Kdbus Will Likely Be Merged Into The Kernel This Year

[phoronix]

Phoronix: Kdbus Will Likely Be Merged Into The Kernel This Year

Lennart Poettering has offered an in-depth look at the kdbus IPC system that's an implementation of D-Bus logic into the Linux kernel. Kdbus has been a work-in-progress for a while and is now finally becoming a reality...

http://www.phoronix.com/vr.php?view=MTYwMzE

[nslqqq]

Video link is broken

[Annabel]

(k)dbus is a cancer

[nlyan]

Video link is broken

There's a low quality preview here:

http://ftp.osuosl.org/pub/fosdem//2014/previews/fosdem/fosdem_2014/dv/H1308_Rolin/2014-02-01/11_47_08.ogv

I couldn't find a mirror of the full thing. Couldn't get the audio working either, but that could be pulseaudio playing up on my new laptop...

[log0]

Video link is broken

Video is not available yet afaik.

There is an awful quality preview version here: http://ftp.osuosl.org/pub/fosdem/201...1/11_47_08.ogv

[mark45]

There are good videos, though not (yet?) about kdbus, like here:

http://video.fosdem.org/2014/H1301_Cornil/Saturday/

Notice: The wayland video is by a french guy with a horrible french accent, so horrible sometimes it seems he's talking french.

[justmy2cents]

(k)dbus is a cancer

i think lack of any decent standardised IPC that would fit needs in post 2K era is way more cancerous than kdbus. there is a big difference between linux and other foss kernels. linux is monolithic and as such it never went trough process that other went when they first made IPC and then started fitting kernel+modules around it. linux IPC is at best in 70s era (not talking about userspace here). not only dbus can't be used in kernel right now to serve as "IPC to fit them all", nothing else can. dbus for kernel purpose is simply too slow as you can't do 0copy, not to mention roundtripping to userland is time costly. beside the fact that kdbus is not "hey, let's stack dbus inside". they are only fitting parts that make sense and leave most in userland.

kdbus simply makes sense if you want linux to progress. not only is lowering gap between non-monolithic kernels, it's also doing the same thing as FreeBSD, Hurd... do by default since creating decent IPC was their 1st item on "TODO" list. and main benefit that you get is dbus can suddenly be used way more effectively in userland too, since implementation is meaner, leaner and faster

saying kdbus is cancer is equal to saying FreeBSD... are cancer, since they use in kernel IPC that is more evolved than fifo...

[newwen]

Lennart shared a couple of kdbus features out on the horizon include sandboxing support, yielding CPU time to destination, priority inheritance, and priority queues.

Those are nice features for embedded systems.

Hoping to see kdbus merged soon!

[Teho]

saying kdbus is cancer is equal to saying FreeBSD... are cancer, since they use in kernel IPC that is more evolved than fifo...

What's this FreeBSD IPC system you are talking about? FreeBSD also uses a monolithic kernel.

[uid313]

Think if they used Intents, the IPC system from Android.

Isn't D-Bus a serialized data format?
Think if they used JSON or Google's Protocol Buffers or Apache Thrift (developed by Facebook) instead?

Doesn't Plan 9 from Bell Labs have any IPC system?
Considereding who is behind Plan 9, its probably good.