I'm doing a bit of rototilling on the inittodr() code in use on the
i386 port, in preparation for attempting to factor out a bunch of MI
code and try to use it on as many ports as possible.

I've noted that there are a bunch of discrepencies in how it is
decided whether to use the passed file system time instead of the time
from the real time clock in setting the system clock.

I'm interested in what people think the algorithm *should* be. I know
that we have several different algorithms in use, and I've seen all of
them -- I'm not so much interested in descriptions of the existing
algorithms, which I can easily read on my own (and have) as much as
what people's opinions are on "how things ought to be".

In particular, how much older should the RTC have to be than the time
from the superblock before we decide that we should use the superblock
time instead of the RTC time?

Perry