16:00:37 #startmeeting fpc 16:00:38 Meeting started Thu Jul 16 16:00:37 2015 UTC. The chair is geppetto. Information about MeetBot at http://wiki.debian.org/MeetBot. 16:00:38 Useful Commands: #action #agreed #halp #info #idea #link #topic. 16:00:38 #meetingname fpc 16:00:38 #topic Roll Call 16:00:38 The meeting name has been set to 'fpc' 16:00:45 Good morning... sitting in the back row today 16:00:50 ok :) 16:01:59 geppetto limburgher mbooth orionp racor Rathann SmootherFr0gZ tibbs|w tomspur: FPC ping 16:02:06 Howdy. 16:02:08 hey 16:02:10 morning 16:02:13 #chair tibbs 16:02:13 Current chairs: geppetto tibbs 16:02:16 #chair orionp 16:02:16 Current chairs: geppetto orionp tibbs 16:02:24 No tomspur today. 16:02:28 * geppetto nods 16:02:38 I just paste that from the page 16:03:01 I haven't finished writing up the python stuff. Seems more stuff will be changing soon. 16:03:28 Hi 16:03:34 #chair mbooth 16:03:34 Current chairs: geppetto mbooth orionp tibbs 16:04:37 Hmm, none of the other 4 are on IRC atm. 16:05:21 Seems to be the trend lately. 16:05:42 I guess people are enjoying the summer? 16:06:35 Maybe, it is grande vacance time of year 16:06:55 * geppetto nods 16:07:19 There's only one new ticket this week anyway: 16:07:22 #topic Schedule 16:07:26 https://lists.fedoraproject.org/pipermail/packaging/2015-July/010846.html 16:07:32 #topic #506 Guideline Draft: Service First-Time Setup 16:07:32 .fpc 506 16:07:32 https://fedorahosted.org/fpc/ticket/506 16:07:33 geppetto: #506 (Guideline Draft: Service First-Time Setup) – fpc - https://fedorahosted.org/fpc/ticket/506 16:07:34 It's 37 outside today, so to enjoy the summer I have to stay insidde. 16:08:07 Oh, nice, I get fedmsg notifications for stuff we do on IRC now. 16:08:14 More mail to turn off. 16:08:18 Ahh 98.6 16:08:22 That is warm 16:08:44 Americans and their units.... 16:08:52 Freedom units! 16:08:59 ROFL 16:09:14 google is also happy to tell me that 37 is the internal rectal temp. 16:09:24 Indeed it is. 16:09:34 Good to know... 16:09:51 When we say 98.6 we think we're being super-precise. 16:10:32 So this draft is two big pages. 16:11:08 The first one is about how to structure systemd units to make sure that self-signed certs get generated on demand. 16:11:39 Well, or anything else that needs to get setup once at the first service startup. 16:11:49 Like ssh keys. 16:11:57 Looks good to me 16:12:59 Note that the mentioned /usr/bin/sscg isn't there yet. 16:13:04 Or at least, it isn't in F22 yet. 16:13:26 sgallagh_afk: ping 16:13:26 geppetto: Ping with data, please: https://fedoraproject.org/wiki/No_naked_pings 16:14:06 I kind of understand why zodbot does that, but in at least 98%+ of cases I do those pings I want to punch it in the face and the other person didn't need more info. 16:14:07 I'm OK with this draft, though I haven't actually tried to implement it. 16:14:40 I'm somewhat confused by the second examplke of what shouldn't be considered system config. 16:15:12 Was waiting for sgallagh_afk to finish it before converting cyrus. 16:15:50 Oh, I guess it's saying that if the daemon/service itself generates the data then you don't need to move that generation into another systemd service. 16:15:52 geppetto: I assume he means any configuration that the daemon itself will create. 16:16:17 Could probably be clarified with a slight rewording. 16:16:50 As for the second page, do we want that as a guideline or just some tips page elsewhere on the wiki? 16:17:27 Because if we just say "use sscg" in the first page, the rest is kind of academic. 16:17:44 yeh the second part feels like it should go somewhere else 16:17:58 not sure where though 16:18:10 We can remove the "Do-it-yourself" bit from the first draft and pretend we never saw the second. 16:18:25 ha 16:18:44 Also, I don't think we need the sscg usage output. 16:18:52 I mean, anyone can run it and get that. 16:19:01 Well, assuming it's actually in the OS. 16:19:40 it's not there by default 16:19:52 I don't know if that's intentional. 16:20:09 We'll need to document the additional dependencies in any case. 16:20:29 So, my suggestions so far: 16:20:34 Nuke the last two sections. 16:20:45 Reword what's now the last section a bit. 16:20:52 it probably is … except I suspect that people won't find it because it isn't there, and will end up running some random openssl script from the internet instead 16:21:10 Add note about the additional dependencies required. 16:21:39 Also, how does sscg know where to put the generated certificate files? 16:21:58 In the example it somehow knows to put them in /etc/Pegasus. 16:22:01 I assume it does it in PWD 16:22:20 although I can't find sscg atm. … even in F21 16:22:36 I assume you'd have to look at F23 or rawhide. 16:22:40 is it just a radnom binary in tog-pegasus package? 16:22:52 No, it's supposed to be somewhere separate. 