From: Henry Spencer (henry@spsystems.net)
Date: Mon May 08 2000 - 09:27:00 CDT
On Mon, 8 May 2000, Paul Overell wrote:
> Is using POSIX time a good choice for canonicalization?
Reasonably so. Imperfect in some areas, but not bad.
> As I recall it runs out in 2038...
No, that only happens if you store it in signed 32-bit binary. Note
that Charles's spec calls for it to be expressed as a decimal number.
Decimal numbers do not run out (although the techniques for generating
them get harder as they get bigger).
By the way, the POSIX spec doesn't say that time_t is 32 bits and doesn't
require it to be signed.
> ...and is ambiguous in that it uses the
> same integer value for both a leap second and the following second.
This is rather unfortunate, although it reflects existing practice. It's
unlikely to be a problem in real life.
> I suggest instead using a canonical RFC822 format...
Hmm. Seems workable, given careful specification. Might be preferable.
Henry Spencer
henry@spsystems.net