If Double is IEEE 754 compatible then it should be perfectly defined for 0, you'd get +/- infinity. There are algorithms which rely on this.