16:05:10 <andreasn> #startmeeting Cockpit meeting 2014-12-01
16:05:10 <zodbot> Meeting started Mon Dec  1 16:05:10 2014 UTC.  The chair is andreasn. Information about MeetBot at http://wiki.debian.org/MeetBot.
16:05:10 <zodbot> Useful Commands: #action #agreed #halp #info #idea #link #topic.
16:05:16 <mvollmer> .hello mvo
16:05:18 <andreasn> #meetingname Cockpit
16:05:18 <zodbot> The meeting name has been set to 'cockpit'
16:05:25 <zodbot> mvollmer: mvo 'Marius Vollmer' <marius.vollmer@gmail.com>
16:05:27 <andreasn> .hellomynameis andreasn
16:05:28 <zodbot> andreasn: andreasn 'Andreas Nilsson' <anilsson@redhat.com>
16:05:49 <mvollmer> ok, agenda?
16:05:55 <stefw> login page message
16:05:59 <mvollmer> * Warning on login screen
16:06:04 <andreasn> #topic Agenda
16:06:14 <andreasn> hups
16:06:16 <andreasn> :)
16:06:18 <mvollmer> * Branding?
16:06:27 <andreasn> * responsive design fixes
16:06:40 <mvollmer> * file API update
16:07:18 <mvollmer> * streaming, etc (stefw)
16:08:01 <andreasn> anything else?
16:08:23 <andreasn> ok, lets start with the first one then
16:08:25 <andreasn> #topic Warning on login screen
16:08:27 <mvollmer> yep
16:08:52 <andreasn> https://github.com/cockpit-project/cockpit/pull/1528
16:09:03 <mvollmer> so I wasn't paying attention much about #1528
16:09:15 <stefw> i wonder if we want to have some sort of warning when you're running as a non-stable release
16:09:27 <andreasn> so minimize the delta between the Fedora 21 stuff and master
16:09:28 <mvollmer> i would say so.
16:09:29 <stefw> see here for stable releases vs pre-release
16:09:30 <stefw> https://github.com/cockpit-project/cockpit/releases
16:10:10 <andreasn> so 0.27 was stable and 0.28 was pre-release?
16:10:11 <mvollmer> our current releases are more like snapshots
16:10:25 <stefw> andreasn, yes
16:10:40 <andreasn> is there any kind of flag for that?
16:10:55 <mvollmer> we could invent one
16:10:58 <stefw> sure
16:11:04 <stefw> we could also drive the documentation uploads off of it
16:11:13 <mvollmer> yeah
16:12:08 <andreasn> sounds good
16:12:58 <andreasn> should we continue that in https://github.com/cockpit-project/cockpit/issues/1075 then?
16:13:05 <andreasn> to invent a flag etc
16:13:12 <andreasn> and I'll close 1528 for now
16:14:02 <mvollmer> fine with me.
16:14:45 <stefw> up to you about closing 1528
16:14:56 <stefw> i guess someone else would need to take the pull req
16:15:00 <stefw> so yeah, closing does make sense
16:15:03 <mvollmer> (of course, we eventually migth want to think about versioning etc the components...)
16:15:04 <andreasn> #info we need a way to warn for pre-releases vs. stable releases. Discussion to continue on #1075
16:15:07 * stefw is sorry for the noise :S
16:15:36 <mvollmer> (and "about" dialogs...)
16:16:05 <andreasn> ok, closed
16:16:12 <github> [cockpit] andreasn closed pull request #1528: remove warning from login screen, fixes issue #1075 (master...sans-error-message) http://git.io/PCJ1-w
16:16:14 <andreasn> lets move on
16:16:26 <andreasn> #topic Branding
16:16:50 <mvollmer> so I was thinking whether we should make branding easier?
16:17:01 <andreasn> yes, that would be great
16:17:02 <stefw> i think we should
16:17:07 <mvollmer> or is it OK that Fedora carries a patch?
16:17:12 <andreasn> right now it's pretty hard
16:17:19 <stefw> Atomic is also carrying a patch :S
16:17:23 <stefw> and it's nasty
16:17:32 <mvollmer> so we need a branding API?
16:17:48 <stefw> hmmm, i think drop some files into the build tree
16:17:50 <stefw> before we run 'make
16:17:54 <stefw> would be such an api
16:18:03 <mvollmer> ./configure --with-branding=/dir/of/branding/bits
16:18:07 <stefw> yup
16:18:17 <mvollmer> ok.
16:18:20 <stefw> obviously details may vary once implemented
16:18:25 <stefw> but i don't think it should be a run time thing
16:18:28 <andreasn> I think there was an issue for this
16:18:30 <mvollmer> no
16:18:31 * andreasn looks
16:18:45 <andreasn> https://github.com/cockpit-project/cockpit/issues/702
16:19:10 <mvollmer> if the "API" is not powerful enough, F and A can still carry a patch additionally.
