Float/doubles are finite. They only approximately model the real. And that's key. Because the fact that approximation works for a problem means it must be reas

Float/doubles are finite. They only approximately model the real. And that's key. Because the fact that approximation works for a problem means it must be reasonably smooth, which means everything will be more well behaved and rules out the worst pathologies.