[groestl]

> maybe this unit of time needed for scientific purposes doesn't need to be the same one used for daily life.

Don't think the unit is the problem, it's the offset. And when it comes to that, we already have a virtualization layer which assures alignment with the sun for social purposes: time zones. No need to add another one over TAI.

[ianai]

Except that’s not entirely the story of time zones. They’re political divisions. Some of them match up with the sun better than others. For instance, the sunset time gets pretty weird in LV over the winter.

[groestl]

> They’re political divisions

Which I think is perfectly fine (if not necessary) for that layer. The human body doesn't care for extra seconds, and science doesn't either. So who do we do the "coordination" in UTC for? If - over hundreds of years - the time zone drifts off into weirdness, it can be changed politically. What more do you need?

[max76]

Oh. In UT1 there number of seconds in a day is constant, so it's value is defined by the rotation of the earth which is slowing. UTC is better than UT1 in that regard. Maybe UT1 is fine for humans, but really humans just assume ut1.

[majewsky]

Time zones need to be defined relative to some reference time, and UTC is (sort of) the least impractical reference.

[groestl]

UTC adds little of value on top of TAI, except some headache for clock synchronization. So I argue for basing timezones directly on TAI, as I see TAI as the least impractical reference.