Subject: Re: Y2K support
From: Ian Lance Taylor <ian@airs.com>
Date: 23 Mar 1999 23:14:26 -0500

   Date: Tue, 23 Mar 1999 21:36:48 -0500 (EST)
   From: Keith Bostic <bostic@abyssinian.sleepycat.com>

   > This is off topic; but is anyone thinking about the C/Unix time_t second
   > counter, which overflows at some instant in 2038 (by my rough reconing)?

   Solaris 7 uses a 64-bit value for time_t.

   Which I think is a waste of time.

   2038 is ~40 years away, which is close to the dawn of computing.
   (Are there any systems built in 1960 that we're running today?)
   So, they're breaking a bunch of software today (including mine,
   to be fair), and the odds of it helping in 2039 are precisely
   zippity-doo-dah.

In Vernor Vinge's most recent book he has a throwaway line about how
the most basic time keeping system in the computer systems thousands
of years in the future is a 32 bit rollover second counter which was
started at the first moon landing--or technically, he says, slightly
later than that, recording the creation of a primitive OS.

I found that funny and all too believable.  The fact that we aren't
using any of the hundreds of programs written 40 years ago doesn't
mean that 40 years from now we won't be using any of the millions of
programs written today.

Ob FSB: of course, source code software is less vulnerable to this
class of problem, since it can at least be recompiled with larger
values.

Ian