Randall Holmes (holmes@catseye.idbsu.edu)
Thu, 27 Oct 1994 09:26:58 -0600

I agree absolutely that floating point numbers, whatever they are,
are not representations of reals :-) A genuine ADT for real numbers
would be an interesting (feasible but hard) project.

--Randall Holmes

P.S. This becomes a serious problem when one attempts, as our
department is attempting with me and my students, among others, as
guinea pigs, to teach elementary calculus using Maple or other
symbolic calculation software. Both bugs in the software and effects
of round-off error can produce phenomena which are very confusing to
students, and have done so in practice, and the students have no
background which would help them to figure out why they are getting
into trouble. I can see how sound and debugged symbolic computation
software targetted at college students could be very valuable in
teaching, but Maple (like any other package I am aware of) has none of
these properties. Keep it away from the QED database :-)