> Parity doesn't reduce the likelihood of an error occuring - actually,
> one could argue that it increases it by 12.5%.  However, what it does
> do is catch a majority of memory errors.  So instead of getting bad
> data (e.g., Ohno's compile errors), you get a system crash or a killed
> process.  I personally would prefer a system crash to a bad compile,
> or, much worse, a bad financial transaction.

Exactly.  And we'll be using Parity memory in the Orion based P6
system we just ordered, though only because ECC is just too hard to
get in high densities (we need 64MB SIMMs to get 256MB of main
memory).  I never regarded the Triton's lack of parity logic a
"feature."

						Jordan