[whatwg] [wf2] Leap seconds, dates in the past

[Ian Hickson]

On Mon, 10 Apr 2006, Henri Sivonen wrote:
>
> Earlier I raised the issue that WF 2.0 does not say what should be done 
> about leap seconds, but I did not suggest a solution. Having read more 
> about the subject, I suggest that date calculations in WF 2.0 be POSIXly 
> correct and ignore leap seconds. (See IEEE 1003.1, 2004 Edition, section 
> 4.14 Seconds Since the Epoch or the corresponding section in ISO/IEC 
> 9945-1.)

Agreed. That's what I did in response to your last mail.


> Another problem I see is that ISO 8601 leaves wiggle room with dates that
> predate the use of the Gregorian calendar. I think there are three politically
> correct ways of approaching this:
>  1) Mandating the use of the proleptic Gregorian calendar all the way to
> 0001-01-01.
>  2) Prohibiting dates before 1924-01-01. (By that time, all countries that
> previously used the Julian calendar had upgraded to the Gregorian calendar for
> non-religious purposes, according to Wikipedia.)
>  3) Require the change from Julian to Gregorian between 1582-10-4 and
> 1582-10-15. (Dates from Wikipedia--caveat lector.) This is the ICU default
> behavior, BTW.
> 
> The politically incorrect way that comes to mind is what the Unix cal utility
> does: Following the British Julian to Gregorian upgrade. (Run the command 'cal
> 9 1752' on a *nix system to see what I mean.)
> 
> I'm inclined to think that the best option for WF 2.0 is to require the 
> use of the proleptic Gregorian calendar all the way to 0001-01-01.

I have no idea what the word "proleptic" means but it sounds good. I've 
added this to the spec on the assumption that you know what you're talking 
about (which has historically been shown to be a good assumption).

-- 
Ian Hickson               U+1047E                )\._.,--....,'``.    fL
http://ln.hixie.ch/       U+263A                /,   _.. \   _\  ;`._ ,.
Things that are impossible just take longer.   `._.-(,_..'--(,_..'`-.;.'