Re: New modules in 2.14

[David Zeuthen <david fubar dk>]

On Jan 20, 2006, at 7:43 AM, Alan Cox wrote:

> On Gwe, 2006-01-20 at 09:05 +0000, Matthew Garrett wrote:
> > > currently has no mechanism for making a distinction between  
> > > background
> > > users and the user that currently 'controls' the machine.
> >
> > I don't think hal's the right layer to make that distinction. I'm
> > working on implementing it at the dbus level.
>
> I question the existance of such a distinction. Take an HP box with  
> four
> monitor heads and four keyboards. Who is "in control".

In this case you'd have four copies of g-p-m, one for each
console. Only one copy obtains the right to manage the
system power, this happens through the PolicyManager as
described here http://blog.fubar.dk/?p=63 ... e.g. the call
ClaimPolicyService() will return FALSE for everyone but
the first one to ask.

Thus, It is up to g-p-m to ensure that only the copy with the right
to manage system power actually does - this does include the
losing copies to somehow communicate to the winning copy
that their display is idle and they're ready to suspend caused
by inactivity. The winning copy will thus invoke the suspend,
due to inactivity, if and only if all sessions have said they are
inactive. How g-p-m does this is up to g-p-m.. I don't see the
need to provide a framework for things like this..

All copies of g-p-m will still manage devices specific to the
console, e.g. the display via DPMS on X. This probably also
includes devices we know are specific to consoles (say,
a subtree of all USB devices.. we can do this) but right
now there's no good answer on run-time power
management on Linux at least so it's not something
that HAL or g-p-m currently does. It's definitely something
we can and will do in the future. Waiting on kernel people.

Yes.. someone needs to code up the support I rambled about
plus add it to g-p-m (and make sure the g-p-m maintainer
agrees so to :-). Right now it's just ideas to support various
not-so-interesting-but-still-important scenarios. For non-corner
cases I think g-p-m is doing pretty well and we can all thank
Richard for that.

> Or consider a
> tape monkey in the server room.

It's busy attempting to write Shakespeare on a good old
fashioned type writer? :-) ... no, seriously, care to clarify?

    David