On 3/11/07, Harald Geyer <Harald.Geyer at gmx.at> wrote:
(me:)
> > Thus, though 1^inf and 0^0 may often be usefully taken as
> > 1, since in some situations (perhaps rare ones) that would lead to an
> > incorrect result, we must leave them as und and ind.
>
> I don't agree that this is the only possible conclusion. Why not
> just leave ambiguous cases like 0^0 alone instead of losing even more
> information?
That would be fine by me if we can make it work. In particular, we
don't want 0^0-0^0 or f(0^0)-f(0^0) to simplify to 0.
Perhaps we need to put a flag on objects like this, e.g. ((mexpt
infinities simp) 0 0) so that detecting this case does not require a
full re-traversal of the expression tree....
-s