FPC Meeting Log 2013-10-17

Thu Oct 17 17:33:55 UTC 2013

[09:09:17] <Rathann> abadger1999: FPC meeting? or was there a time change that I missed?
[09:09:45] <abadger1999> Rathann: yeah sorry, my calendar reminder didn't go off.
[09:09:51] <abadger1999> #topic FPC Meeting
[09:09:56] <abadger1999> #topic Roll call
[09:09:59] <abadger1999> Who's here today?
[09:10:08] <tibbs|w> I'm around.
[09:10:08] <abadger1999> I know spot and geppetto are out.
[09:10:12] <tibbs|w> Much more awake than last week.
[09:10:16] <abadger1999> Cool.
[09:10:24] * jsmith lurks
[09:10:41] <abadger1999> #chair Rathann tibbs|w limburgher SmootherFrOgZ
[09:11:07] <abadger1999> limburgher, SmootherFrOgZ, RemiFedora: you guys around for FPC meeting?
[09:11:12] <abadger1999> #chair RemiFedora
[09:11:37] * RemiFedora here
[09:11:47] <abadger1999> Almost to quorum.
[09:13:27] <abadger1999> while we wait -- yesterday geppetto, dgilmore, mmaslano, bkarbrda, langdon, and I talked on the phone about scls.  I think that the SCL people are sufficiently flexible that we can work something out although it might be a compromise all around.
[09:13:38] * limburgher here
[09:13:47] --> LoKoMurdoK has joined this channel (~soporte at fedora/LoKoMurdoK).
[09:13:50] <abadger1999> geppetto is away today so we figured we'd defer the next SCL discussion until next week.
[09:13:55] <abadger1999> Cool, that's quorum.
[09:14:10] <limburgher> abadger1999:  Heart. . .not broken.
[09:14:20] <abadger1999> #topic https://fedorahosted.org/fpc/ticket/336 Please clarify the General Naming Guidelines for packages
[09:14:29] <abadger1999> mschwendt asked us to prioritise this.
[09:15:08] <tibbs|w> I think I'm on record as always preferring lower case.
[09:16:22] <tibbs|w> I certainly wouldn't ever want to require CraZyCasE just because upstream is a bunch of 13-year-olds.
[09:16:23] * abadger1999 prefers lower case and dash as well
[09:16:33] <abadger1999> yeah.
[09:16:35] <limburgher> I personally prefer lowercase.
[09:16:35] <RemiFedora> +1
[09:17:21] <abadger1999> what would we like to have happen here?
[09:17:43] <abadger1999> We could say "always use lowercase"
[09:17:50] <tibbs|w> At minimum let's explicitly allow that.
[09:17:53] <abadger1999> or we could give maintainers a little wiggle room
[09:17:57] <limburgher> With or without a grandfather clause?
[09:18:06] <abadger1999> with grandfather
[09:18:10] <tibbs|w> We pretty much always grandfather everything.
[09:18:35] <limburgher> Say " Always either use lowercase or follow upstream, with preference in that order?"
[09:18:48] <limburgher> Or is that too wiggly?
[09:18:55] <jsmith> limburgher: I like that :-)
[09:19:12] <tibbs|w> Certainly we don't want anyone adding uppercase when upstream doesn't.
[09:19:27] <limburgher> tibbs|w: Oh heck no.
[09:21:36] <tibbs|w> So, for case, I'm +1 to limb's proposal.
[09:21:38] <abadger1999> "use lowercase and turn underscores and into dashes unless there's compelling reasons to follow a different upstream convention.
[09:22:08] <RemiFedora> +1 for "turn underscores and into dashes"
[09:22:18] <RemiFedora> (especially when both are used in a package name)
[09:22:21] <tibbs|w> I'm +1 to that as well
[09:22:42] <tibbs|w> Basically I'm automatically +1 to anything that gets us closer to using all lower case and dashes.
[09:22:55] <abadger1999> evas-generic-loaders needs to have the dash piece clarified
[09:22:59] <abadger1999> <nod>
[09:23:35] <Rathann> +1 to abadger1999's proposal, but what would be a compelling reason?
[09:23:42] --> Martix has joined this channel (~martix at eduroam-147.fi.muni.cz).
[09:23:50] <Rathann> already existing package in another distribution
[09:24:12] <Rathann> explicit upstream recommendation
[09:24:14] <tibbs|w> We usually trust the review process for that kind of thing.
