Marcos (zeekar) wrote in snobss,

TAI vs UTC, leap seconds vs leap hours, thoughts?

This may be OT: it's actual science, but the bullshit is on the scientists' part in this instance, rather than the pseudoscientists'.

So yesterday the UTC new year was rung in a second late for the first time in seven years. And possibly the last time, since the PTB are mulling eradication of the leap second system, apparently in favor of a leap hour system, which seems like a terrible idea - sure, they'll be extremely rare, but they'll be a lot more disruptive when they do occur.

I don't understand why we don't just use TAI; screw the discrepancies with UT1. Who care if it's not true solar noon over Greenwich at 1200Z? It's not anyway what with nutation and obliquity. The difference won't add up to a half hour for 1000 years, and Atlanta is already further off than that since it longitudinally belongs in Central rather than Eastern time. Sure, astronomers do calculations over much longer timespans, but they use TD instead of UT precisely to avoid such issues.

What do y'all think?
  • Post a new comment


    Anonymous comments are disabled in this journal

    default userpic

    Your IP address will be recorded 

Oh, jeez. Leap *hour*? That'll never fly. We hope. Only problem is, TAI is already off UTC by 30+ seconds. To move *back* to TAI is just going to cause *more* problems.

I know it's busted, but it's not *that* busted, in that the algorithms and services we have now are adapted to it. I say we don't mess with it, simply because it'll be more pain to really fix it than it will be to leave it slightly broken.

"Perfect is the enemy of good enough."
    -- Linus Torvalds

p.s. good post. I would say expand your acronyms, but this *is* a science forum, and there *is* Wikipedia (and Google)...
Alternatively, we could just say that UTC where it is now is the new basis of civil time and that its offset from TAI is constant henceforth, despite the variability of UT1.

The problem is that many algorithms and services we have now are *not* adapted to it. Leap seconds are not included in UNIX time_t values, for instance, so every time we have one the clock has to be set back by a second - or, to be POSIXly correct, the corresponding time_t value has to last for two seconds without changing. I don't think either of those solutions is particularly satisfactory.

And science forum or no, I should define my terms. I can assume y'all are clueful, but not necessarily clued in to chronology minutiae. :)

TAI = International Atomic Time (French acronym). TAI seconds tick by at a constant rate determined by the SI definition of the second in terms of cesium emissions, which is what atomic clocks measure.

UT = Universal Time, which is time that theoretically represents the Earth's actual position in its daily rotation. It comes in various flavors.

UT0, UT1, and UT2 are based on measurements of the Earth's rotation and differ only in the number and type of measurements used; in each case the basic unit is the observational second of 1/86,400 of a rotation. Because the Earth's rotation is irregular, this is a variable-length unit.

UTC = Coördinated Universal Time (compromise between English and French acronyms, but usually interpreted as "Universal Time (Coördinated)" in English), by which all clocks are set for civil timekeeping. The unit of UTC is the same as that of TAI: the SI second. Because the Earth's rotation is slowing down overall, one rotation generally takes slightly more than 86,400 such seconds, and whenever the cumulative error adds up to more than half an SI second, a leap second is added to UTC, increasing the difference between it and TAI.

TD = Dynamical Time, which is tied to the relativistic reference frame of an earthbound clock at sea level, including the changes to such a frame over long periods of time. TD is therefore the timescale of choice for astronomical calculations. For times around the present TD is essentially equal to TAI plus a constant offset.
What, a leap hour? Every 25,200 years?

(a) they're barking mad.
(b) this is just saying, "Fuck it, let someone else deal with it."
Nah, the leap hour would come more often than that. On average the Earth's rotation is slowing by about a second a year, despite little discontinuities like the last seven. So we'd have our first leap hour less than 4,000 years from now. And more often as time goes on. Your point (b) still applies, of course. :)