BL> For memory CRC-32 gives odds of a million... you'd need a file a
BL> million long to get an even-money chance of duplicating the CRC with
BL> two files one-bit different. 

Nope.  There is ZERO chance of a 1-bit error getting through undetected. I
*think* 3-bit errors are also zero chance.  I think 2-bit errors are 1 in a
billion or two.  BFN.  Paul.
@EOT:

---