Update on DeviceKit

[Matthew Garrett]

On Wed, May 07, 2008 at 01:23:41PM -0400, David Zeuthen wrote:

> By the same token should we also care about voltage / power management
> on the GPU etc? Of course, not, leave that to the _driver_. Which is X.

No, we've learned that lesson. Policy decisions belong in userspace, but 
the mechanism belongs in the kernel - you really don't want scheduling 
latencies hitting you when you're playing with dynamic power management 
states. Think cpufreqd compared to the ondemand governor. You can argue 
that policy decisions should be made through the X server, but at that 
point it's really just an argument about who's echoing a value into 
sysfs.

> Omitting a backlight interface on o.fd.DeviceKit.Power also has a lot to
> do with keeping things simple (or simpler) and make it easy for people
> new to hacking on Linux on where to fix bugs / fix their system. It's
> _confusing_ the way it is right now. And that confusion is hurting us
> insofar that people wanting to contribute / fix their system waste time
> looking in the wrong place.

There's nothing simple about backlight control. Anyone who claims 
otherwise hasn't spent enough time dealing with the hardware...

-- 
Matthew Garrett | mjg59 at srcf.ucam.org