In message: <[EMAIL PROTECTED]> Neal McBurnett <[EMAIL PROTECTED]> writes: : I still haven't seen any good data on predictions for periods of : longer than 9 years.
Neal, thanks for the excellent summary of the current state of the art in prediction. I think this shows that a 20 year time line is unrealistic at this point, but 5-10 years would keep things fairly close, and 4 years should be able to keep the current tolerances. It might be worth an experiment where over the next 5 years IERS publish 12 new months of data every 6 months. (eg Jan 2006 publish both the June 2006 and Dec 2006 correct, July 2006 publish the June 2007 and Dec 2007 correction, Jan 2007 publish Jun 2008 and Dec 2008). We'd hit 4 years in advance in Jan 2009. This would phase in the predictive timeline for leap second insertions, and would also give the IERS control to end the experiment if the time horizons exceeded their ability to predict with confidence. This would be an evolutionary step, rather than a revolutionary one. Of course this would make them even more entrenched than they already are, because to kill them would require waiting many years... Warner