Re: gnome-screensaver 0.0.23

[William Jon McCann <mccann jhu edu>]

Jaap Haitsma wrote:
> > What is the advantage of having it in gnome-settings-daemon? 
>
> Detecting if a keyboard or mouse is idle seems like a general service to
> me which several other programs might want to use. E.g. a disk
> defragmentation tool or a backup tool.

Since gnome-screensaver broadcasts on the session bus any D-Bus aware 
application is able to use these signals.

> > At the moment gnome-settings-daemon doesn't have idle detection.  It has
> > a very crude type of activity detection.
>
> Can´t we just move the code that handle keyboard/mouse detection you have
> in g-s to g-s-d?

Perhaps, but I think that creates far more problems than it solves.

> > In order to decouple idle detection from the screensaver you will need
> > to solve a number of tricky technical problems.  How do you handle
> > "inhibitors"?
>
> If we just have keyboard / mouse idleness detection in g-s-d, g-s can just
> handle screensaver inhibitors like totem like it is doing now, can´t we?

The inhibitors not only block the activation of the screensaver but also 
prevent entering the idle state.  I suppose this is somewhat debatable 
but it makes the most sense to me.  For example, if you are watching a 
movie or playing a video game you are not idle and you would not expect 
that power management to kick in.

> > How do you handle races between idle and screensaver
> > activation states?  What is the sequence of events when the lock screen
> > button is pressed?  How do you do idle activation warnings?  And so on.
>
> Since I´m not familiar with all the internals of g-s I can´t answer all
> these questions. But it sounds to me like g-s is also looking at CPU
> power? Should it do that? I personally think it shouldn´t it should just
> look at mouse and keyboard activity and have a list of inhibitors which
> can block the screensaver.

No, it isn't concerned with the CPU - only session user activity.

> > Seems more reasonable to make the typing break program gnome-screensaver
> > aware.
>
> I find it very weird that typing break and g-p-m should depend on an end
> user app (a screensaver) to know if the keyboard/mouse is idle. Sounds
> much more logical that keyboard/mouse idle detection is done some where
> deeper in the software stack.
>
> Later on we could also move stuff like network idle or cpu idle detection
> into g-s-d. Gnome power manager could use these for instance

Why are typing break and g-p-m lower/deeper in the software stack?

Jon