[E-Lang] Relying on H/W
Mon, 29 Jan 2001 12:02:07 -0700
> I think the place to be most concerned is where some security value is
> changed by the error. In the EROS case, if a DomainKey was changed to a
> NodeKey by a memory error, then you would have a "magic" wand to wave over
> that domain, which will persist as long at the domain persists.
> Does anyone have any information about how often these kinds of errors
> actually occur? (Or are all my Windows re-boots software caused?)
Anecdotal evidence: I had a Windows machine I had manually upgraded to a
faster CPU, and it started crashing all the time. Figuring I'd created some
sort of edge-condition clocking-mismatch or some such, I turned the computer
into an experimental Linux machine that I could putter with, but would not
endanger my data.
It hasn't crashed since.
It's the software :-)