suspend-to-disk

[C. Scott Ananian]

On 7/9/07, Mitch Bradley <wmb at laptop.org> wrote:
> If you fill the FLASH with useful data, you cannot hibernate.

We could imagine alternate UIs, where hibernation was an optional,
rather than mandatory, feature.  This would allow us to fail to
hibernate under disk-full or large-memory conditions, which might
result in better average performance.  I'm not sure I believe this,
but it's worth thinking about.

For example, when the battery is low, we could try to hibernate, but
if that fails, ask the applications to save their data and cleanly
shutdown.  (Of course, this adds complexity and hard-to-reach corner
cases.)

Possibly a better alternative: we already know we need to incorporate
better 'available memory' feedback to prevent the user from opening
too many applications and invoking the OOM killer.   We could expand
this to include available hibernation space, so that Sugar would not
let the user open more applications then could be safely hibernated
using the available disk space.  This idea sounds nice, but I'm afraid
that actually computing the appropriate limits would be very
difficult.  (What about a program with a slow memory leak, that was
safe to open initially, but gradually expanded to the point where
hibernation was impossible -- and some time after that, invoked the
OOM killer.)
 --scott

-- 
                         ( http://cscott.net/ )