I\'ve got some questions regarding hole punching with UDP. Based on the wiki http://en.wikipedia.org/wiki/UDP_hole_punching
1) To set up a UDP session between two partie
1) Yes. However, you don't need hole punching if you're contacting a non-NATted server. Your client application just behaves normally.
2) Yes.
3) Some NATs do indeed restrict a public port to just one sender-receiver pair. If you need to hole-punch in such a scenario, your only chance is to guess the public port the NAT will choose for the direct connection.
However, NAT is not a security feature. Therefore, accepting any packets to the public port is not a security hole as there is no difference to the simple case of a client directly connected to the internet.