Dear Mr. Fateman,
>> Obviously, the mathematically correct answer would read abs(b-a) or
>> abs(a-b).
> Actually, this is not obvious.
of course you are right; I should not write emails concerned with
mathematics late in the evening... sorry!
The conceptual difficulty I had was how to tell maxima/radcan that the
argument to sqrt in fact is a squared distance and therefore, a
non-negative, real result for sqrt is required in the following
calculations in order to yield meaningful results. I now proceed by
substituting the argument by a helper variable d^2. The calculations now
require a lot of subst()-commands, but I'm not frightened anymore. The
final result turned out to be identical to that obtained using radcan.
> Richard Fateman (who first wrote radcan in 1969 or so)
I wasn't even born at that time! I think I wrote my first Fortran code
in 1994.
Regards
Dragan Stoikovitch