[09:24:42] <-- fuwei has left this server (Quit: Leaving.).
[09:26:09] <abadger1999> yeah, possibly.  or the package is just for a program you'd run on the commandline and it is /usr/bin/Uppercase
[09:26:43] <abadger1999> we've said the mod_* for httpd is allowed so following that convention for future apache modules would seem compelling.
[09:26:57] <tibbs|w> Another option is to force lowercase but explicitly permit Provides: CraZyCasE
[09:27:21] <Rathann> ugh
[09:27:26] <Rathann> I'd rather avoid that
[09:27:29] <limburgher> tibbs|w:  I'm not in love with that.  What does that do to yum search space?
[09:27:53] <tibbs|w> No clue, honestly.
[09:28:01] <tibbs|w> I think there are a couple of packages which do that already.
[09:28:54] <abadger1999> "Alwways use lowercase and dashes in package names [with the following, grandfathered exceptions for dashes LINK].  if a well known name with uppercase or underscores exists for the package you can use a virtual Provides for that."
[09:29:59] <RemiFedora> tibbs|w, yes, ex phpMyAdmin provides phpmyadmin
[09:30:21] <Rathann> or if upstream name contains _ it becomes -
[09:30:23] <-- paragan has left this server (Remote host closed the connection).
[09:30:33] <abadger1999> I think people can and will use virtual provifdes regardless...  the quesetion to me is whether we're comfortable with disallowing uppercase and underscores...
[09:30:43] <Rathann> I'm not
[09:30:47] <tibbs|w> I am.
[09:30:57] <RemiFedora> discourage seems anough
[09:30:57] <abadger1999> pointing out you can use virtual provides would just be to head off a frequent question.
[09:31:02] <tibbs|w> But anything that gets us closer is fine with me.
[09:31:05] <limburgher> RemiFedora: +1
[09:31:09] * abadger1999 with tibbs|w
[09:31:49] <abadger1999> So it sounds like to pass something today we want to go the discourage route.
[09:32:16] <abadger1999> Rathann or limburgher: Do you have some better wording than "compelling reason"?
[09:32:52] * SmootherFrOgZ is half-here
[09:33:04] <Rathann> hm
[09:33:22] <-- Martix has left this server (Ping timeout: 268 seconds).
[09:33:37] <abadger1999> If not,  I think we just vote to move the chains and clarify what those reasons could be if people still argue about it.
[09:34:13] <abadger1999> "move the chains" == get us closer to our goal (sorry, american idiom)
[09:34:16] <Rathann> "unless upstream project name doesn't follow the convention and there software is packaged in at least one distribution under that name"?
[09:34:29] <Rathann> s/there/the/
[09:34:44] <abadger1999> I'd still rather we used lowercase and dashes in many/some of those cases.
[09:35:41] <Rathann> it's a fight between KISS and least surprise principle ;)
[09:35:47] <abadger1999> <nod>
[09:36:05] <abadger1999> and kiss is least surprise to a different audience too :-)
[09:36:38] <abadger1999> ie: if I come from debian I may expet the name to match the debian name... but if I come from just Fedora I probably expect it to match other Fedora packages.
[09:36:56] <Rathann> *sigh* all right, I'll go with mandating lowercase and s/_/-/g just because I can't think of better reasons
[09:37:14] <Rathann> this is a big change for us though
[09:37:22] <abadger1999> Proposal: "use lowercase and turn underscores and into dashes unless there's compelling reasons to follow a different upstream convention."
[09:37:26] <abadger1999> +1
[09:37:42] <Rathann> that gets my +1
[09:37:46] <tibbs|w> +1
[09:38:09] <RemiFedora> +1
[09:38:12] * SmootherFrOgZ is also +1 here
[09:38:32] <Rathann> just fix the grammar before committing
[09:38:35] <abadger1999> limburgher: if you want to vote for the record.  we're +5 otherwise.
[09:38:37] <abadger1999> <nod>
[09:38:41] <abadger1999> s/and//
[09:39:02] <Rathann> s/'s/are/
[09:39:12] <Rathann> or s/reasons/reason/
[09:39:44] <abadger1999> <nod>
[09:39:48] <abadger1999> singularizes
[09:40:19] <abadger1999> "Use lowercase and turn underscores into dashes unless there's a compelling reason to follow a different upstream convention."
