I don't think CL supports setting the floating point rounding mode (maybe
CMUCL does as an extension); at least for now, I think it would be okay
to have error bounds that were almost certainly correct. Really it seems
that the hard part of doing interval analysis is blending an interval type
into Maxima. If you still have a copy of your code and if you could donate
it, I'll look at it. The specfun name was inherited from Macsyma.
Barton