A TCP handshake is distributed by having an initiator device send, to a server SYN(m) with the IP address of a terminator device as source address. The initiator device can then forget any TCP state for the SYN(m). The server responds with a SYN-ACK(m+1, n) according to the normal TCP handshake, but the response goes to the terminator device that receives the message, reconstructs the TCP handshake as if it had sent the initial SYN message, and sends an ACK(n+1) to the server. The TCP handshake method can be used to avoid allocation of resources in for example device monitoring.