuse to be part of the community?
<cweyl> jwb: I just ordered a "laser mouse" from newegg. I was so disappointed...
<nirik> moderator makes reccomendation to fesco, fesco votes, can be appealed to board sounds ok to me.
<abadger1999> |DrJef|: I don't know.
<|DrJef|> here's my biggest concern with teeth.. it is my impression that moving forward the most problematic packagers are going to be packagers who have a large set of maintained packages and will use that large investment to pushback against any teeth in any enforcement
<jwb> |DrJef|, valid concern
<tibbs> I think we can deal with that, though.
<|DrJef|> are you prepared to watch a maintainer leave 100+ packages over an enforcement issue?
<tibbs> I'd choose that over a packaging free-for-all with no rules.
<|DrJef|> tibbs, because those are the people who will be the first test cases of an arbitration breakdown
<cweyl> |DrJef|: that's why it's important to have a fair, reasonable process where deviations can be approved.
<jwb> |DrJef|, we need to be healthy enough to handle that anyway
what if that maintainer dies?
<|DrJef|> jwb, my suggestion to you is.... as long as we have maintainers with 100+ package workloads..its not healthy
<nirik> or disappears from the interweb...
<spot> damn, i get stuck in traffic and you people get... morbid.
* nirik looks at ivazquez. ;)
<jwb> and no, I'm NOT suggesting we should just willy-nilly drop maintainers
<tibbs> The thing that's sadly hidden in this discussion is that Patrice has a valid point that deserves further scrutiny.
<jwb> tibbs, which point?
<tibbs> People who use his packages often want to link statically.
<abadger1999> I am more worried about the community as a whole feeling like the enforcement isn't fair and having the community fracture.
<cweyl> abadger1999: ++
<bpepple> abadger1999: agreed.
<tibbs> Because they're running computational code on loads of different machines.
<cweyl> yeah. along those lines, moderator should be a peer, that is someone neither on fesco nor the FPC, yet someone generally well respected in the community.
<tibbs> Unfortunately Patrice was blown off by Ralf, and so Patrice decides that the guidelines shouldn't apply to him.
<|DrJef|> tibbs, i very much agree with the underlying issue... but i violently disagree with patrice's approach to getting it resolved
<bpepple> ok, we should probably move on. we can discuss this some more after pjones finishes his write-up,=.
<|DrJef|> tibbs, i'm a consumer of some of the packages at question..i have a personal stake in seeing it resolved
<abadger1999> tibbs: I would be more than willing to have the guidelines adapted to allow those packages to ship static libs... but he needs to present a definition of those libraries so we can add it to the guidelines.
<tibbs> Unfortunately his communication was mainly with Ralf, hence the breakdown.
<bpepple> abadger1999: +1
moving on, unless anyone objects.
<f13> please do
<|DrJef|> tibbs, no comment
--- bpepple has changed the topic to: FESCO-Meeting -- MISC -- F7 Prep
<bpepple> anything we need to discuss regarding F7?
<f13> Branching today!
(at some point)
<bpepple> f13: sweet!
actually, i have a question on that
<f13> branching == outage
<bpepple> f13: how long do you think the outage will be?
<jwb> so after when we branch all the koji tags will be in place correct?
<f13> bpepple: dunno, there are a _lot_ of branches to make.
<f13> jwb: I think all the koji tags are in place already
dist-fc7-updates(-candidate) dist-f8(-candidate) dist-f8-build
<jwb> f13, and rawhide is still composed from f7-final yes?
<pjones> Ok, I sent an email about it. Now I'll go look at raid some more.
<bpepple> pjones: thanks.
<f13> jwb: yeah, rawhide will still compose from f7-final until we go "gold"
<f13> I'll have to unlock some of those tags but that's quick/easy.
<cweyl|work> um. we're not going to email on branch are we? :)
<f13> and once branched we'll make changes to the build.map
<nirik> is there a key for the tags? what each means?
* nirik gets confused by them for some reason.
<f13> nirik: not exactly. I should create one, but -ETIME :(
and where to put it....
<bpepple> anything else? or should we move on.
<jwb> f13, under the Koji/ page in the wiki
<f13> nirik: if you start a page somewhere listing the tags, I'll happily fill in what they're there for.
<nirik> perhaps on koji? when you add a tag have to add a description?
<f13> nirik: not a horrible idea, but would need patches to koji
<jwb> and a redeployment on the server
<f13> bpepple: I don't think there is anything pressing on it.
<jwb> which means more downtime
<f13> yeah, I'd like to avoid that
--- bpepple has changed the topic to: FESCO-Meeting -- MISC -- Broken Upgrades paths - jwb
<jwb> things are improving there
i'm running daily reports and bugging people
<bpepple> how many are still broken?
<f13> and will continue to improve (:
<tibbs> Were the issues with Ralf's packages resolved in some way?
<f13> we tagged them.
