16:00:09 <geppetto> #startmeeting fpc
16:00:09 <zodbot> Meeting started Thu May  4 16:00:09 2017 UTC.  The chair is geppetto. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:00:09 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
16:00:09 <zodbot> The meeting name has been set to 'fpc'
16:00:09 <geppetto> #meetingname fpc
16:00:09 <zodbot> The meeting name has been set to 'fpc'
16:00:10 <geppetto> #topic Roll Call
16:00:43 <tibbs> Wow, time already.
16:01:27 * limburgher here
16:01:31 <mbooth> Hi
16:02:23 <geppetto> #chair mbooth
16:02:23 <zodbot> Current chairs: geppetto mbooth
16:02:26 <geppetto> #chair limburgher
16:02:26 <zodbot> Current chairs: geppetto limburgher mbooth
16:02:28 <geppetto> #chair tibbs
16:02:28 <zodbot> Current chairs: geppetto limburgher mbooth tibbs
16:03:53 <orionp> sort of here
16:04:45 <tibbs> In case I get called out, I'm +1 on the default services change, but I will note that it's slightly vague.
16:05:42 <geppetto> #chair orionp
16:05:42 <zodbot> Current chairs: geppetto limburgher mbooth orionp tibbs
16:05:58 * geppetto nods
16:06:03 <geppetto> #topic Schedule
16:06:04 <geppetto> https://lists.fedoraproject.org/archives/list/packaging@lists.fedoraproject.org/message/N3C5ETS7BPAEZBJAQXJL6D25UQ2CWWCZ/
16:06:23 <geppetto> #topic  #683 Update to DefaultServices guidelines
16:06:30 <geppetto> .fpc 683
16:06:31 <zodbot> geppetto: Issue #683: Update to DefaultServices guidelines - packaging-committee - Pagure - https://pagure.io/packaging-committee/issue/683
16:07:32 <geppetto> What do you think is vague, what would you reword?
16:08:16 <tibbs> Well the whole thing is this section "Restrictions":
16:08:22 <geppetto> Yeh
16:08:23 <tibbs> Installation of the package providing the unit auto-started by this preset may not change the behavior of any other service running (or potentially running) on the system.
16:08:54 <limburgher> Run-time behaviour or configured systemd service status?
16:09:03 <tibbs> And, well, I've thought about it for a day and I'm still not entirely sure what that means.
16:09:11 <limburgher> Like, disabling apache is a no-no but can it provide QOS?
16:09:15 <limburgher> Hypothetically?
16:09:36 <limburgher> Not a great hypothetical.
16:09:40 <geppetto> Yeh, I assumed "change negatively"
16:09:45 <limburgher> But you get what I'm laying down.
16:09:55 <geppetto> So you can't turn other stuff off … or disable random features
16:10:04 <tibbs> The specific question I saw asked about this was relating to a performance management daemon.
16:10:38 <limburgher> Ooh, that could be sticky.
16:11:09 <geppetto> in what way?
16:11:36 <tibbs> Well, a performance management daemon is something we might want, and it definitely impacts, well, everything else.
16:11:57 * geppetto might just be being dense … what do people think it'll do
16:12:10 <tibbs> But does it actually change the behavior of anything?  Does "performance" count as "behavior"?
16:12:20 <geppetto> So the assumption is that this perf. management daemon will start turning things off for performance?
16:12:32 <geppetto> Wouldn't that be user controlled in some way?
16:12:57 <limburgher> yeah, but I would assume it would have defaults.
16:13:12 <geppetto> I'd guess so, if the other apps. started going slower without any feedback
16:14:14 <geppetto> I might be being optimistic, but I'd assume the defaults would be "do no evil" … but would have a desktop notification like thing that says "for N hours more battery life click this button and everything will go slower" or whatever
16:14:19 <tibbs> Anyway, I'm pretty sure I understand why this is there.  It's just that someone is going to look at this guideline and have to stop and ask about it.
16:15:12 <geppetto> I mean … as with many things I feel like it could be replaced by "don't be a dick" and we'd be good … but ¯\_(ツ)_/¯
16:15:35 <geppetto> I've long given up on expecting that, so +1
16:16:36 <tibbs> You're completely correct there.  I just don't know if there's any other way to say what this change says so that it won't confuse people.
16:17:24 <orionp> Is tuned enabled by default?
16:17:31 <geppetto> Maybe something about "if you are going to change the behaviour of another package, discuss it with that maintainer and make sure they approve"
16:18:00 <tibbs> Or "if unsure, ask <something>"
16:18:31 <orionp> Are we concerned about behavior of "system" or "service" or "package"?
16:18:57 <tibbs> I'm not even sure, honestly.
16:19:35 <tibbs> Obviously the idea is to avoid the case where the mere installation of package A breaks package B because A enables some service by default.
16:20:03 <geppetto> Or disables, but yeh
16:20:29 <tibbs> So, yeah, our ticket has a link to https://pagure.io/fesco/issue/1698
16:20:36 <tibbs> Which I will now confess to just reading.
