The recent post discusses an algorithm in quaternion and rotation matrices conversions.There is a risk of taking square root of a nearly negative number in floating point arithmetic.The post explores the need to handle cases where the argument for square root could be small or slightly negative.Authors suggest an alternative expression that is numerically less sensitive to avoid precision loss.