How would floating-point constants in interval expressions be handled? The standard requires the value of a constant to be independent of the rounding mode. That would appear to be a problem for a proper interval implementation (it violates containment). Bob Corbett