[09:40:48] <abadger1999> #info "Use lowercase and turn underscores into dashes unless there's a compelling reason to follow a different upstream convention." (+1:5, 0:0, -1:0)
[09:43:12] <abadger1999> #topic #352     BLAS and LAPACK packaging   https://fedorahosted.org/fpc/ticket/352
[09:44:01] <limburgher> Sorry, AFK, +1 for 336.
[09:45:02] <Rathann> not sure what bundling exception is fkluknav talking about
[09:45:05] <-- LoKoMurdoK has left this server (Ping timeout: 268 seconds).
[09:45:35] <Rathann> but haven't we mandated recently that ABI-compatible alternative implementations must be runtime interchangeable?
[09:46:01] <Rathann> well, unless upstream makes it impossible
[09:48:51] <RemiFedora> iiuc, openblas is a fork, which have change ABI, but not soname ? Ask upstream to change soname.
[09:50:09] <Rathann> it's a fork of gotoblas
[09:51:02] <abadger1999> Sounds more like the two projects are changing soname but not abi.
[09:52:17] <Rathann> no, actually they don't provide the standard libblas and liblapack
[09:52:42] <Rathann> while providing the blas and lapack symbols inside their libs
[09:53:02] <Rathann> atlas used to provide libblas and liblapack but doesn't anymore
[09:53:06] <limburgher> And the rational basis for this is. . .?
[09:53:16] * Rathann shrugs
[09:53:33] <Rathann> Susi Lehtola says it should be a compile-time decision
[09:53:33] <limburgher> Yeah.
[09:53:45] <Rathann> which blas/lapack implementation to use
[09:54:21] <Rathann> the issue is when one package uses one implementation and another, dependent one uses another implementation
[09:54:34] <Rathann> like the mess we had (still have?) with ssl libraries
[09:54:36] --> LoKoMurdoK has joined this channel (~soporte at fedora/LoKoMurdoK).
[09:55:26] <Rathann> but the Susi is not upstream although she does maintain a number of scientific packages
[09:55:45] <Rathann> and is my sponsoree
[09:55:58] <Rathann> *Sussi
[09:56:03] <Rathann> sorry for misspelling
[09:56:36] --> KageSenshi has joined this channel (~kagesensh at fedora/KageSenshi).
[09:58:15] <abadger1999> Hmmm... sussi mentions copying the MPI guidelines.  Would that work here?
[09:58:58] <Rathann> it would, but then we'd have builds of each liblapack/libblas consumer
[09:59:04] <Rathann> three builds
[10:00:24] <abadger1999> <nod>  But on one hand, that seems to be what upstream for these libraries is promoting.
[10:00:39] <abadger1999> since they're no longer providing separate liblapack/libblas
[10:07:25] <Rathann> both Kevin's and Sussi's arguments have merits
[10:08:30] <abadger1999> yeah.
[10:09:04] <abadger1999> Perhaps we should defer this to when spot is around since he kicked the mpi guidelines into their present shape
[10:09:27] <limburgher> abadger1999:  That's not a bad idea.
[10:09:56] * Rathann looks at packages which depend on libblas, liblapack and libatlas in F19
[10:10:24] <abadger1999> #topic https://fedorahosted.org/fpc/ticket/353 Update autodep filter guidelines to mention changes to rpm in F20+
[10:11:14] <abadger1999> Looks like RemiFedora found the guidelines that have filters and tested that the new way works.
[10:12:13] <RemiFedora> I think the descript in the ticket description is fine to be added in the Guidelines for explanation
[10:12:35] <abadger1999> <nod>
[10:12:44] <RemiFedora> and then we can open a bug against perl and ruby to have them to simplify their macros
[10:13:22] <abadger1999> Let's vote to update the https://fedoraproject.org/wiki/Packaging:AutoProvidesAndRequiresFiltering#Private_Libraries  guidelines with that description and open bugs against the packages providing the macros
[10:13:24] <abadger1999> +1
[10:13:55] <RemiFedora> +1
[10:14:15] <Rathann> +1
[10:15:39] <abadger1999> limburgher, tibbs|w, SmootherFrOgZif you're still around?
[10:16:03] <tibbs|w> +1
[10:16:05] <tibbs|w> Sorry, work.
