> > first and foremost, I think we should make sure that no code should
> > produce a mathematically incorrect result, even if it fails to
> > produce the "desired" result (i.e., it's better to not simplify an
> > expression than to simplify it incorrectly.) I believe this
> > attitude is essential if Maxima is intended to be used as a serious
> > research tool.
>
> I think you underestimate the difficulty of this.
I didn't mean to suggest that it's an easy thing to do, since I know better.
What I said is that it should be our goal. Whether or not this goal is
always attainable is another matter. Sometimes, the best you can do is to
warn the user in the documentation that certain results are ambiguous or
questionable. But that should really be a last resort, because a) users,
being the creatures that they are, don't always read (gasp!) the
documentation, and b) during a complex calculation, the user may not even
realize that an interim result (perhaps not even necessarily displayed) is
one that triggers such a special case. Verifying important results (e.g.,
with another CAS) is always a good idea, but nevertheless, we should strive
to make Maxima's results as trustable as possible.
Viktor