Two general thoughts

[david at lang.hm]

>_Big_ disclaimer: I'm a G1G1 owner and lurker who hasn't contributed
>anything yet. Just been lurking and learning. That said, one minor
>nit:
>
>2008-02-19T23:45:14 Ricardo Carrano:
>> - Suspend and resume should be a user command and much less
>> aggressive when it happens automatically.
>
>I'm not positive exactly what you meant by this, but if I've been
>understanding what I've been reading, the core devs are hoping to
>get very aggressive indeed with suspend and resume, to the point
>where the common use pattern of spending lots of time looking at the
>screen, reading and thinking, would sit there with the DCON, the
>EC, and (if active) the autonomous mesh side of the wireless as
>the only things powered on except when you hit a key, when it wakes
>up (in a couple hundred ms max), repaints, and goes back to sleep.
>The dream being lovely hours and hours and hours of ebook, or web
>reading, or anything else but active processing or input.
>
>-Bennett

when the resume gets fast enough that the suspend is transparent to the 
user (and the applications the user is running) super agressive suspend 
will be just fine.

for this to work the suspend/resume cycle needs to be speed up by a factor 
of 10 or so, and it needs to wake up for scheduled events so that apps 
polling will keep it alive

the problem is that right now the suspend is not transparent to the user, 
and we keep finding apps that break. and this is not just the presense 
server, but normal client/server apps like mail clients that want to sleep 
for a while between checking for new mail. In addition suspend seems to 
cause the system to loose track of the network it's on, break network 
connections, distract the user by dimming the screen and being slow to 
respond to keystrokes, and otherwise make things difficult for the user.

I am all in favor of the eventual goal, but I really disagree with the 
approach of setting it to sleep as much as possible now while these 
problems are so common. the problems need to be addressed or you will find 
that most people will learn how to do 'touch /etc/ohm/inhibit-suspend' and 
not realize that the problems ever get fixed (I've done it on one of my 
machines, and if I used the other frequently I'd do the same there)

David Lang