Re: Long-lived software?

["Roy T. Fielding" <fi...@liege.ICS.UCI.EDU>]

> I was just reading through Roy's new util_date stuff, and I have a
> question: what if someone decides to use Apache through the
> twenty-second century? I mean, sure, 114 years is a long time, but
> that's what they said about the year 2000 twenty years ago, and now
> some people say that the year 2000 problem is going to bankrupt half
> of all Fortune 1000 companies. (I'm exaggerating here)
> 
> My point is, the code, as stands, will be a full day off as of March
> 1, 2100, since it does not accurately take into account that this is
> not leap year. It even says this in the comments (twice). Do we really
> want to be this short-sighted?

Ummmm, no, it will fall over and die in 2038, as will every Unix system
on the planet (assuming nobody ever fixes the time_t limitations).
That isn't a problem for this particular usage (protocol dates).
Most people don't even realize that, which is why the comments are so
extensive.

I suppose we could write a set of routines that assumes that eventually
there will be unsigned long versions of the Unix time routines, but there
is nothing to gain until such an interface exists.  Besides, by the time
that happens they might make it a double.

.....Roy