BL> In a general sense, I never did see any point in parity. If a
BL> system is so unreliable it needs 12.5% redundancy then 12.5% is
BL> not enough! The minimum workable redundancy is 100% - or you
BL> may as well not bother.

PE> It isn't redundancy, it's designed to see if an error has
PE> occurred, ie the memory has been corrupted.

BL> How does that work? If the other 8 chips handle one bit each, then
BL> how does the one-bit in the parity chip keep track on 8? I assumed
BL> that it had to work on averages... a false parity meant it was likely
BL> one of the others had failed too.

Well, the idea is first of all that it's highly unlikely for memory to go
bad ever.  So on the one-in-a-billionbillion chance that a cosmic ray has
zapped one BIT of your memory, you only need to be able to detect a single
corruption.

So if you have 8 data bits and 1 parity bit, then what you do is, say your data is:

1 0 1 1 1 0 0 1

Well you then have a 9th bit (the parity bit) which records whether the
data has an even number of "1's".  In this case, there are 5 1's
and 3 0's, which is odd, so you put in the parity bit a "0"
meaning "not even".

Now when the cosmic ray zaps your data, and you have:

0 0 1 1 1 0 0 1

then when you go to fetch the memory, you can tell that the parity bit says
it is not even, but you have an even number of 1's. Therefore a cosmic ray
has zapped your memory.

There are many variations on this theme, and the above is just one that I
made up.  In actual fact, a similar scheme exists for hard disks, can't
remember the name of them, but you can have say 5 disks, but you only put
your data on 4 disks, and you keep the 5th disk as an XOR'd sum (or you
could use parity etc) of the other 4 disks.  When one disk COMPLETELY DIES,
you can tell what data used to be stored on it, ie whether it was a 0 or a
1 (whichever one makes your 5th disk's maths work out correctly). And
they're so clever that you can actually take any one of the 5 disks out,
throw it in the garbage bin, and get another disk, and stick it back in,
and after a short while you're back to being fully protected against a
single hard disk failure!

Very nice!  BFN.  Paul.
@EOT:

---