> >> does some object to this change ?
> >
> > That would slow the reboot process down even more.   If it were
> > interruptible, I guess I wouldn't mind, but otherwise, ick.

Maybe now that I have a brand-new Pentium, two seconds would be a significant
fraction of the reboot time.  But I doubt it.

> It strikes me as the wrong approach.  In many (most?) cases
> there will not be a human staring at the console to see whatever
> flashes by, plus there is often too much to fit on a screen and/or
> memorize anyway.

In about 75% of the cases *I've* cared about, a two-second delay before
a reboot would be enough to at least understand what went wrong.  I'd
dearly love to be given a machine that can reboot and hit multiuser mode in
no more than 10 times that, so it doesn't strike me as a significant cost.

Make it depend on a kernel option.