> > It seems to me that a no-such-object like "bottom" is logically most
> > consistent, where indeed f(...,bottom,...) is bottom for all f.
>
> I have tried this idea too. The main problem of a new object, I think
> is, that we have to modify the simplifer and meval to look into every
> expression for such a new object.
Well, you might introduce a new object without having all
simplifications in place and it would still be better then returning
an empty list.
Actually what's so special about lists? They are just ((mlist simp) ...).
You can very well introduce something like
((mnan simp) divergent gensym offending_expression).
I think this would unify all suggestions we had in this thread so far
and it should be consistently extendable for inf/minf/und etc.
Also it might help to implement some kind of nan-arithmetic.
Harald