Expand description
A simple rendezvous protocol implementation to help NAT traversal or hole punching.
To connect a node behind firewall or NAT(such as home gateway), which only allow outbound connection. You not only need to known its gateway ip, you also need the node send you traffic first.
This apply not only to ipv4, but also ipv6. As ipv4 need to deal with NAT, both need to deal with firewall.
How rndz works
Setup a public accessable server as rendezvous point, to observe all peers addresses and forward connection request.
Each peer need a unique identity, server will associate identity with the observed address. A listen peer will keep ping server and receive forward request. A connect peer will request server to forward connection request.
After received forward connection request from server, the listening peer will send a dumy packet to the connecting peer. This will open the firewall or nat rule for the connecting peer, otherwise all packets from the peer will be blocked.
After that, we return native socket type std::net::TcpStream and std::net::UdpSocket to caller.
The essential is, we must use the same port to communicate with rendezvous server and peers.
The implementation depends on socket option SO_REUSE_ADDR and SO_REUSE_PORT, so it is OS depends.
For tcp, the OS should allow listening socket and connecting socket bind to the same port.
For udp, the OS should correctly dispatch traffic to connected and unconnected udp all bind to the same port.
Feature flags
For convenient, the crate include both client and server code by default.
Mostly you only use client or server code, set features to client or server instead.
rndz = { version = "0.1", default-features=false, features=["client"]}client: tcp, udp clientserver: rendezvous serverasync: tcp async client, returntokio::net::TcpStream