On Sat, Nov 09, 2002 at 11:59:29PM +0100, Julio Merino wrote:

> Hardware problems should be managed by the kernel. For example, if a media
> fails repeatly (like a wrong CD or a floppy), the kernel should be able to
> stop the read/write, send an error signal to the requesting process and
> reset the drive. What I mean is that the kernel should detect these error
> conditions and stop the processes. Isn't it?

absolutely

Not that it wouldn't be tricky under some circumstances, but it's all kernel
responsibility to behave properly when hardware timeouts and such.
Unfortunately it implies possible semantic changes and drivers would have
to be migrated I beleive

Matt