On Wed, Mar 14, 2012 at 12:58 PM, Alan Watson <alan@alan-watson.org> wrote:
> After an illuminating discussion with Steve Allen, I have two further remarks.
>
> Steve points out that the predecessor of TAI was not called TAI in 1970. Therefore, it might be better to say that current-second returns "The number of seconds elapsed since 1970-01-01 00:00:00 as measured in the atomic time scale then maintained by the BIH. This time scale was later renamed TAI, and it has been maintained by the BIPM since 1988." (Alternatively, you might use the words from IEEE 1588-2008, but I'm not going to plunk down 180 bucks just to find out what they say.)
>
> Steve also reminds us there is some doubt as to the long-term existence of TAI. The CCTF has stated[1]: "In the case of a redefinition of UTC without leap seconds, the CCTF would consider discussing the possibility of suppressing TAI, as it would remain parallel to the continuous UTC."

We use "TAI" for lack of a better term.  What the draft
current-seconds does is return a monotonic time - an
unmolested number of seconds since an epoch.  Your
formal comment suggests changing that epoch, not
the definition of a second.

The continued existence of "number of seconds" is
not in any danger.  Converting to TAI or UTC calendar
time is a matter for WG2, and we will likely want to
provide both of these, as well as Julian time and
other historic calendars for completeness, even should
they become obsolete.

> This suggests that it might be wiser to define current-second to return the number of UTC seconds since an arbitrary epoch and provide a means to convert this to and from UTC calendar dates (either explicitly or by exposing the epoch).

The number of UTC seconds is the same as the number of TAI seconds.

--
Alex