Hi Larry, all,

As Denis points out, one can (in theory) go to the Web site and
dig through the source code (delta_t.cpp, in this case) and see
exactly how Delta-T is computed. As source code goes, this is
(in my humble opinion as its author!) reasonably clear and easy
to understand... _if_ you happen to be a C programmer. It's
probably rather opaque to the rest of the world, I'm afraid.

Anyway. 'delta_t.cpp' has a table of values for Delta-T from
1618 to 2004, at two-year intervals, and interpolates within
that table. After 2002, it just does a linear extrapolation.
That is, for 1 January 2002, it has Delta-T = 64.30 seconds;
for 1 January 2004, it has Delta-T = 64.57 seconds, i.e.,
Delta-T increased by .17 seconds over those two years. So,
for a given time t (expressed in years),

Delta-T = 64.57 + (t - 2004) * .17 / 2

That rate of increase corresponds to 8.5 seconds/century, and
to Delta-T = 64.13 seconds on 1 January 2000. So in the more
conventional system where t_cen = centuries from J2000,

Delta-T = 64.13 + 8.5 * t_cen

Usual warning: Delta-T becomes increasingly ill-defined as
you extrapolate it. For example, a decade or so ago, it
looked as if Delta-T was going to keep increasing by about a
second or so each year, requiring a new "leap second" each
December. Then it stopped increasing so briskly, and we've
not had a new leap second since 1999. The above formula is
there simply because Guide has to say _something_ about what
Delta-T is apt to be, even if that something isn't very
meaningful.

-- Bill