On Sat, Jan 22, 2005 at 12:52:39AM -0500, der Mouse wrote:
> "Because things are so much better now that the chance of a bug in the
> checking code causing trouble is higher than the chance of the hardware
> failing", perhaps?
> 
> I don't know whether that's actually *so*, mind you.  But it seems to
> me to be a stance that at least warrants consideration.

Having debuged a number of nasty issues involving bad RAM I'm somewhat
hesitant to to completely trust the hardware to be correct.  (Including
tag RAM on an ASIC, so ECC isn't a catch-all.)

A quick search of the net shows that 1 soft error/month of uptime....

-- 
Kentaro A. Kurahone
SIGUSR1 Research and Development