-----maxima-bounces at math.utexas.edu wrote: -----
>It could be that defining Bessel_J of fractional order by taylor series is
>not a good idea.
Downward recussion + normalization (the Miller algorithm) works pretty
well. This
method is described in A&S (available online). Usually, the method is
described for only integer
order Bessel functions, but if I recall there is an normalization that
works for fraction
order too. I don't know how Maxima does numerical evaluation for integer
order.
Barton