16:23:07 I meant F22, not F21 16:23:24 I assumed it'd be in latest stable release, if we had it in policy 16:23:52 I would assume he just hasn't pushed it yet. Maybe it's in testing. 16:24:07 Anyway, I can't figure out how it would know where to put the certs. 16:24:30 There's no specification of the working directory in the unit files and hardcoding that into sscg would be dumb. 16:26:35 weird 16:26:52 It's probably just an oversight; it needs an outdir option or something. 16:26:58 It doesn't matter much if we are just going to +1 everything before the special case section. 16:28:03 I wasn't suggesting removing anything other than "Easy implementation with sscg" and below. 16:28:13 The examples sure need to be in there. 16:28:27 yeh, the examples are fine 16:28:32 And of course they need to be correct. 16:28:55 I just wasn't sure about the special case bit … 16:30:28 I can hack on the draft a little bit and maybe talk to sgallagh_afk when he's back, but I can't think of anything else to say about this currently. 16:30:47 it seems like a lot of text to say "we'll generate self signed certs, but they might not help you anyway" 16:31:17 which I'd expect most packagers know 16:31:20 * geppetto shrugs 16:31:37 I'm happy to +1 that too, if you want. 16:32:09 mbooth: orionp: Any questions/concerns? 16:32:37 Not really. I agree with your points 16:32:52 No, actually this draft if pretty fine 16:33:26 I see your point about the "Special Case" section. 16:33:58 Maybe just say "you should use the sscg tool to generate self-signed certificates" or something in there. 16:34:15 Well I still can't find it in F22 16:34:26 So I'm not desperate to mention it :) 16:34:37 It is sufficiently complicated and security-sensitive that we should mandate using the tool if it's available. 16:34:45 * geppetto nods 16:34:46 But not having the tool currently is kind of a blocker. 16:35:08 Maybe we should just table this until the tool is actually in there. 16:35:31 Sure, I mean we can't pass it today anyway 16:35:33 sscg is in rawhide. 16:35:37 ahh 16:35:47 * geppetto looks in updates-testing 16:36:06 No updates that I can see. 16:36:28 Hmm, the repo is four months old. 16:36:36 yeh, whatprovides gives 0 results 16:36:49 which repo? 16:37:00 sscg in git. 16:37:13 It's branched for f21 and el6 (and newer). 16:37:20 But I guess it was never pushed. 16:37:39 Or even built for anything other than f23. 16:38:02 There isn't even a spec in the f21 or f22 branches, so.... 16:38:12 yeh, that's a bad sign :) 16:38:19 I guess we'll see what sgallagh says. 16:38:57 Yeh, koji says 2 builds both for rawhide/f23 16:39:00 I'm here now. 16:39:01 Sorry, had a family conflict 16:39:29 ok, give you a few minutes to catch up :) 16:39:41 This is the only real ticket today anyway 16:41:36 OK, yeah. I haven't built sscg for F22 or earlier yet because IIRC I only made it py3 compatible thus far. 16:41:53 ahh 16:41:56 I need to make it dual-stack before we should mandate it on Py2 default platforms 16:41:58 Something wrong with python3 in F22? 16:42:22 tibbs|w: Only that we're not supposed to be having the /usr/bin/ stuff require py3 16:42:23 I guess people would complain about dependencies. 16:42:42 I'm not familiar with that rule. 16:43:26 Maybe I misinterpreted the Python guidelines on that point 16:43:32 I guess the server folks might complain about having to pull in py3+modules just to generate a cert. 16:43:45 But yeah; I need to finish the dual-stack work 16:43:56 I don't think anyone's going to switch over existing services do this in F22 anyway. 16:44:06 /me nods 16:44:25 If it were me I wouldn't bother, but perhaps that's because I can never remember what I have to change to make things work in old python. 16:44:35 Well … it'd be nice to just merge any updates back if you change this in rawhide and then have a fix for something 16:44:51 "(12:38:12 PM) tibbs|w: [12:21:38] Also, how does sscg know where to put the generated certificate files?": The destination is a mandatory argument 16:45:04 Am I missing that in the example? 16:45:16 Or in the usage info (which we don't really want, but still....) 16:45:48 I swear it isn't mentioned in either. 16:45:58 --cert-file 16:46:08 and --cert-key-file 16:46:13 * geppetto assumes 16:46:18 Which... don't have directories in the example. 16:46:24 oops, the example should have had an absolute path 16:46:33 Ah, that makes more sense. 16:46:41 /me fixes 16:47:02 Otherwise it'll be relative to the $CWD, which would be wrong for a unit file 16:47:28 * geppetto nods 16:47:35 It _would_ be shorter to take the directory as an argument. 16:48:05 tibbs|w: True; the alternative would be to set CWD in the unit file and use that, I suppose 16:53:09 So what are the remaining questions? 