16:19:35 <stefw> yes especially if we don't want to spend eons figuring out every corner case
16:19:43 <stefw> i think that this would not be stable
16:19:47 <mvollmer> yes, that would be the concern with the API
16:19:47 <stefw> going forward
16:19:59 <stefw> we can adapt this to cover necessary cases as they come up
16:20:04 <stefw> and any branders are expected to follow along
16:20:05 <andreasn> right
16:20:06 <stefw> mostly us
16:21:20 <mvollmer> ok, what's the priority of this?
16:21:36 <andreasn> I heard Arch started packaging Cockpit
16:21:39 <stefw> depends how long it will take
16:21:50 <stefw> if it's a small bit of work
16:21:53 <stefw> then we should do it now
16:21:54 <andreasn> no idea how often they release
16:21:56 <stefw> to save us time in the long run
16:22:08 <andreasn> next release of Fedora is 4-6 months away
16:22:21 <stefw> yes, but every time i release for Atomic
16:22:27 <stefw> i blow time on these branding patches
16:22:48 <mvollmer> alright so you know the pain points! :-)
16:23:41 <mvollmer> stefw, so what needs to be done?
16:23:50 <stefw> i haven't thought much about it
16:23:51 <mvollmer> do we need to run sed on index.html?
16:24:01 <stefw> hmmm, interesting
16:24:02 <stefw> maybe
16:24:24 <stefw> it's a shame we have two layers of templating then
16:24:29 <stefw> build time and run time, in login.html
16:24:44 <stefw> but we probably can't avoid that
16:24:45 <mvollmer> we _can_ do the branding at run-time.
16:24:58 <mvollmer> that would also open the door for some other configuration
16:25:03 <mvollmer> that distros might want to do.
16:25:11 <stefw> my main concern is slowing down cockpit loading for no reason
16:25:12 <mvollmer> no idea what, though.
16:25:18 <stefw> there's already enough things that slow it down for good reasons
16:25:23 <stefw> so i want to avoid the unnecessary ones
16:25:50 <mvollmer> can we do the branding via l10n?
16:26:01 <mvollmer> ugly...
16:26:08 <stefw> yeah, that does seem ugly
16:26:18 <stefw> and it's also images, image paths etc.
16:26:29 <mvollmer> same problem, only more obscure
16:26:39 <mvollmer> running sed on the po files
16:27:02 <mvollmer> or having a language just for the branding words
16:27:11 <andreasn> so in the fedora patch I put the images beside the regular ones, but it might be better to just replace the images
16:27:13 <mvollmer> anyway, so, do we spend time on this now?
16:27:50 <stefw> well every week we spend about 10 - 15 minutes on it
16:27:51 <mvollmer> we should at least document current practices, even if they are ugly
16:27:54 <stefw> so i guess not that much
16:28:01 <stefw> the current practice is just a patch
16:28:09 <stefw> i switched to using git to patch
16:28:12 <stefw> and that makes it much nicer
16:28:15 <stefw> since i can patch binary files in
16:28:17 <mvollmer> right
16:28:20 <stefw> rebasing the patch is a pain when it breaks
16:28:26 <stefw> but we haven't changed the login screen much lately
16:28:34 <stefw> although a commit like 1528, would cause pain
16:28:42 <stefw> because then the branding needs to be rebased and manually fixed
16:28:57 <stefw> but i don't mind dealing with the pain a little longer
16:29:07 <stefw> there are more pressing things
16:29:11 <mvollmer> ok
16:29:11 <andreasn> all right
16:29:45 <andreasn> I'll put it on the trello somewhere
16:29:56 <andreasn> and then we can take on that at some future point
16:30:01 <stefw> ok
16:30:17 <andreasn> ok, next topic
16:30:19 <andreasn> #topic responsive design fixes
16:31:10 <andreasn> https://github.com/cockpit-project/cockpit/pull/1537
16:31:16 <andreasn> https://github.com/cockpit-project/cockpit/pull/1536
16:31:32 <stefw> are they ready for review?
16:31:35 <andreasn> yes
16:31:43 <stefw> ok, i'll review them
16:31:45 <andreasn> thanks for the reviews so far
16:31:49 <stefw> no wortries
16:31:59 <stefw> i hope i haven't been too picky ...
16:32:20 <andreasn> not at all
16:32:21 <stefw> but i found myself avoiding cockpit :S
16:33:05 <andreasn> yeah, and the current behaviour stinks, doesn't feel stable at all
16:33:11 <andreasn> so it will be good to have these fixed
16:33:26 <andreasn> https://github.com/cockpit-project/cockpit/issues/1522 is still in the works, will create pull request soon
16:33:34 <stefw> cool
16:33:38 <stefw> thanks for knocking these out
16:33:41 <stefw> i think there's a few more
16:33:41 <andreasn> np
16:33:42 <stefw> will file
16:33:47 <andreasn> thanks!
