On 10 Jan 2012, at 0023, Rob Seaman wrote: > A day for civil timekeeping purposes is a mean solar day.
Clearly a day which consists of 86400 SI seconds isn't a mean solar day over any extended period of time. One of "86400" or "SI" or "a day runs from mean noon to mean noon" has to go at some point many thousands of years out. So we are necessarily discussing the trimming of keeping 86400 SI Seconds +/- some adjustments being equal to the current mean solar day, not a solution which will last for eternity. > > "Distant relative" is a statement of an engineering requirement. Over lo > these many years that's all I've been trying to say. "Approximately aligned" > is a statement of an engineering trade-off. We obviously disagree about the > necessary tolerances. If the tolerances are of the order of unit X, then the solution is a leap or step of unit X, whether expressed as a step in the underlying time scale (as with UTC) or in the offset between the timescale and civil time (as with timezones). I don't dispute that civil time needs to keep mean solar time to some rough approximation. I don't see why the underlying timescale needs to: we already have timezones with resolution of five minutes (Nepal), and the case that civil time needs to align with mean solar time to a precision better than five minutes is almost impossible to make (as already it's true for less than a 12th of the surface area of the planet). > >> for everyone else, it just doesn't matter. > > This is an unsupported assertion. Nobody has looked. Part of "it just doesn't matter" is that people won't even bother to look. The UK's planning to shift civil time by an hour further away from mean solar time than it already is, something which it's done --- in different ways --- several times in the past. Claims that demons lurk under the mattress to make this difficult are going to be hard to sustain, given it's been done before all over the world, and happens, mutatus mutandis, every spring anyway. The distinction between civil time and underlying time scales is lost on most people, and the experience of DST shows most people realise the relationship is arbitrary within loose bounds. No-one expects the sun to be at its zenith at noon, either on a particular day or on average, just that it be light at noon and dark at midnight. > > So you are saying that the word "day" could be set to mean some duration > distinctly different from 86,400 SI-seconds? Please consider the > implications should a civil day be taken as 86,408 SI-seconds - less than > one-hundredth of one-percent different from the synodic day. Eight leap > seconds per day or one leap hour per 450 days or timezone shifts (assuming > this is practical - I'm still waiting to be convinced) every > year-and-a-quarter. Is this practical? Well, no-one's suggesting it, and as a thought experiment it fails the "why would this happen?" test. But to answer your presumably rhetorical question, I don't think 8 leap seconds per day is practical, but I see no particular reason why a zonetime shift every year would be be an insurmountable problem. They'd all be in the same direction, so all you need is to drop one of the DST changeover dates (depending on whether your day is 8s too long ot 8s too short). Introduced at no notice there might be some fun and games, but I don't see anything that wouldn't sort itself out after a day. "Spring forward...don't fall back". > >> You either need to justify it a lot more carefully, including explaining >> why the ~2 hour offsets routinely seen today don't cause anyone any >> difficulty, > > Because they are not offsets in the underlying time scale, which remains > stationary with respect to the synodic day. But for the purposes of civil time, that doesn't matter. Civil Time is whatever the law says it is. How it's derived is a technical question, not a legal one. Does my watch, set to widely available timescales, allow me to catch trains on time? That's pretty much the beginning and end of civil time. Literally, as that's pretty much how civil time arose in the first place. ian _______________________________________________ LEAPSECS mailing list LEAPSECS@leapsecond.com http://six.pairlist.net/mailman/listinfo/leapsecs