16:20:55 <orionp> I'm not sure that's the idea behind this request though - I'm not sure what thermald does, but presumably it only affects scheduling/cpuspeed somehow?
16:21:31 <tibbs> Well, maybe you'd have to get inside sgallagh's head to know the true idea behind this.
16:21:52 <tibbs> Essentially, the kernel folks have chosen a set of config knobbs.
16:22:10 <sgallagh> *looks up*
16:22:22 <tibbs> If you install thermald (and it was allowed to start by default) then merely installing the package overrides the kernel team's choice.
16:22:43 <orionp> but it what way does this affect another "service"?
16:23:05 <tibbs> I certainly can't make that argument.
16:23:22 <sgallagh> orionp: Basically, FESCo was treating "'Kernel Power Management" as a service.
16:23:37 <tibbs> I guess the idea is to give FESCo an out if they want to deny something.
16:23:54 <geppetto> Yeh, service == something running providing a service … not just the things in systemd
16:23:58 <sgallagh> This would be a change to that behavior that the customer might not expect if it was pulled in accidentally or as a dep
16:24:42 <sgallagh> tibbs: Sort of. It basically was just another specific case we want to have always brought to FESCo rather than waved through.
16:25:02 <sgallagh> That page just lists the conditions it has to meet to skip a FESCo decision
16:25:19 <orionp> So should ksmtuned need to get an exception?
16:26:04 <sgallagh> orionp: IMHO, yes. The default behavior should be "if unsure, ask FESCo"
16:26:06 <tibbs> OK, so basically, anything that doesn't have "negligible impact" on the _running system_ needs to go through fesco.
16:26:26 <sgallagh> tibbs: that's probably a good summary.
16:26:29 <tibbs> Personally I'd prefer that every single change to the presets to through fesco.
16:26:38 <geppetto> Again, I assume there'a s hidden negative here
16:26:41 <tibbs> I'm not sure why anything gets a pass at all.
16:27:26 <geppetto> Like if you want to auto start keybase and it provides the /keybase virtual FS, I'd guess that's fine
16:27:27 <tibbs> But I guess the owners of the package providing the presets would notice any change and let someone know if they saw something going in with which they didn't agree.
16:27:40 <geppetto> Even though it technically changes the FS
16:27:45 <sgallagh> Right, which is usually me or dgilmore.
16:27:59 <geppetto> It would be hard to see it as anything but an expected positive
16:28:58 <sgallagh> tibbs: that's what happened in this case; it came in and technically met the previous guidelines we set, but didn't feel right.
16:29:33 <tibbs> It's just that I don't think adding something weirdly vague as a catch all is the best way to address that.
16:29:35 <orionp> How about "changes kernel parameters/behavior" or something like that? (still doesn't know what thermald actually does)
16:29:37 <sgallagh> So we tried to set precedent within the restrictions of imperfect language
16:30:06 <tibbs> I mean, sure, it makes the people wonder if they're breaking it and therefore more likely to go to FESCo....
16:30:09 <sgallagh> tibbs: if you have a suggestion for better phrasing, I'm open to it.
16:30:29 <tibbs> But I'd prefer that people get more answers than questions from packaging guideline pages.
16:31:03 <tibbs> Oh, I'd personally redo the whole deal by requiring all preset changes which enable additional services to go through fesco.
16:31:44 <tibbs> But this is basically FESCo's page, and if FESCo likes the language then I don't have a big problem with it.
16:31:57 <sgallagh> It was intended partly to speed things up. Waiting one or more weeks for FESCo to meet for trivial decisions was not ideal
16:33:15 <tibbs> True, but there's no technical reason why FESCo has to meet.  The bar could simply be an ack by a committee member.  That would kind of be up to FESCo.
16:33:17 <sgallagh> Of course, it's not like the fedora-release package gets updated instantly either.
16:33:32 <mbooth> How about just adding an additional clause, e.g.:
16:33:33 <mbooth> "Installation of the package providing the unit auto-started by this preset may not change the behavior of any other service running (or potentially running) on the system or otherwise negatively impact the running system."
16:33:50 <mbooth> (Taken from tibbs phrasing)
16:33:52 <tibbs> Sure, negative impact is the point here, isn't it?
16:34:01 <geppetto> Sure, I'maso +1 on that
16:34:08 <geppetto> I'm also, too
16:34:15 <sgallagh> mbooth: Is any requesting person going to think they are having negative impact?
16:34:38 <orionp> Is thermald really a negative impact?  "It keeps my system from overheating!"
16:34:48 <mbooth> Good points :-)
16:35:17 <sgallagh> orionp: Unless I happen to have a system that thermald causes to spin the fans until they break off and cut through wires...
