> There was a proposal to add the third behaviour, (c):
> don't draw our function in such places in R where some intermediate
> results during the function calculation are not real, even if the final
> result is.
This proposed behavior really has nothing to do with plotting. It
would be some sort of "real-only" mode, where for example sqrt(-1),
asin(2), etc. would give an error. Though this could be done in
principle, it would require changes to large parts of Maxima. For
example, (sqrt(x))^2 and sin(asin(x)) could not simplify to x.
-s