> > What I don't understand is why there's an insistence on keeping the > rate of UTC identical to TAI and inserting leap seconds. Why not just > define the UTC second to be the advancement of Earth's 0-longitude > line by 15 arcseconds relative to the sun (in other words, make the > Earth's rotation the reference clock for UTC), and have NTP supply UTC > and TAI separately (so that systems that need either one have access > to it)? If Earth's accumulated rotation angle is ever non-monotonic, > we'll have bigger and more thermally significant problems than > timekeeping. > The eternal question: to smear or to step over :-)
I appreciate the arrangement currently in use. Though I may be biased. Let's respect that TAI is the ultimately precise clock, derived from some atomic normal (defined that way). If you had UTC floating free from TAI, and have it follow the earth's rotation... how would you "keep" that time? What would be your reference to align the oscillators to? Have a myriad astronomical observatories around the globe, measuring the sunrise and sunset down to the nanosecond? :-) You would then need to distribute two different time domains, mutually asynchronous. Or just distribute TAI, and report after the fact, what the agreed UTC nanosecond offset was the day before yesterday? :-) Then again, this is probably the approximate way that the UTC offset is arrived at nowadays. Only averaged over many days (samples), and stepped a whole second every once in a while, based on political decision, rather than milliseconds. Well... it would make UTC more like "the wall time for the wetware who won't know the difference of a whole second, unless you hold a stopwatch in front of their face." Any application aspiring to industrial or scientific status would just run TAI on the inside. Because in UTC, a second today would be a slightly different length than two months ago. UTC intervals would not be the same length, if measured based on precise atomic clocks. The UTC offset declared to be an integer number of seconds, changing once in a few years, looks a little more practical to me. This way or that way, it doesn't make the problem go away. The fact that our astronomical time, in spite of the curious numerical system, does not align perfectly to the precise clock that we have arbitrarily defined based on some atomic wavelength :-) Frank _______________________________________________ Freedos-user mailing list Freedos-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/freedos-user