Computer bugs in the year 2000

[Doug Pardee]

> >>       I have a friend that raised an interesting question that I
> >> immediately tried to prove wrong.  He is a programmer and has this
> >> notion that when we reach the year 2000, computers will not accept
> >> the new date.  Will the computers assume that it is 1900, or will
> >> it even cause a problem?...
> 
> Your friend is probably aluding to the leap-century correction
> in the Gregorian Calendar.

Oh, dear oh dear.  Folks, there is an outside world out there and
that world uses computers to do REAL STUFF.  One of the "real stuff"
things that computers do out there is to store data in files, both
on tape and on disk.  Things like the balance in your checking account
(or the amount that it's overdrawn :-)

There is SO MUCH data in those files, and tapes and disks cost SO MUCH
to buy and store, that those files have "expiration dates", at which
time a program (run daily, as a rule) will see that they have expired
and will remove all traces of them from the various directories, and
will return the disk space or reel of tape to the "available" pool.

I imagine you are aware that IBM's System/360/370/30xx machines
handle nearly all such transactions (to the unending dismay of
Honeywell, Burroughs, Univac, etc.)   In the IBM world, the date
of December 31, 1999 is the highest (latest) date that can be
specified.  So if you have stuff that you want to keep forever,
you put a date of 99365 on it.  I leave it to your imagination
what will happen on 12/31/99 when all of those computers find
all of those disk files and tapes are to be scratched.

A variation results from the natural cycle of many such files.  For
example, a monthly backup tape in a 4-month cycle will be kept for
four months, no?  Although IBM doesn't supply any routine to
compute such a date, virtually every site has written or bought one.
So on, say, 10/01/99 a 4-month file will be set to expire on
02/01/00.  Guess what happens the next morning?  Bye-bye file!

There are a number of other effects which will result, all from
the fact that the computer will NOT be able to compare two dates
to find out which one is later.  Unless the programmer anticipated
the problem, the formula for figuring out how many days elapsed
between two dates won't work.  How do you figure, e.g., interest
earned, if you don't know the time period involved?

Dates and time ARE of the utmost importance to the business world!

There are minor effects, too.  Like when your company's ten-year
forecast says that you'll be making a good profit in 1903.  Looks
really professional on the ol' annual report.
-- 
Doug Pardee -- Terak Corp. -- !{hao,ihnp4,decvax}!noao!terak!doug