Intelligent result checking before asking positive/negative/zero...?

> In general wouldn't this be a bad idea?  ...
> wouldn't there be a danger of Maxima "churning" for a while and not
> providing the result it can provide?

I was not thinking of this as the default behavior.  It could be
chosen either by some sort of operator, e.g.

       eval_by_cases( integrate(x^a,x,0,1))
or
       eval_by_cases( integrate(x^a,x,0,1), [a] )
                      -- only do all cases for variable a

or by a global variable, e.g.

      integrate(x^a,x,0,1), prederror: all;

> "well, this looks like it is solvable for all cases of x, so I'll try all of them?"

No.

> (I don't even know - is Maxima able to make
> decisions to ask questions based on real vs. complex assumptions?)

No.  Maxima doesn't know much about real vs. complex variables.

          -s