IRC Logs for #circuits Friday, 2014-03-28

pdurbin~8:20pm here :)00:18
prologichi pdurbin00:22
c45yprologic: got the irc bot working, just figuring how to start a web at the same time and hook all the events00:50
c45yI think I'm going to have to use websockets00:50
prologichave a look at for an example of this01:01
prologicit comines an irc client and web server01:02
c45yoh wow thats simple01:15
prologicit’s meant to be :)01:18
prologicc45y: wanna do lunch/beer soon-ish?01:19
c45yI do feel like pub lunch01:19
c45ylemme ask around the office01:19
prologicI could do a pub lunch if you’ll help me descide what to have — i.e: I can’t see the enu :)01:21
c45yhaha np01:24
c45ynow good for you?01:24
c45yok im heading there now, cya down there01:25
c45yblack shirt if it helps find me01:25
prologicme too01:25
prologicomfg we just ran into an issue with case sensitive autocomplete ajax autocomplete widgettry in our plone app and the bug was fixed in colelctive.z3cform.widgets only a year ago!03:33
prologicffs why do we keep doing this03:33
prologicsuch quality my god :(03:33
prologicso now my plone project is using yet another (we have dozens now) of a git bleeding edge dep03:33
c45ytbh, still better than sharepoint03:41
prologicyes okay I’ll give you that one03:42
prologicbut I am seriously sick and tired of using bleeding edge crap03:42
prologicit’s okay when it’s just one dependency03:42
prologicbut once you started getting into hunderds of them03:43
prologicand you start having a dozen of those bleeding edge03:43
prologicthings break horribly03:43
c45ymigrate it to something else03:43
prologicin ways you cannot possible imagine03:43
c45ydon't tell anyone03:43
prologicloose coupling comes with a cost03:43
prologicsometimes a rather high cost03:43
prologicdude lol if I had the time/energy outside of work03:43
prologicI’d have already re-designed and re-implemented TerraNova (this stupid Plone project) to something far better by now03:43
prologicwith recent RESTful API(s) and a nice ssext Bootstrap UI03:44
prologicwith 1/10th of the no. of dependeices03:44
prologicand 10x better performance and scalability03:44
prologicbut hey03:44
c45ywhats stopping you?03:44
prologicwhtaever right :)03:44
c45yyour project lead :P03:44
prologictime/energy outsode of work03:44
prologicproject lead on a project that’s a) over budget and b) over shcedule03:44
prologicand don’t get me wrong03:45
prologicI’ve already started such efforts outside of work03:45
prologicWanna see a demo of this running?03:46
prologicIt doesn’t do anything obviously as none of the functionalirty is there yet03:46
prologicbut the layout/structure of the app and ui is there03:46
c45ynot just at the moment03:46
c45ysilly friday productivity03:46
Worksterah the joys of replacing screens on iphones03:47
Worksteri hear ya on the software bugs03:48
prologicyeah tell me about it03:59
prologicbut OTOH you’ve given me a little motivation to continue my efforts outsid eof work to rewrite this stupid project04:00
prologicor at least get a quick prototype up and running04:00
prologicbtw Romster, Workster: if you focus on your efforts on getting a Docker environment up and runnig at home or on DO or whereever you want04:11
prologicI plan to build a Docker image for sahriswiki that you can “just spin up"04:11
prologicoh ffs Plone is so slow05:28
prologicand overly complex05:29
prologicso is our server05:29
prologicour server suffers massive Disk I/O performance problems05:29
robert_I hate d-bus05:31
prologicI honestly don’t even know what it does05:36
prologicI believe I have it running on my desktop at home (CRUX)05:36
prologicbut f’d if I know what it does :)05:36
prologicI think I run hald too to make some warnings go away05:36
robert_it's your init replacement.05:44
prologicno it’s not :)05:45
robert_oh wait05:45
prologicD-Bus is used for eventing in Xorg05:46
prologicand HAL is the Hardware Abstraction Layer for Xorg05:46
prologicthey are used by several applications and such05:46
robert_and systemd05:46
prologicsome of which I don’t actulaly use05:46
prologicyes and systemd is a evil monstronsity that most of the moajr distros have switched to05:46
prologictihgtly binding too many things into one05:46
prologicall of a sudden anything you write against systemd05:47
prologiccannot use any other init system05:47
prologice.g: CoreOS’s fleet component is tightly bound to systemd05:47
robert_I don't mind unifying the init systems05:47
robert_init is one of those things that even we shouldn't touch.05:48
robert_unless you feel like doing LFS05:48
robert_(which, no thanks.)05:48
robert_I just bleh; it's a real PITA to deal wiht05:49
robert_I just bleh; it's a real PITA to deal with05:49
robert_(DBus, I mean.)05:49
prologicNo I disagree05:52
prologicbecause I come from a background where I like control over things05:53
prologicif you have to touch it (and you will)05:53
prologicsomething that’s tightly bound and overly complex05:53
prologicand introduced all sorts of facny configurations and coencpts05:53
prologicis hard to maintain05:53
prologicsystemd is exactly that05:53
prologicinstead what I’d rather see05:53
prologicis a api system built on top of existing init systems05:53
prologicwith support for multiple backends05:53
