CK: more paranoia, idle status, and creation-time

[David Zeuthen]

On Mon, 2007-02-19 at 18:34 +0000, Richard Hughes wrote:
> On Mon, 2007-02-19 at 10:40 -0500, William Jon McCann wrote:
> > The property change notification is emitted from the CK session
> > object.  In the case where the session is GNOME related this property
> > is set by gnome-screensaver.  So, I don't think there is much
> > advantage to gnome-power-manager not listening to gnome-screensaver
> > directly.  There are some disadvantages like adding more points of
> > failure, the fact that this property is not necessarily authoritative
> > (at the moment anything can use the SetIdle method - not just a
> > screensaver or presence monitor), and that we don't track the
> > lifecycle of the caller (don't handle if it calls SetIdle and then
> > dies). 
> 
> ConsoleKit runs as root and monitors sessions. Why not just make the
> idle stuff system wide?

No, no and no. The authority on whether a session is idle in GNOME is
gnome-screensaver.The information passed to CK is only advisory; IOW
g-p-m in a session can/should trust the screen saver in the same session
(for some definition of trust) but not necessarily CK in the system.

IIRC, CK have a notion of what information it passes about a session
that can be trusted (such as "is_active" and "what seat do I belong to")
and what can't (e.g. "idle time", "number of unread messages"). The
latter is supposed to only be used be login greeters for informational
purposes - e.g. to display how long time a session have been idle, to
display number of unread emails etc. etc.

      David