[tommiegannert]

As much as I appreciate the topic, I fear this will be one of those 14 to 15 standards transitions. The proposal is complex in that it proposes three time units.

I see a lower-bound case for two: sun/Earth time and universal time. Farmers must be able to express "I will plant seeds at noon on August 29, 2329". They care about how the sun affects Earth. The worker trying to schedule a meeting is the same, because they'll want to guarantee the meeting happens during daylight. As such, I find the obsession over "always use Unix timestamps" a bit weird. Yeah, they're simple, but semantically wonky for some practical things. Everyone else, dealing with durations and time deltas unrelated to geological conditions want TAI.

For the farmer, the problem is the Gregorian calendar is deeply flawed/inaccurate, not that UTC has/had a leap second. Sun and Earth rotations are not aligned, so we cannot expect these to line up. We need year, day of year and time of day. The year is just the "number of solar turns" integer. The first and last day of the year would have to be cut up in unsavoury ways to account for misalignment. The time of day only needs to line up with Earth's rotation, but this means the speed relative TAI will need to shift as Earth is affected/slowed by the universe. (Now, I would suggest making the clock zenith based on something around the poles or leaning of the Earth axis, to end the Greenwich bias: that's pretty damn arbitrary.)