NavList:

First, David Iwancio, thank you for explaining what you were getting at regarding an "artifact" based definition of time. Sorry for not getting back to you before now.

John Clements, you wrote:
"This was interesting. I always thought that TAI was not tied to one particular clock. ... And indeed, after reading the relevant wikipedia article, I think it's fair to say that TAI is not in fact tied to one particular clock or set of clocks; specifically, it appears to me that if you set up a good number of atomic clocks and corrected for their altitudes, you would arrive at the same second duration that the existing TAI network does. No?"

Right. The definition of the second is reproducible and not "strongly" tied to any particular implementation. But there's a related issue. In order to reproduce TAI, atomic time, as a running clock, any surrogate system would hve to start with a current value of TAI accurate to some exceedingly small fraction of a nanosecond. And any such surrograte implementation would also need to match up perfectly with the "official TAI" after centuries of elapsed time, down to some tiny fraction of a nanosecond. This probably can't be achieved in practice at the level of accuracy advertised for atomic time so the surrogate time standard would slowly random walk itself away from the official standard over the course of centuries. Even if the difference is only a nanosecond after a few decades, it would validate the idea that there is really only one time standard which must be consulted, and that set of clocks would constitute the "artifact" that David Iwancio described.

The second is well-defined and reproducible, independent of any specific device, in a metrology sense, right? But a running count of seconds could never be independent of a single standard after a long-enough period of elapsed time (how long? how many years before the uncertainty would be one millisecond?). And yes, this does contrast with the traditional time-scale based on the rotation of the Earth. If the passage of time is defined by reference to distant astronomical objects (quasars --like the ICRF), then in the event of some sort of "collapse of civilization" or global disaster, global time could be recovered to high precision by observation. We could pick up the second count with high accuracy as if disaster had never occurred. With the passage of time based on counting seconds, if the count is lost then we have start over. In a global disaster scenario, we might permanently lose a second or more... Is this a concern in any practical sense? Or is it just an amusing theoretical complaint? Do we need to design civilization's T with the ability to recover it if our civilization collapses? Maybe we should just land a few dozen solar-powered Timex watches at the Moon's south pole for future generations (if any) to visit. Oh, but then there's that "artifact" problem again! :)

Frank Reed