Dear Jonathan,

I think it would be better to gently deprecate `gregorian`, since the Bureau 
International des Poids et Mésures, who apparently have some influence in this 
area, define the Gregorian calendar to include leap seconds. A name such as 
`gregorian_fixed`, meaning "Gregorian with fixed length days" might work. 

I think you are missing the fact that Jim would like there to be a defined 
mapping from the true UTC calendar to the fixed day calendar and back. I'm 
still of the opinion that we cannot define such a mapping in CF. The situation 
is analogous, in some ways, to the use of the `no_leap` calendar in climate 
simulations of the 20th century. In such cases there is a mapping from the 
calendar with 365 days years to the real world dates. There may even be an 
ideal one, which matches the mapping used by the modelling group to transfer 
real world forcing into the artificial calendar (this matches, I think, Jim's 
use case in which observations are transferred into an artificial calendar with 
no leap seconds). However, CF, has nothing to say about how you get from one 
calendar to another because there are too many use-case specific options 
available.

Because the fixed day calendar is artificial, I think it is important not to 
refer to the time stamp as UTC --- it is a ISO8601 format time-stamp, as for 
the other artificial calendars. I've already suggested in another thread  that 
we need to clarify the wording in the convention to make it clear that while 
the units of measure in the units string have (apart from a few exceptions) the 
meaning given to them by UDUNITS, the time-stamp only has the format used by 
UDUNITS and may have a different meaning depending on the calendar (this is 
sort of obvious, but not well expressed in the convention).

-- 
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-436701377

Reply via email to