16:53:49 The wording on "A few examples that should not be considered system-specific configuration: " is a bit confusing 16:55:08 I initially read it as "this stuff isn't configuration" … but it's actually "this stuff doesn't need a special service to create it" 16:55:24 Not sure how to reword it though 16:56:21 geppetto: Well, I was trying to differentiate "stuff that is configuration but not system-specific" 16:56:28 And yeah, words is hard :-P 16:57:05 Yeh, but often stuff created by the daemon will be system-specific … it's just that it's fine for it to be created by the daemon 16:58:01 Right 16:58:17 Maybe "A few examples that should not be considered for a new system-specific configuration service: " 16:58:24 This is attempting to solve problems like all of the cloud images having the same UUID :) 16:58:28 yeh 16:58:52 also just further down the long hard road of removing %post entirely 17:00:05 Wouldn't that be nice. 17:00:12 No more ldconfig. 17:00:42 geppetto: That would be the stealth benefit, yes :) 17:02:00 So … I think that was it for the bits before "special case" 17:02:23 Did you want to keep the last 3 sections, or put them somewhere else? 17:02:47 Afraid my time is up, but y'all look like you're on top of it :-) 17:02:51 geppetto: Well, considering the certificate case was one of the primary drivers of this, I think scrapping it is kind of problematic 17:03:11 And there's not enough of us for quorum anyway, I think 17:03:18 no, there isn't 17:03:31 Was going to vote anyway and put it in need more votes state 17:03:37 but we can just do it next week 17:04:05 sgallagh: If you could get sscg in F22 for next week that that'd be cool 17:04:19 geppetto: I'll see what I can do 17:04:37 My dance card already looks like I let my two-year-old color on it 17:04:42 Having it work but require py3 is probably fine … just so that packagers can see it and run it 17:04:49 Welcome to the club. 17:04:58 But yeah, if the dependencies get smaller later then great. 17:05:21 If they shout about wanting to use it and prefering py2 … you can deal with that when you get them (or never ;) 17:05:34 Fair enough 17:07:07 Feel free to #action me 17:07:52 #action sgallagh sscg should be in F22 before we publish policy saying to use it. 17:08:24 Also … do you call all your packages ss? ;) 17:08:45 geppetto: No, that was just an amusing coincidence 17:08:53 Self-Signed Certificate Generator 17:08:57 * geppetto nods 17:09:20 #topic Open Floor 17:09:28 Anyone want to talk about anything? 17:10:05 Right now I'm trying to stop fedmsg from spamming me every time the topic changes or something gets actioned. 17:10:37 "fallout" from the rawspeed bundling request was/is being talked about on the ML 17:10:56 it just really likes talking to you 17:11:17 tibbs|w: fedmsg just thinks you're a good listener 17:12:25 Well it's throwing HTTP 500s when I try to add a rule to shut it up, so I guess it really wants me to get those notices. 17:12:52 I've had relationships like that... 17:12:53 There we gol 17:13:26 Looks like I'm going to flock. 17:13:37 cool, see you there! 17:13:44 Oh, cool. It'll be nice to see you 17:13:48 Travel request signed and everything. Booked my room yesterday. 17:14:02 I don't think I met either of you at the last fudcons I attended. 17:14:30 Boston (BU), Raleigh, then Boston (MIT). 17:14:30 the last one I attended was arizona 17:15:15 tempe 17:15:37 Anyway … going to close in a minute, unless anyone has something to bring up 17:15:41 tibbs|w: I was at the MIT one, but I doubt we knew each other back then 17:16:01 Well, we'll see how it goes. 17:16:20 My wife is going to fly up and we'll spend a few days after things wind down. 17:16:27 * geppetto nods 17:16:34 I was hoping for Colorado Springs, I could have made that.... 17:17:02 My wedding aniversary is right in the middle of this one, so my wife is going too :) 17:17:45 Mine just passed. 19 years. 17:18:05 congrat 17:18:10 +s 17:18:23 Feels like (only yesterday|an eternity). 17:18:33 ha 17:18:53 Celebrating 11 years tomorrow :) 17:19:08 Seems like July is the month, then. 17:19:22 where will it be? 17:19:25 everyone likes summer weddings :) 17:19:38 racor: Flock? It's in Rochester, NY. 17:19:41 racor: rochester NY 17:19:46 Not far from Niagara Falls. 17:20:10 I've been to upstate NY before (Cornell) but never went to the falls. 17:20:19 tibbs|w: I think it's further than you think it is from Niagara 17:20:33 yeh, like 2-3 hours by car 17:20:42 Looks like arond 150KM or so. 17:20:51 lol with your non-freedom units 17:20:54 ;) 17:20:54 Was planning to drive for a few hours. 17:20:58 * geppetto nods 17:21:16 I have mostly switched my life over to metric. 17:21:28 Except for thermostats. 17:21:40 My car gets 40 rods to the hogs-head and that's fine by me 17:21:40 Don't want to give up the accuracy. 17:21:51 sgallagh: :p 17:21:56 14 years last week (sorry, lagging) 17:22:03 Well shit. 17:22:45 I'm 16 while at flock … so double digits for everyone 17:23:29 Anyway … I'll see you all next week. Thanks for coming. 17:23:35 Be well 17:23:42 #endmeeting