**** BEGIN LOGGING AT Sun Mar 29 02:59:59 2015 Mar 29 09:12:33 given that we really have low traffic here these days, would anyone oppose to bringing back the commit messages here? Mar 29 09:12:44 I'm kind of missing those Mar 29 09:19:23 dbus sucks :/ Mar 29 09:27:01 heh Mar 29 09:27:34 PaulFertser: what in particular? Mar 29 09:35:46 should we roll our custom IPC? (jk Mar 29 09:39:37 mickeyl: the whole history of FSO proves that. D-Bus was apparently defective by design :/. Without clear understanding of what and how and why should be done, without clear plan or clean implementation. The very fact gdbus was so broken that FSO just had to migrate to qgdbus (or whatever it was called) says a lot. It also is not really properly network-transparent, probably not endian-clean e Mar 29 09:39:43 tc Mar 29 09:39:46 etc. Just a big PITA in the end. Regarding custom IPC, you know, OpenWrt did just that with ubus, and for their constrained usecases they got a very nice and accessible functionality, serving them just fine. Mar 29 09:41:59 Now they claim dbus is broken because it should be implemented inside the kernel. Seriously? Mar 29 09:44:19 PaulFertser: hmm, i somewhat agree. having a bit more experience with true REST, these days I'd wish we had a normal REST interface for IPC, but that is properly a bit far fetched. Although it could facilitate the idea of FSO being something to easily experiment and tinker with Mar 29 09:44:27 Regarding FSO API, in retrospect, it looks like the idea to rely on signals was not that nice. Signals are not supposed to be reliable. Mar 29 09:45:13 (signals) Yes, however the alternative is that every client must also be a server, which hinders easy experimentation on the command line IMO Mar 29 09:45:43 OpenWrt mocked up a minimal IPC with json, and they work with it just fine. And it's helluva easier to implement a ubus service in C than that gobject mess that dbus requires. Mar 29 09:46:01 hmm Mar 29 09:46:04 that sounds pretty interesting Mar 29 09:46:13 * mickeyl makes a mental note to check that out ASAP Mar 29 09:46:21 Core components need reliable IPC, signals just don't cut it. AFAICT, having used FSO daemons through all these years. Mar 29 09:46:29 *nod* Mar 29 09:46:36 mickeyl: it's for embedded usecases, probably not that fancy for high-level languages etc. Mar 29 09:53:34 i see, well bindings can always be created Mar 29 10:06:58 "DBus, FreeDesktop, and lots of madness": http://gentooexperimental.org/~patrick/weblog/archives/2014-11.html#e2014-11-23T09_26_01.txt and a HN discussion: https://news.ycombinator.com/item?id=8648437 where people share their experience trying to contribute to FDO projects etc. Mar 29 10:07:18 thanks Mar 29 10:08:20 interesting that this only comes up many years after introduction of it, i guess the recent dispute about kdbus made people put their eyes dbus again Mar 29 10:21:49 The article isn't that accurate, but the whole impression seems to be correct: FDO is a mess, and always was. D-Bus, HAL, ConsoleKit and it seems like just about everything that came from FDO never got production-quality. Mar 29 10:22:49 Yes, I'm running dbus-daemon and console-kit-daemon right atm, and I guess most of us are, but that doesn't mean this software is worthy. Mar 29 10:24:21 CK tools do not even have a man page! Mar 29 10:39:31 For the reference, some ubus docs: http://wiki.openwrt.org/doc/techref/ubus Mar 29 11:09:27 that blog contains a bunch of nice articles Mar 29 11:09:44 (gentooexperimental) **** ENDING LOGGING AT Mon Mar 30 02:59:59 2015