On Mon, Jul 16, 2001 at 02:22:55PM -0700, Matthew Jacob wrote:
> > > Still, it's the target driver who really should be deciding how important a
> > > selection timeout is.
> > 
> > I can't see why, sorry. Can you give an example of why selection timeout
> > should be handled differently for 2 different type of device ?
> 
> Different access methods:
> 
> Method 1: SCSI Scanner- just single shot stuff, assumed a non-critical
> application, just return ENXIO to app. End of story.
> 
> Method 2: SCSI Disk- delay and retry.
> 
> 
> It's possible that these are the wrong places for such policy. It might
> should be the case that the applications or the kernel caller should decide
> what to do. However, from a pragmatic point of view neither applications or
> kernel subsystems are typically written robustly enough to figure out quite
> what to do with such information.

Well, it the example above, it wouldn't hurt to use policy #2 for both
devices :)
This is my problem: I can't see why a policy acceptable for "important" device
(disks or tape) wouldn't be good for other devices too.

--
Manuel Bouyer, LIP6, Universite Paris VI.           Manuel.Bouyer@lip6.fr
--