ACCURACY INFO:
 
The distinctions between UTC, UT1, TDT (etc.) are ignored
except that a rough correction to TDT is used for the moon.
The solar ephemeris used is good to << 1 arcmin.  Moonrise/set
are based on low-precision algorithms, but explicitly printed
moon positions are +-1 arcmin.  Moonrise/set is generally
+-2min, sunrise +-1 min; non-level horizon, site elevation,
and refraction uncertainties are often larger than this.
 
The planetary calculations are truncated, but the error should
seldom exceed 0.1 degree; MV are best, MJSU ok, Pluto worst.
 
The Local Mean Sid. Time accuracy is set by the time argument 
which is the julian date in double-precision.  Workstation-class
hardware generally gives it to < 0.1 second.
A JD very near midnight can sometimes be converted into one calendar
date and *another* day of the week; this only happens within a
roundoff error of midnight and is flagged on output.
 
Daylight savings time, if selected, is established using a
site-specific convention (e.g., USA post 1986).  Beware of ambiguities
and nonexistent times when the clock is reset.  In this case,
use the 'g' option and enter times and dates as Greenwich (UT),
or disable DST in site params.
 
The *precession* routine used is a 'rigorous' rotation matrix routine,
taken from L. Taff's Computational Spherical Astronomy.
It uses IAU1976 constants, is good to < 1 arcsec in 50 years,
and has no troubles near the pole.  Proper motion corrections
are done crudely as x = x0 + mu * dt; this is inaccurate near
the poles.  Use another routine if sub-arcsec accuracy is critical.
Apparent place (with nutation, aberration) is NOT given.

The parallactic angle follows Filippenko (1982, PASP 94, 715).
 
The barycentric ('heliocentric') corrections are computed using
an elliptical earth orbit with a few periodic perturbations
including lunar recoil.  The helio-to-barycentric transformation
uses the same algorithms as the planetary postions.  Overall max error:
< 0.2 sec and < 0.005 km/s.  Velocity corrn. includes earth rotation.
 
The galactic coordinate routine is rigorously accurate, and
precesses to 1950 before transforming.  The ecliptic coord.
routine is for coordinates of date and is good to < 1 arcsec..
 
These routines are **not correct** at times very far from
the present (1990s).  The program rejects input outside 1900-2100.
 
When porting to a new machine, run tests to ensure
correctness and accuracy.  Experience shows that compiler
peculiarities arise surprisingly often.
 
Use with reasonable caution, at your own risk.