prologic(i.e: backends == init systems)05:53
prologicand on another note05:54
prologicsystemd should never have decided to mix in udev05:54
prologicthat’s just a stupid horrible idea05:54
prologicbecause now you’re tightly bound to udev as your device management system as well05:54
prologicyes I agree - in an ideal world05:54
prologiceverything would just work05:54
prologicand you’d have the one init system05:54
prologicthe one dev management system05:54
robert_Some things I like to touch, but I don't like making server systems unbootable as a result of it :p05:54
prologicthe one remote api on top of init05:54
prologicreality doesn’t work like that05:54
prologicbeing true to the UNIX philosophy05:55
prologicwhich I don’t think I will ever move away from05:55
prologicComponent decoupling and loosely coupled sytems05:55
prologic_are always a good thing_05:55
prologicnot just loosely coupled components in a software05:55
robert_They're trying to standardize some of the lower-level parts.05:55
prologicbut also loosely coupled components in s aystem05:55
prologicstandards are great05:55
robert_For better or worse, init sucked hairy goat balls.05:55
prologicexcept when the standard becomes a dictated implementation05:55
prologictake a look at Python for example05:56
prologicPython is a specification for a language05:56
prologicit has many implementations05:56
robert_That's true.05:56
prologicthat are all compatible with one another05:56
prologicbut have added things on top05:56
prologicCPython is the python implematnaion that I use most of the time05:56
robert_well varying degrees05:56
prologicout of choice and necessity05:56
robert_you have the oddballs like stackless python05:56
prologicbut the moment you start dictating a standard as an implemtnation05:56
prologicyou bacjk yourself and others into a corner05:56
prologicCPython, PyPy, IronPython, Jython and Stackless are all good implentations05:57
robert_well then python backed itself into a corner because it has an implementation.05:57
robert_though systemd could have been part init replacement, part system interface library, sort of like how you had a usermode interface to the NT SCM06:00
robert_so long as you conformed to the specification, you could be a compliant implementation06:01
robert_We can agree on that at least, right?06:01
robert_oh btw06:02
robert_so we should work on that admin app framework I was talking about last night, lol06:02
prologicno Python never backed itself into a corner at all06:17
prologicIt’s okay to have a “reference implementation"06:17
prologicand that’s exactly what CPython is06:17
prologicand  BTW06:17
prologicCPython and PyPy are almost identical in every way06:18
prologicFWIW I do not see systemd as a specification for an init system06:18
prologicnor even a specification for a device management system06:18
prologicor even a specifciation for a remote api to an init system06:18
prologicnamely because a) it’s three things in one tightly bound and coupled and b) there are no specificaitons06:19
prologicbut yes I do agree with you06:19
prologicsystemd would have done all Linux distros a favor by firstly designing a specifcation for an init system06:19
prologicand a specification for an api06:19
prologicbut they didn't06:19
prologicand now it seems like it’s being forced down our throats by the likes of Debian, Ubuntu and RHEL/CentOS06:20
prologicas well as folks like CoreOS (but they ahve good reason to use it)06:20
prologicand to be fair (and honest) I quite like what CoreOS are doing with Docker06:20
prologicand how they have nitegrated systemd nicely into their whole setup06:20
prologic-but- OTOH they are still building a tightly boudn system06:20
prologicthe features of CoreOS you could never port to anything else without “a lot of work and oeffort"06:21
prologicand what’s this about an admin app framework?06:21
prologicsorry I must have missed that… :)06:21
prologicc45y: got the code for your new bouncer somewhere handy?06:23
Romsteryeah systemd i hate it.06:33
Romstergo eudev06:33
Romsteri'll try docker again this weekend it's been a pain but to be honest i have't bothered todo any more on docker or my version sort that i really want to get done.06:34
robert_systemd cold still have become the dominant implementation, so long as it allowed for others.06:34
Romsterrobert_, the realhellson eh :D06:34
Romsteri just hope the kernel dones't remove support for others to use there own init wiht eudev06:35
Romsterand then the gtk3 crap that's going on.06:35
Romsterwith the ABI breaking every release for themes and stuff.06:35
robert_I can't ever see Linus going along with that.06:35
Romsterlinux is about molding it to your own use. why lock us into only one init system06:36
Romstersounds like windows06:36
robert_init was old and crusty; parts of Windows Server were running circles around linux in the server arena.06:37
prologicthe kernel will never adopt such things06:37
prologicbecause it’s outside of the kernel’s scope06:37
Romsterhow.. you can run runit sysvinit any other init.06:37
robert_Of course not.06:37
prologicthe kernel is the kernel06:37
prologicand needs something to start06:37
prologica process to spawn06:37
prologicthat’s typically by convention called “/sbin/init"06:37