[10:16:18] <abadger1999> no problem... we're running late too
[10:16:43] <abadger1999> okay, we're at +4 .  I'll record in the ticket and we can try to get the remaining vote there.
[10:17:01] <abadger1999> Let's try to get one more:
[10:17:12] <RemiFedora> 354 ?
[10:17:13] <abadger1999> #topic https://fedorahosted.org/fpc/ticket/354 PHP Guildelines small addition
[10:17:17] <RemiFedora> :)
[10:18:11] <abadger1999> This looks okay to me.
[10:18:13] <Rathann> looks reasonable
[10:18:15] <abadger1999> +1
[10:18:55] <Rathann> +1
[10:19:00] <RemiFedora> +1
[10:19:28] <limburgher> +1 for this one.
[10:19:38] <tibbs|w> +1 in theory; I'm guessing the pecl section of http://fedoraproject.org/wiki/Packaging:PHP#Requires_and_Provides would just change to match what the pear bits do.
[10:19:45] <limburgher> +1 to previous as well, I'll comment in ticket.
[10:19:52] <limburgher> Sorry i keep getting called away.
[10:20:14] <RemiFedora> tibbs|w, yes, exactly
[10:20:24] <abadger1999> Cool.
[10:20:41] <abadger1999> So that's +5 to 354 and to 353.
[10:20:50] <abadger1999> #topic Open Floor
[10:20:51] <RemiFedora> so I think we have +5 for 353 and 354
[10:20:59] * RemiFedora is too slow
[10:21:06] <abadger1999> anyone want to bring something up?  If not I'll close out for today.
[10:21:37] <Rathann> I saw #355 in the agenda
[10:21:47] <abadger1999> Rathann: yeah -- but we're over time.
[10:21:51] <RemiFedora> I'd like to find a way to avoid "fast" review....
[10:21:54] <abadger1999> Official this is jut a one hour meeting.
[10:22:05] <Rathann> right
[10:22:10] <Rathann> nothing from me then
[10:22:17] <RemiFedora> like review submitted / reviewed and approved in a few minutes....
[10:22:19] <limburgher> Same here.
[10:22:23] <abadger1999> oh one thing.
[10:22:43] <tibbs|w> I don't see any way around fast reviews at this point.
[10:22:52] <tibbs|w> Submit, someone runs fedora-review, says OK, done.
[10:23:32] --> inode0 has joined this channel (~inode0 at fedora/inode0).
[10:23:36] <abadger1999> FYI: This week, with the ideas about some body needing to approve scls, I thought that maybe we should create a new group composed like the FPC that approves SCLs and maybe bundled libraries as well.
[10:24:25] <RemiFedora> +1 for having a group working / approving SCL. We need a filter to avoid to much uneeded SCL
[10:24:37] <abadger1999> neither one is really how to package which is what we concentrate on but both need to have consistency, attention to precedent, and someone to write up the decisions in a more general form which makes unsuitable for fesco.
[10:24:40] <tibbs|w> I don't know if it should just e FPC that does that, though.
[10:24:58] <abadger1999> tibbs|w: yeah, geppetto was of the same opinion.
[10:25:01] * Rathann needs to review old logs to find out why he was convinced that we should go forward with the idea of SCLs
[10:25:28] <tibbs|w> The problem is that group would have to work really closely with FPC to avoid a disconnect between the guidelines for bundling/SCLs and what actually gets approved.
[10:25:40] <abadger1999> but spot pointed out to me once that the majority of FPC's open tickets are usually bundled libraries.
[10:25:46] <abadger1999> I think this would add to the same pile.
[10:26:04] <abadger1999> tibbs|w: I would turn over that particular criteria to them as well.
[10:26:59] <abadger1999> that criteria's question is: "when can you bundle/create an SCL".  whereas we try to stick to "how do you package something that bundles/an SCL".
[10:28:15] <abadger1999> But yeah -- we should discuss whether we as a group want to push that stuff off to another group or if we want to keep it to ourselves.
[10:28:43] <abadger1999> I'll open a ticket so we can discuss it in the main part of the meeting instead of when we're all anxious to get back to other things.
[10:30:05] <abadger1999> Okay, thanks for coming everyone!
[10:30:07] <abadger1999> #endmeeting
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: not available
URL: <http://lists.fedoraproject.org/pipermail/meetingminutes/attachments/20131017/f928a36d/attachment.sig>