One thing that is clear from inspecting code that actually
uses FLOOR, CEILING, etc, is that there is a very strong
usage pattern; the procedure inexact->exact is almost ALWAYS
called on the result.

I think standard library macros ought to capture and standardize
this usage.

(xfloor x) <==> (inexact->exact (floor x))
(xceiling x) <==> (inexact->exact (ceiling x)
(xround x) <==> (inexact->exact (round x))

			Bear