From: Bill Davidsen (davidsen@prodigy.com)
Date: Wed May 17 2000 - 06:49:31 CDT
In <Fup4n0.Bo4@clw.cs.man.ac.uk>
chl@clw.cs.man.ac.uk (Charles Lindsey) warned:
> In <KKVjUYAiQRI5QAV8@turnpike.com> Paul Overell <paulo@turnpike.com> writes:
>
> >The advantage of my suggestion of using a canonical RFC822 format is
> >that it isn't a new wheel and it already covers leap seconds.
>
> No, your proposal was silent on the issue.
>
> Is, or is not
> 31 Dec 2000 23:59:60 +0000
> to be canonicalised into
> 1 Jan 2001 00:00:00 +0000 ?
>
> Because, for sure, some munger en route is going to do it and break your
> signature otherwise.
That may actually be good. Sites which reject the post on failed sig
may then be able to get the correct post from a site which didn't
rewrite it.
The issue may be more a question of allowing seconds >59 at all. since
there is such a gaping opportunity for bad implementation. Perhaps the
seconds should be 0..59, with a note that if you post during a leap
second the posting software should take certain action.
Does any application ever see a leap second, or are we ti;ting at a
windmill? My impression has been that when they are added, systems
simply go off by one on seconds until the mechanism for drifting the
clock against standard gets the system time right again. Certainly on
POSIX systems I wouldn't bet that there is a way to tell a leap second
from the others, and the "seconds since the beginning of time" is not
really a coult, but rather a handly way to express a date in a single
number.
The struct tm returns a flag for daylight time, but not for leap
seconds, so this problem may not ever some up on these systems. Am I
forgetting some part of the discussion which detailed which systems
actually support a leap second, or are we looking at something which
might happen someday, somewhere?
-- -bill davidsen (davidsen@prodigy.com) "The secret to procrastination is to put things off until the last possible moment - but no longer" -me s