TCP hole punching for Bitcoin nodes behind home NATs?

Option 6 proposes a method to facilitate connectivity without utilizing a board, instead depending on the propagation of address self-announcements through addrv2(a, P) as referenced in this discussion. Success with this approach hinges on both parties announcing themselves and subsequently having each other’s address made available through network relay. The reliability of this mechanism is tightly coupled to how broadly these announcements are disseminated throughout the network.

Analysis using simulation tools indicates that nodes operating from home networks, or those which are generally unreachable, experience significantly less reliable delivery of their own address announcements compared to nodes that actively listen for incoming connections. Listening nodes benefit from a greater number of peer connections, thereby increasing the likelihood of receiving relevant self-announcement information. In contrast, home-nodes are typically limited to just eight connections, restricting the potential pathways through which they can receive relayed addresses.

The network protocol further constrains this process by forwarding any given address to a maximum of two peers. Additionally, both peers request address information via the getaddr message solely over outbound connections. Given these compounding limitations, the mechanism underpinning option 6—wherein successful communication depends on mutual reception of self-announcements between two home-nodes—emerges as highly unreliable. The structural and practical barriers within the address-relay system diminish the feasibility of this method for consistently enabling peer discovery and connection establishment among nodes situated behind typical residential network configurations.