16:35:24 <orionp> I tend to fall back also to - "I installed it, I want it to run" , but meh
16:35:51 <sgallagh> orionp: which would make sense if we could indicate intent at the installation level
16:36:17 <sgallagh> But we have no way to know if packages are installed on purpose or just as part of a dependency chain
16:36:23 <geppetto> Again, everything affects the system … so this feels like a weird way to randomly exclude things, or at least send them to fesco … but sgallagh already did that anyway.
16:36:25 <sgallagh> (Possibly even a malicious one)
16:36:56 <sgallagh> To be *clear* it's not for exclusion.
16:37:06 <sgallagh> It's to say it's not an auto-accepted request.
16:37:23 <sgallagh> Those are different things.
16:37:35 * geppetto nods … yeh, I meant "exclude from the auto accepting nature of the policy"
16:39:04 <geppetto> Anyway … does anyone object enough to try for other wording … or to not vote +1?
16:39:11 <sgallagh> Honestly, much of this is about covering *me* by having something to point to when I don't wave things through. (Since I'm usually the one who processes preset requests)
16:39:33 * geppetto nods
16:40:23 <limburgher> geppetto: not really.
16:40:26 <tibbs> I'm still +1; I'm only pointing out that it hasn't even been written up yet and people have already asked questions about it.
16:41:17 <geppetto> "If you have any questions/complaints, send them to sgallagh." ;)
16:41:30 <sgallagh> geppetto: I can work with that
16:41:36 <limburgher> Same
16:41:38 <mbooth> "Installation of a package providing the unit auto-started by this preset may not change the behaviour of anything outside of that package on the system."
16:42:23 <geppetto> That seems even more generic, and confusing
16:42:36 * mbooth stabs in the dark :-)
16:42:45 <tibbs> "If in doubt, please consult FESCo"
16:42:47 <limburgher> AAAAAAAAAAAAAAAAAAAAAAAaaaaaaa my tummy
16:42:54 <sgallagh> tibbs: that
16:43:07 <racor> mbooth: I believe to have understood sgallagh's intentions, but this sentence doen't parse to anything, I understand.
16:44:11 <mbooth> racor: I meant to make a clear boundary of "don't change stuff outside of your package with this preset"
16:44:19 * mbooth shrugs
16:44:31 <geppetto> "Installation of the package providing the unit auto-started by this preset may not change the behavior of any other service running (or potentially running) on the system. If there are any questions defer to the bugzilla owner for fedora-systemd-request, or FESCO"
16:44:52 <geppetto> #chair racor
16:44:52 <zodbot> Current chairs: geppetto limburgher mbooth orionp racor tibbs
16:45:23 <racor> the tough part I am chewing on is "the behavior".
16:45:28 <sgallagh> geppetto: just say defer to FESCo
16:45:29 <racor> what is it=
16:45:41 <racor> s/=/?/
16:45:57 <geppetto> sgallagh: Fair enough
16:46:18 <geppetto> sgallagh: I kind of meant it as the more generic, any questions/problems speak to sgallagh
16:46:54 <geppetto> So they'd have someone obvious to contact, without opening fesco tickets
16:47:11 <sgallagh> Well, it's just redundant. A FESCo member will always be the bugzilla monitor for that.
16:47:19 <geppetto> ahh
16:48:18 <limburgher> Also, "FESCO" scales better than "sgallagh".  No offense to sgallagh. :)
16:48:53 <mbooth> I can +1 -- I don't mind the original wording, and turns out I can't make it clearer :-)
16:49:08 <limburgher> +1
16:49:18 <orionp> +1
16:49:31 <geppetto> Ok, still +1 on original wording
16:49:38 <geppetto> tibbs: You too I assume?
16:49:43 <tibbs> Yep, +1.
16:49:58 <geppetto> racor: orionp Ok, so just one o you needs to +1
16:50:12 <orionp> +1 .. since I don't have to explain it
16:50:24 <geppetto> :)
16:50:52 <geppetto> racor: want to vote for the record?
16:51:25 <racor> 0 ... I think this sentence will only lead to further confusion and discussions
16:51:30 <sgallagh> Ok; I'll go back to working my booth at RH Summit now. Thanks folks
16:52:08 <geppetto> sgallagh: no problem … thanks for joining
16:52:14 <geppetto> #action Update to DefaultServices guidelines (+1:5, 0:1, -1:0)
16:52:22 <geppetto> #topic Open Floor
16:52:41 <orionp> anything about the new comments in https://pagure.io/packaging-committee/issue/509 ?
16:55:05 <geppetto> Doesn't this break everything … the bootstrap change?
16:56:07 <geppetto> I also want to -1 just because re-opening 2 year old tickets
16:59:39 <orionp> Yeah, I agree
17:02:47 <geppetto> Ok, I'm going to ask him to open a new ticket given it looks like he wants to change things non-trivially.
17:03:07 <geppetto> Anything else?
17:03:10 <limburgher> Excellent.
17:03:16 <limburgher> Not I, said the duck.
17:03:50 <tibbs> Hadn't seen those yet.
17:05:13 <geppetto> #endmeeting