Unexpected behavior about kcp-go

[xtaci]

    for {
        c, err := l.Accept()
        if err != nil {
            panic(err)
        }
        c.Write([]byte("hi"))
        n, err := c.Read(b)
        fmt.Println("LISTEN", string(b[:n]), err)
    time.Sleep(time.Second)   /// this will let listener send data to client
        c.Close()
    }

your output tells that client tries to retransmit hello message, but the listener never had a chance to send data, because it's closed immediately after Read().

KCP doesn't define session control messages like SYN/FIN/RST in TCP, a real world example is to use some multiplexing protocol over session, such as yamux.

[AudriusButkevicius]

I am more worried that the dial part keeps trying to send data even after Close(). I fully understand that there is no session control, but TIME_WAIT could be implemented based on the convID on the listen side, essentially ignoring all packets for some time.

Regarding the timeout, not sure why you've pointed at this issue as it's somewhat a different question.
I saw that there was a local var related to connection timeout, but it's not used anywherr.

Are you saying application level timeouts is the right and only way to solve it? Or are you saying yamux solves it?

[xtaci]

application level timeouts is the right and only way to solve it.

[AudriusButkevicius]

I see yamux has some tcp like control messages which might help solve this, I'll see where that takes me.