The real-time clock chip used in older PCs only has two digits for the
year counter.  The BIOS doesn't care (except that it might only know how
to read the two-digit clock chips), since it only passes the time to
DOS/WIN.  DOS/WIN doesn't care much (except file dates.)  Whether your
applications do the right thing is up to them.  There's reasonable info
on this sort of stuff at Dallas's web site, since they've sold both good
and bad clock chips :-)

Mainframes have less of a problem GETTING the correct date, but are likely
to have databases with 2-character fields, read by programs that carefully
rangecheck and/or have actual 2-decimal-digit data types, and then do
comparisons.  I don't remember any of the little cobol I learned once, but
(for example) you have a pascal program with "var year 0..99;" (is that
close?) and then "if claimyear > lawchangeyear then" there isn't any way
that will end up correct after 2000.  You'd have been better off in "awful
C" with "int year;" - at least then all you'd have to change was your I/O.

I mostly see the silly side of this: "is your router with no senses of date
to start with certified to be year-2000 compatible"?, I talked with a
financial planner the other day, and he had a significant worry that there
would be financial crises related to the "year 2000 problem". (~30%, I
think.)  I think "Crisis" in this context means anything that will spook
the stock market, as opposed to "real" crises...

BillW