<jwb> last count as of this morning was 30-ish, however i know at least 1/3 of those got fixed and the majority have newer builds that just need to be tagged
<jwb> i need to tweak my scripts and find the packages that aren't in rawhide in any way because they were added after freeze
shouldn't be hard and that number should be fairly small
i'll continue to send out reports and bug people until things get fixed
<bpepple> great. thanks jwb!
<jwb> tibbs, i submitted a request on Ralf's behalf
<f13> pjones: awesome mail!
<notting> what about other things not tagged yet?
<jwb> notting, ?
<f13> notting: the general hidden packages thing?
<notting> other random newer stuff in dist-fc7 aside from upgrade-path stuff
<f13> spot: how goes porting of find-hidden-packages ?
<jwb> notting, that's what i just said i need to tweak my scripts for
<spot> f13: well, i'm not sure what it is supposed to look like
<pjones> f13: yeah. I left out 2 things, one of which on purpose. ("if some little shit won't leave you alone after abritration..." on purpose; "Don't be an asshole." just because I forgot.)
<spot> but its going. :)
<f13> spot: can you put your in progress results somewhere so that I can tweak with it and try to help?
we have that handy dandy releng git repo now...
<jwb> in the git repo would be good so i can steal it
<spot> sure, i'll do it this afternoon
<jwb> spot, this is in python right?
<jwb> good. i know you're a perl freak so i was just checking
<warren_treo> sigh, internet died
<jwb> damn tubes
<bpepple> warren_treo: gotta hate that.
<nirik> f13: if you have a few minutes at some point: http://fedoraproject.org/wiki/Koji/CurrentKojiTags (Or someone who understands the tags and can fill in that)
<bpepple> anything else regarding broken updates?
<f13> nirik: thanks, I'll give it a go.
<jwb> nirik, cool thanks
<jwb> bpepple, not from my end
<bpepple> ok, moving on then.
--- bpepple has changed the topic to: FESCO-Meeting -- MISC -- compat policy for future
<nirik> might be better to put them somehow so it shows which inherit from which, but thats a first cut.
<jeremy> bpepple: I started working on a draft for that last week, but since getting back have been a little preoccupied with fixing f7 blockers instead
<f13> nirik: the bigger trick is going to be remembering to update this page when adding new tags :(
<jeremy> http://fedoraproject.org/wiki/JeremyKatz/DraftCompatPackages is where I got to
<bpepple> jeremy: that's fine, it's certainly less pressing than F7 work.
<nirik> so this is the compat-python24 issue? is someone interested in drawing up guidelines for compat packages?
<bpepple> nirik: that's what jeremy is doing.
<f13> nirik: more generic than compat-python24
<jeremy> nirik: I'm trying to make it be much more generic
<nirik> f13: yeah, thats why it would be better in koji long term.
<jeremy> since it's not just about python
<f13> but driven by that somewhat, in a 'oh god oh god make it stop please' kind of way.
<nirik> cool. Glad to hear it.
<bpepple> ok, we can probably move on then.
--- bpepple has changed the topic to: FESCo meeting -- Free discussion around Fedora
* |DrJef| shivers at the thought of compat-python
<bpepple> Anything in general people want to discuss?
<f13> Can we somehow find an extra week in May?
<tibbs> Well, there's always the issue I brought up on the list.
<bpepple> jwb: I don't think lmacken is about.
<abadger1999> 13: I'll talk to the Pope about that.
<tibbs> Regarding packages that download non-redistributable content.
<|DrJef|> f13, sure.. come to alaska....and enjoy the 20 hour days
<jwb> tibbs, i think we should move that to either fedora-advisory-board or -maintainers
<f13> I think our current guidelines are adaquite with this right? If it can't be used with free content, as in there is no free content available for it, it shouldn't be in Fedora.
<wwoods> let's just redefine "May 29" to actually mean "June 12" or so
<jeremy> (and related to the compat stuff, if anyone wants to help flesh out it, feel free... that's why it's on the wiki ;-)
<abadger1999> My take on that is if it uses autodownloader then it's all about licensing.
<jwb> jeremy, careful what you ask for
<bpepple> tibbs: how about we wait for some more discussion on the mailing list, and then add it to the agenda for next week?
<tibbs> Sure, no problem.
<jwb> bpepple, is there a reason it needs to be on the fesco list?
<bpepple> jwb: I don't see any.
<tibbs> That just happens to be where I sent it originally.
* nirik would prefer it go to maintainers or even deve.
<abadger1999> jwb: +1 to moving it.
<jwb> tibbs, oh i know. i'm just asking if anyone has a preference
<bpepple> tibbs: would you mind sending it to the devel or maintainers list?
<tibbs> Not at all.
<tibbs> But which list?
<jwb> -maintainers would be my preference
but -devel works
<bpepple> tibbs: either works for me. devel would probably have a larger audience.
<abadger1999> I think devel...
<tibbs> Most of the people who card are probably already on -maintainers, but you never know.
I'll use -devel
<bpepple> anything else, or should we start to wrap up this week's meeting?
* nirik has nothing else.
bpepple will end the meeting in 60
bpepple will end the meeting in 30
bpepple will end the meeting in 15
<bpepple> -- MARK -- Meeting End