Dear Jim

Please could you elaborate on your reply to me? I appreciate, as Chris says, 
that this thread is long and complicated, like our previous email discussion 
was, so I may have missed something. However we do need to keep all this serial 
reasoning, I think. From time to time, the moderator (or the proposer, if there 
is no moderator, as is often the case) should restate where we have got to, in 
order to see if everyone agrees. Maybe I'm implicitly asking you to do that now.

I use the names tai and utc, as you first said, because they're more mnemonic 
than a and b. The names we eventually choose for the convention are a different 
matter. They don't have to be tai and utc.

If I understood Martin correctly, I agree with him that timestamps should be 
interpreted in the calendar stated. That's what we do in model calendars, and 
with gregorian and julian. Apart from that point, I think I agree with what you 
proposed, don't I?

gregorian_tai. No leap seconds in the conversion between time coordinates and 
TAI timestamps, reference time is a TAI timestamp, time coordinates are metric.

gregorian_utc. Leap seconds in the conversion between time coordinates and UTC 
timestamps, reference time is a UTC timestamp, time coordinates are metric.

gregorian (the existing calendar). For the real world, reference time is a UTC 
timestamp, but no leap seconds in the conversion between time coordinates and 
timestamps, and time coordinates are therefore not metric. Data written up to 
now with this calendar might or might not have allowed for leap seconds in the 
conversion (if it did, it was gregorian_utc). We don't know, but as you have 
argued several times, most applications don't require such precision, so it 
doesn't matter very much. (For the ones that do, we are providing the two new 
calendars.) For model data, reference time is UTC in the model world, 
conversion has no leap seconds, but the model world has no leap seconds, so the 
time coordinates are metric.

How does that differ from what you have in mind?

Cheers

Jonathan


-- 
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-435587942

Reply via email to