> With memory reliability as good as it is, parity is probably of dubious
> utility; it is no longer to be *expected* that memory will fail during the
> life of a computer.

Memory failure rates are under ideal conditions -- PCs
provide hardly that.  A speck of dust or humidity or
vibrations or a slightly bent connector or a host of such
conditions can cause intermittent failures.  Parity atleast
improves your chances of catching such an error.

Also, if you catch a parity error, there may be other ways
of recovering (other than ECC).  A parity error in a page
read from a file (and not yet written to) can be handled by
reloading from the file.  A parity error in a cache can be
handled by reloading from memory (provided the cache is
write-through).  A parity error in generated data can be
handled by regenerating the data. [Why these things are not
being done or how hard they are is a separate discussion].

> "Parity is for farmers."
>        - Seymour Cray, on why the Cray 1 supercomputer did *not* have parity.

I have heard one of the biggest investors in Micron Tech.
is a farmer :-)