16:34:07 <andreasn> ok, next topic
16:34:12 <andreasn> #topic file API update
16:34:38 <mvollmer> right
16:34:55 <mvollmer> i opened a pull request, but it still "needslotsofwork"
16:35:09 <mvollmer> i am quite happy with the general API, though
16:35:32 <mvollmer> i'll start writing tests now, and then make the code a bit nicer.
16:36:05 <mvollmer> (upps, phone,  sorry)
16:37:04 <stefw> i haven't taken a look yet, but will shortly
16:39:32 <mvollmer> (ok,back, sorry)
16:39:47 <mvollmer> open issue: binary files and text encoding
16:39:58 <mvollmer> binary files should be straightforward.
16:40:00 <stefw> yes
16:40:07 <mvollmer> i would love to ignore text encodings
16:40:12 <stefw> we should
16:40:14 <stefw> we do everywhere else
16:40:16 <stefw> right now
16:40:22 <stefw> and just coerce to utf-8
16:40:28 <mvollmer> and just say, you get strings, and the file is written in the right encoding.
16:40:38 <mvollmer> where right == UTF-8 for now.
16:40:46 <stefw> yes, i think that's correct for now
16:40:53 <stefw> it's identical behavior to all our other channels
16:41:10 <mvollmer> and then we can make the bridge look at LANG etc and dtrt.
16:41:25 <stefw> perhaps
16:41:31 <stefw> would need some seriously good reasons
16:41:40 <mvollmer> but I really would like to keep that can closed and just force everything to utf-8
16:41:45 <stefw> yeah
16:41:53 <mvollmer> filenames also.
16:41:58 <stefw> yup
16:42:27 <mvollmer> ok, I'll frst write some tests and then add the binary handling.
16:43:48 <andreasn> next topic?
16:43:53 <mvollmer> yep
16:44:12 <andreasn> #topic streaming, etc
16:44:27 <stefw> so this comes from the http work
16:44:30 <stefw> which is now ready for review
16:44:48 <mvollmer> ok
16:44:50 <stefw> basically taking the docker http code and generalizing it so that we can use it with kubernetes and openlmi
16:44:51 <stefw> and so on
16:45:06 <stefw> and doing this, unfortunately, highlighted several weaknesses in the protocol
16:45:14 <stefw> which marius and i discussed last week a bit
16:45:25 <stefw> some of these fixes have been merged, others are in pull requests
16:45:52 <stefw> one big change is the way handle end of file
16:46:01 <stefw> for example to a process, or an http request
16:46:08 <stefw> another one is how we buffer data
16:46:34 <stefw> slowly but surely this has gotten more efficient, and i'm pretty happy with how it's done in #1545
16:46:56 <stefw> i also did a bit of research on higher performance throughput of data
16:47:01 <stefw> for things like VNC, SPICE etc...
16:47:20 <stefw> and last week on friday had a proof of concept where i connected Cockpit to a Virtual Machine display
16:47:36 <stefw> and was driving a RHEL VM via a cockpit component :)
16:47:51 <stefw> this wasn't too much work
16:48:02 <stefw> and you can see the resulting code changes in #1540
16:48:20 <stefw> i'll try to make an example which uses this as a proof of concept
16:48:30 <stefw> i know we don't have the design for listing or driving VMs
16:48:52 <stefw> the basic idea for the high performance throughput is to open additional WebSockets for those use cases
16:48:53 <mvollmer> soooooon.....
16:49:45 <stefw> a lot of this work is so we can stabilize the protocol
16:49:49 <stefw> i know this seems premature
16:49:58 <mvollmer> i don't think so.
16:50:06 <stefw> but since i'm not some kind of genius guru, many of the ideas that i implement need lots of vetting and t esting
16:50:14 <stefw> for example the package stuff needs some tweaking now that we've had experience with it
16:50:51 <mvollmer> we will always have ideas for improvements :-)
16:51:02 <stefw> yeah, but once the protocol is stable, we have to be much  more conservative
16:51:06 <stefw> especially to do with packaging
16:51:08 <stefw> and stuff like that
16:51:16 <mvollmer> and let's be happy that right now we have the space to look at those
16:51:24 <stefw> right
16:52:14 <mvollmer> so, super work, stef! :-)
16:52:35 <stefw> thanks
16:53:03 <mvollmer> i _am_ looking forward to reaching a stable point again with our architecture and protocols, but no need to rush that, I would say.
16:53:33 <mvollmer> t will be all kinds of awesome to use the new features to refresh the storage, for example.
16:54:29 <andreasn> sounds cool
16:54:34 <stefw> mvollmer, yes that'll be cool
16:55:08 <mvollmer> are we done :-)
16:55:24 <andreasn> I guess
16:55:25 <andreasn> so
16:55:34 <andreasn> #endmeeting