Romsternow if they remove shit in the kernel but only keep systemd stuff were screwed06:37
prologicor simply “init"06:37
prologici.e: init=06:37
prologicit will never change (unlikely)06:38
prologicpretty much all kernels are designed this way06:38
robert_particularly in services management. Linux had no notion of "services" and had no real userspace support for virtualization or containerization outside of not-well-supported libraries like libvirt.06:38
prologicwhetehr monolithic or micro06:38
Romsterdynamic node creation06:38
prologicRomster: give the latest docker port a try06:38
prologicI have made fixes to the kernel checking script, etc06:39
prologicand now that it’s at 0.9.106:39
prologicit’s quite rock solid for me06:39
robert_I like the advances that systemd is making; I just wish they'd have drawn up a spec first.06:39
prologicI’m consdering moving all my servers apps and serivces to docker soon06:39
prologicas I consdier it production ready and stable06:39
prologicrobert_: so do i06:39
prologicon that note06:39
Romsterbinary log file no thanks06:39
prologicI wouldn’t mind impelemting an API06:39
prologicdesigning a specification for o ne06:39
prologicand writing a reference implemtantion06:39
robert_(and let's be serious; udev? really? You should have exposed a storage api via DBus)06:39
prologicfor _any init system_06:39
prologicand yeah06:40
prologicthere are several things I will never accept about systemd06:40
prologiclike Romster said06:40
Romsterwell they are doinga bus in the kernel now?06:40
prologicbinary log file06:40
prologicget lost06:40
robert_You know, that'd actually be really bad-ass06:40
prologicit just makes things harder than it has to be06:40
prologicI would do it too06:40
robert_take hardcore design aspects of Windows (their heart was in the right place, just not their heads.)06:40
prologicsimply because I’d like to see another implementation of what CoreOS are doing06:40
Romsternext they'll write some binary blob and compile some abstraction layer around systemd06:41
prologici.e: a distributed init system06:41
prologicCoreOS is basically06:41
robert_(Services, Storage, Authentication) and roll it all into a singular, all-encompassing spec.06:41
prologiccustom gentoo ports + systemd + etcd (distro agnostic fortuantely) + fleeet (tightly bound to systemd)06:41
prologicso I’d like to design a sepc for a remote api on top of any init system06:41
prologicthen deisgn a new fleet-like system that uses this api06:41
prologicto build a distbruited init system06:42
prologicthat can use any init backend06:42
prologicincluding sysvinit06:42
prologicbecause then I can easily and trivially turn CRUX into a distirbuted Docker cluster06:42
prologicand anyone else can turn their favourite non-systemd distro into a Docker clsuter too06:42
robert_the problem is you need pid 1 to be useful for anything.06:42
prologicby using teh same set of decoupled tools/daemons06:42
prologicNo you don’t :)06:42
prologicthat’s the beatty about loose coupling06:43
prologicpid 1 can be syvinit’s init for all I care06:43
prologicand I don’t care :)06:43
robert_I see.06:43
prologicwhat’s more ...06:44
prologicI can actually implement references implemtantions of all this06:44
prologicin Python06:44
prologicthat can support any init backend system06:44
prologicstart with sysvinit — the simplest06:44
prologicthen support upstart, runit and systemd06:44
prologicpretty much covers all bases06:44
prologicand you’re done06:44
prologicmake it a nice HTTP/RESTful set of API(s)06:44
prologicand then re-design and re-implement a new fleet-style tool on top of that utilzing said API(s)06:45
prologicand voila06:45
prologicyou’re golden06:45
prologicDistributed Docekr on any and I mean any UNIX/Linux platform that Docker supports is then possilbe (without CoreOS and tightly boudn/coupled tools)06:45
prologicanyway home time06:46
prologiccyas when I get home06:47
robert_I need sleep.07:17
Romstersounds good but also sounds like alot of work07:50
*** Ossoleil has joined #circuits08:08
prologicRomster, what sounds like a lot of work?09:09
Romstermaking an init abstractiion layer API09:30
Romsteralong wiht your work and your lack of time to do circuits09:30
Romsterand everything else.09:30
*** Workster has quit IRC10:10
prologicwell sure12:41
prologicbut it's not a complex problem to solve I don't think12:41
prologicjust time/energy/effort required12:41
prologicit's 12am14:12
prologiclet's see if I can't build a simple dns caching server in circuits14:12
prologiceriknw, morning :)14:12
prologichow's this?14:42
prologicdon't you just loving /ping #channel14:42
ircnotifier19c87a23cb53 by prologic: Added a simple little dummy DNS Server Example that responds with an A record given any query of any type. Great integration example with dnslib14:54
ircnotifiera71d91f0e2d9 by prologic: Removed sync Fabric Task. We'll find another way to synchronize the Github mirror. Perhaps a hook?14:54
prologicg'night all15:01
*** REV117 has joined #circuits16:01
*** REV117 has quit IRC16:12
*** Ossoleil has quit IRC17:56
ircnotifier5352e23aa7eb by prologic: Updated: tox 1.7.0 -> 1.7.1, Fabric 1.8.2 -> 1.8.321:17

Generated by 2.11.0 by Marius Gedminas - find it at!