Hello Libp2p Hackers,
I wanted to share a P2P networking framework I’ve been building. Like Libp2p, it addresses connectivity, but my approach diverges in key ways.
When I last examined Libp2p, I noticed a few areas worth highlighting:
NAT Type Detection
Libp2p’s AutoNAT is often compared to STUN but primarily determines if a node is behind a NAT. It doesn’t implement RFC 3489 to identify specific NAT types (e.g., full cone, restricted, port restricted, or symmetric), which is crucial for fine-tuning connectivity strategies.
Hole Punching
Libp2p’s hole-punching mechanisms work well with simpler NAT types (e.g., full cone, restricted), but can struggle with more restrictive NATs. Symmetric NATs, in particular, generate unique mappings for each connection, complicating conventional punching techniques. Emerging research offers ways to improve connectivity in these cases, but they require deeper NAT behavior analysis.
With these observations in mind, I designed my library to focus on NAT enumeration. By identifying NAT types and mapping behaviors, I developed improved algorithms optimized for different network scenarios. For instance, my framework supports multiple interfaces and features strategies tailored for enhanced direct connectivity, even across LANs or on the same machine.
The library is written in Python (≥3.6) and works across most platforms. I’ll attach a demo below for those interested.
Lastly, I respect Libp2p and its contributions to P2P networking. My project is not an implementation of Libp2p but rather a complementary effort focused on underexplored areas like NAT enumeration.
PS: I used AI to shorten my original post.