How about including the above in a future version as internals? I have
some fairly good .M files for integer and half-integer orders
(J and Y, and spherical j and y), but real/complex orders are more
difficult, and should really be implemented in FORTRAN.
BTW, I suppose people are aware that the Matlab FTP server has a large
collection of .M files which are not copywrited and can be very
easily used in Octave.
I think SLATEC library has Bessel functions also. (Some of the functions from this
library are included in Octave). So it should be be difficult to include Bessel functions
into compiled code. I too required Bessel functions, but I used the code from Matlab
Bessel functions from our local Matlab installation. But that is copyrighed so can't
be included with Octave. So compiled Bessel functions from SLATEC might be
a better choice besides being faster :-)