I agree, 121 seconds is the correct answer in the real world. The difference 
between the time coordinates is 120 seconds and this is wrong. That is what Jim 
means by it not being metric[al]. However, the timestamps can be encoded to 
coordinates, which can be decoded reversibly and correctly, except that 
`2016-12-31 23:59:60` cannot be encoded. Software supporting this calendar 
should object to that timestamp. We all know this is a kludge, but it's widely 
used and easy to use, so we should (continue to) support it. CF is mainly 
concerned with providing ways to for people to do what they want to do as 
clearly as possible.

In the model world, the difference is truly 120 seconds.

The CF `calendar` attribute specifies the calendar for the reference timestamp 
and other timestamps, and the rules for converting between timestamps and time 
coordinates. This seems convenient and clear to me. We just need to add the new 
ones for (proper) UTC, TAI and GPS, I think.

My daily contribution to this discussion. :-)


-- 
You are receiving this because you commented.
Reply to this email directly or view it on GitHub:
https://github.com/cf-convention/cf-conventions/issues/148#issuecomment-437106123

Reply via email to