16:00:26 <geppetto> #startmeeting fpc
16:00:26 <zodbot> Meeting started Thu Jun 29 16:00:26 2023 UTC.
16:00:26 <zodbot> This meeting is logged and archived in a public location.
16:00:26 <zodbot> The chair is geppetto. Information about MeetBot at https://fedoraproject.org/wiki/Zodbot#Meeting_Functions.
16:00:26 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
16:00:26 <zodbot> The meeting name has been set to 'fpc'
16:00:26 <geppetto> #meetingname fpc
16:00:26 <zodbot> The meeting name has been set to 'fpc'
16:00:26 <geppetto> #topic Roll Call
16:05:10 <tibbs> Hello.
16:05:31 <tibbs> Guess it's just you and me.
16:05:43 <geppetto> #chair tibbs
16:05:43 <zodbot> Current chairs: geppetto tibbs
16:05:45 <geppetto> yeh, so far
16:07:51 <carlwgeorge> .hi
16:07:52 <zodbot> carlwgeorge: carlwgeorge 'Carl George' <carl@redhat.com>
16:07:56 <geppetto> #chair carlwgeorge
16:07:56 <zodbot> Current chairs: carlwgeorge geppetto tibbs
16:08:03 <geppetto> #topic Open Floor
16:08:12 <geppetto> Either of you have anything you want to talk about?
16:08:27 <carlwgeorge> just a heads up about that gpsd thing
16:08:41 <carlwgeorge> https://pagure.io/packaging-committee/issue/1283
16:09:32 <tibbs> I think I said all I can say in that ticket already.  I don't think there is any good solution.
16:09:34 <carlwgeorge> i closed the package review bug and the private bug the discussion started in as wontfix, based on the results of the fpc ticket.  i expect the maintainer will not be happy and will keep pushing for us to tell him his approach is allowed.
16:10:08 <carlwgeorge> i'm worried he won't stop unless we explicitly say "your way is not allowed", not merely not recommended
16:11:41 <tibbs> I'm not entirely sure it's not allowed on its face, just that any resulting poor user experience might be disallowed.
16:12:33 <tibbs> But we just don't know what that's going to look like.
16:13:12 <tibbs> Guidelines here are designed to prevent a terrible user experience where you can't install various unrelated things at the same time because there is a conflict somewhere down in their dependency trees.
16:13:19 <geppetto> Seems like it should hit rules on where shared libraries go, or something.
16:14:11 <carlwgeorge> what i told him early on was that naming a package with a certain string doesn't exempt you from following other guidelines
16:16:15 <tibbs> This is primarily an EPEL thing, isn't it?
16:17:32 <geppetto> yeh
16:18:53 <tibbs> Was there any input from the EPEL steering committee folks?  I suspect they wouldn't like this much either.
16:19:01 <carlwgeorge> we don't
16:19:17 <carlwgeorge> we talked about it, but landed on "let's get fpc's take first"
16:19:53 <carlwgeorge> our recommendation was to follow the existing guidelines and do the incompat update process when needed
16:20:22 <tibbs> That's what I figured.  I guess that is all linked in the initial comment.
16:20:35 <carlwgeorge> the maintainer's outlook is that the incompat process is too much work because he would need to do it too often, which just lends credence to the package being a bad fit for epel in the first place.
16:20:54 <carlwgeorge> if the fpc stance is to let epel steering decide, i can work with that
16:21:11 <geppetto> <pastes the same comment from two weeks ago about using COPR and EPEL>
16:21:45 <carlwgeorge> the response to that has been "I suspect a copr repo wouldn't be acceptable for the customers or our support."
16:22:05 <geppetto> I think we are pretty firmly against it … but I guess if the EPEL sterring committee wanted to do something different we'd look at that.
16:22:33 <carlwgeorge> which to me points clearly to it going into rhel, but that was already decided against, which should make people think about the stability of this upstream
16:22:59 <tibbs> My personal stance is that in general this kind of thing is OK as long as the user experience is not bad (i.e. leaf packages still work regardless of which underlying gpsd package you have installed).  But if applications have to choose which they will use, and then get partitioned into two groups that can't be installed together, then that would be very bad.
16:23:24 <tibbs> We have things like "foo" and "foo-minimal" in various places and this isn't a huge problem.
16:23:32 <carlwgeorge> the proof of concept of how he wants to do it is here https://copr.fedorainfracloud.org/coprs/mlichvar/gpsd-minimal/
16:23:42 <tibbs> I suspect we'll end up with a "tzdata-minimal" package, for example.
16:24:52 <carlwgeorge> but even that implementation includes so files, so breaking changes would still technically require the incompat process, so there is no point
16:27:31 <tibbs> Fundamentally this just shows that gpsd is problematic.
16:27:52 <tibbs> I don't think there is any good answer besides "don't break the user experience".
16:28:06 <tibbs> If you can't switch out gpsd-minimal for gpsd without things breaking, then that breaks the user experience.
16:30:05 <tibbs> Otherwise I think that guidelines-wise it would be OK unless there is something else I'm missing.
16:31:37 <geppetto> EPEL has supportyeh
16:32:03 <geppetto> Blah … meant to just say yeh to tibbs
16:32:21 <tibbs> I was searching for "supportyeh".
16:32:29 <geppetto> First words were in reponse to COPR being less "supporty" than COPR, which meh
16:33:29 <geppetto> And to the extent that it's true, part of the reason for that is due to not being able to do user breaking changes … like this.
16:34:44 <carlwgeorge> this would have been a perfect fit for epel-playground, but we shut that down due to lack of adoption
16:35:35 <geppetto> probably doesn't look much better than COPR from the outside.
16:36:05 <carlwgeorge> that was a big factor in the shutdown, no need to duplicate what copr does
16:37:27 * geppetto nods
16:37:49 <geppetto> Ok, well I'm not sure there's more to say … so unless anyone has anything else I'll close at 12:40
16:38:03 <carlwgeorge> what's the fpc stance then
16:38:15 <carlwgeorge> say no outright or defer to epel steering
16:38:28 <carlwgeorge> or something else
16:38:54 <geppetto> Well there's only 3 of us … but I'd say no, but I guess they can complain to steering comittee and
16:38:57 <geppetto> …
16:39:31 <carlwgeorge> i can work with that.  i'll retag the epel ticket for meeting.
16:39:36 <tibbs> To me it's all about the user experience.  It doesn't seem to me that this can be done without breaking it, but I'm willing to look at evidence to the contrary.
16:39:52 <tibbs> I will say that again in the ticket just in case it helps.
16:39:55 <carlwgeorge> great way to look at it
16:39:59 <carlwgeorge> yes it would, thank you
16:40:29 <tibbs> There is one other interesting possibility that I can see:
16:41:47 <tibbs> All packages in the distribution use only the "unstable" (or whatever) gpsd version and if you want the old version for use with your own development or self-compiled stuff then that locks you out of any distro-packaged applications which depend on gpsd.
16:42:32 <tibbs> But then that is probably more OK for Fedora than for EPEL.
16:43:05 <carlwgeorge> yeah i can't imagine that solution would be popular with epel steering
16:43:54 <tibbs> The guidelines come from the stance of not making users choose between disjoint sets of packages because of conflicts down in the dependency trees and that's about it.
16:45:13 <geppetto> Yeh, I'd also mostly side on the EPEL doesn't break or update things … rather than update a lot and break whatever.
16:45:25 <geppetto> Anyway …
16:45:27 <geppetto> #endmeeting