[AFS3-std] Re: A call for consensus on draft-deason-afs3-type-time-02

[Russ Allbery]

Andrew Deason <adeason@sinenomine.net> writes:
> Russ Allbery <rra@stanford.edu> wrote:

>> Or, hm, I suppose if you squint at it right, you can decide that
>> "number of seconds" isn't just elapsed actual time, but includes the
>> leap seconds that were inserted.  Which would also work for our
>> phrasing.  Maybe we could just say that explicitly.  Something like:
>> 
>>     the number of seconds and nanoseconds since midnight or 0 hour
>>     January 1, 1970 Coordinated Universal Time (UTC), including any
>>     leap seconds inserted into UTC.

> It's very possible I am backwards on this, but shouldn't this be
> "excluding any leap seconds"?

Oh, yes, you're exactly right.  Thank you.

> That is, in our time representation, there is a difference of 1 "second"
> (however we define "second") between the times 31 Dec 2005 23:59:59 and
> 01 Jan 2006 00:00:00, even though there was a leap second at 31 Dec 2005
> 23:59:60. That is, I thought we'd be following UTC more than TAI.

Yes, correct.

> And also, I did find another instance of this being mentioned in IETF
> RFCs. RFC 4049 states in Section 2 (at the top of the second page):

>     The integer value is the number of seconds, excluding leap seconds,
>     after midnight UTC, January 1, 1970.  

> Would that work for us?

That sounds great to me.

We could take the same approach if we use the other epoch, although I
think we should be more wordy about it:

    The integer value is the number of seconds, excluding leap seconds,
    after midnight, January 1, 1601 Coordinated Universal Time (UTC), in
    the Gregorian calendar.

        NOTE: Neither the Gregorian calendar or the modern UTC time zone
        were in use in most of the world at that date, but times are
        represented as if they were, using the obvious backwards
        projection of the current UTC time zone and Gregorian calendar
        rules to January 1, 1601.  Be aware that any real-world times from
        that era will likely require Julian to Gregorian calendar
        conversions to be represented in this format and probably cannot
        be simply converted using normal time conversions from the modern
        era.

I think that phrasing would resolve my objections to the epoch, along with
the additional bits that are in the current draft about how to convert.

-- 
Russ Allbery (rra@stanford.edu)             <http://www.eyrie.org/~eagle/>