I don't know the UDP punching method of netmaker, but the DNS SD method is worth studying. It doesn't need to consider where the client comes from, how many layers of NAT it goes through, the type of NAT and the intermediate links, because the client can always connect with the server and only need to notify other clients of the final IP: port information. I think this is a good solution, simple and efficient. In netmaker, however, it often occurs that the server can be pinged, but the client cannot be pinged, and the punching fails. In the complex network environment, the punching scheme of transmission may be out of date.