Old V1 Remotes

Please Hang Up and Try Again

The BIP324 spec mentions how if a V2 client tries to connect to a V1 only node, the remote node will probably close the connection immediately. Modern versions of the bitcoin core implementation validate the first few bytes they receive from a new peer connecting to them. Before the V2 protocol, they check if the first four bytes match the network magic for the network the node is running on. V2 adds another validation branch, but I am curious about how a V2 client handles really old bitcoin core remotes.

If you go back to some pretty old versions of core, talking before 0.9.0 (released around 2014), the client connecting code was much more forgiving. Instead of having a disconnect immediately policy, it tried to push the client back to the protocol. Core changed the overall strategy around 0.7.0 to be less forgiving, since forgiveness opens windows for attacks to toss anything at a node.

At the end of 2024, less than 0.1% of bitcoin nodes were running a core version older than 0.9.0. I don’t think the bip324 library needs to have the “V1 only node keeps connection alive and keeps sending V1 messages” as a first class support path. It can safely assume a remote will close the connection if it doesn’t understand V2. The worst case scenario, for implementations with a more forgiving policy, the connection fails hard as the client chokes on the handshake. The V1 version packet which starts the V1 protocol is larger than the 64 byte key to start the V2 protocol, so at least the client won’t hang.