GitLab

(Originally noticed in comment:3:ticket:18628.)

Proposal 196, which defines the ExtORPort protocol, implies that the USERADDR command must include a port number, here:

     [0x0001] USERADDR: an address:port string that represents the
       client's address.

and here:

3.1.2.1. USERADDR

  An ASCII string holding the TCP/IP address of the client of the
  pluggable transport proxy.

But connection_ext_or_handle_cmd_useraddr calls tor_addr_port_split, which makes the port number optional.

It seems that connection_ext_or_handle_cmd_useraddr in fact accepts all these formats for USERADDR:

• 1.2.3.4 (implied port=0)
• 1.2.3.4:5678
• 1:2::3:4 (implied port=0)
• [1:2::3:4] (implied port=0)
• [1:2::3:4]:5678 If this is intended, then I'd like proposal 196 to say that the port is optional, and square brackets are optional in the case of IPv6.

For what it's worth, obfs4proxy and meek-server take proposal 196 at face value and always include a port in USERADDR (meek-server always uses the fictitious port number 1 because it doesn't know the true remote port).