Confusion around `J1900_REF_EPOCH` and `J2000_REF_EPOCH`

[AljoschaMeyer]

Hi,

I am not at all well-versed in the world of timekeeping, but had to delve into using TAI recently, and I have a hard time understanding some of the documented constants in hifitime.

The docs for J1900_REF_EPOCH state that this is "The J1900 reference epoch (1900-01-01 at noon) TAI." This contradicts the NAIF SPICE J1900, which is exactly 24 hours earlier. The JD_J1900 constant agrees with NAIF SPICE, and thus contradicts your J1900_REF_EPOCH. Is this an actual inconsistency, or am I missing something? Does this mean that all times in hifitime are off by one day? I'm assuming this is not the case, but I do not really understand why.

The docs for J2000_REF_EPOCH state that the time is midnight, but shouldn't it be noon? Since it is defined as being exactly one century after J1900_REF_EPOCH, it cannot shift by twelve hours, right?

CC @joakim

[ChristopherRabotin]

Hi there,

Rest assured, the time computations are validated and match SPICE.

However, you've likely found the exact reason why I've had an issue implementing #294 ! I tried to implement it before version 4.0.0 and the validation tests kept on failing. I now suspect that's because the reference epochs are wrong. They aren't actually used in the code:

[joakim]

Good to hear! Hopefully that cleared things up for all of us :)

With the correct value (nanoseconds: -43200000000000?), J1900 should be a sunday.

The only issue I can see with J2000_REF_EPOCH is that the documentation says midnight when it should be noon.