Suspend and NetworkManager

[David Zeuthen]

On Wed, 2006-01-11 at 16:17 +0100, Jaap Haitsma wrote:
> >>This seems just one 
> >>extra layer which is not really necessary.
> > 
> > 
> > But it is necessary in order to do this. What's the problem?
> > 
> There's no problem. It's only weird that we do hibernate/suspend in HAL 
> and sending "wake"/"sleep" signals in gpm. Since both of these are 
> heavily related I'd expect they would be in the same module.
> 
> I.e. sending the signals and implemeting the hibernate/suspend method 
> both in gpm or both in HAL
> 
> Is it a problem to both in HAL?

Yes, it violates layering and introduce policy into HAL as in how long
the timeout should be for applications requesting a delay. (This
shouldn't be more than 2-5 seconds - I think that MS changes this from
infinity to 2 seconds going from Windows XP to Vista btw).

It's also more correct doing it in g-p-m as we then have access to a
display where we can put notification bubbles a'la "the application Foo
is preventing standby" though that may or may not be a good idea from a
UI point of view. 

Another thing is that g-p-m has a lot more information - g-p-m *knows*
whether the suspend request came from 1) closing the laptop lid; 2)
inactivity; 3) someone pressing the sleep hot-key; 4) request from other
app in the session; 5) battery critically low. For example for 3) it
might make sense to put up a bubble, whereas it don't make sense for at
least 1) and 5).

We shouldn't overengineer things though :-)

Ideally the interface that g-p-m publishes on the session-bus should be
standardized on freedesktop.org so we'd have interoperability between
for example KDE and GNOME.

Cheers,
David