Regarding nat traversal - hole-punching

[jimmylei]

This is may not a valid issue, but I don't know where to post my questions, hope someone can help on it.

I've tried to understand how nat traversal works in libp2p, especial hole-punching. Per specs, it says libp2p-nat is the module to resolve NAT traversal, I tried to trace the code and seems only uPnP and NATPMP are implemented and did not see any implementation about hole-punching like STUN/TURN.

I also checked libp2p-circuit, but per relay document: "the circuit-relay mechanism that allows nodes that don't speak the same protocol to communicate using a third relay node", seems it not design for traversal.

So, my question is that how hole-punching works with libp2p? let's say: We already have a STUN like server(Bootstrap Node) S, now I have a node B and node C, B and C ping S regularly to keep the hole active so that S can call back B,C any time, B and C both behind NAT(Gateway devices may not uPNP device) and can't communicate directly. Ideally, I can have B to send request to S and S tell C to send a request to B, then finish the hole punching.

[zhuwei]

I have the same question

[Stebalien]

Hole-punching work is being tracked in libp2p/go-libp2p#1039. This library just does UPNP, etc.