"all floats can be represented perfectly in decimal"? not sure i understand