[En-Nut-Discussion] NutTcpAccept() time-out proposal

[Harald Kipp]

Henrik,

At 11:48 28.03.2006 +1000, you wrote:

>     return NutEventWait(&sock->so_pc_tq, sock->so_read_to);

I assume that most existing TCP servers set a read timeout to detect
dead clients. I further assume that they do not expect NutTcpAccept()
to return an error. Thus, using the read timeout value to limit
accept waiting times may break a lot of existing applications.


>         if(sock->so_state == TCPS_SYN_SENT)
>                 NutEventWait(&sock->so_ac_tq, sock->so_read_to);

Not sure, but I think that the state machine will handle the timeout.
Are you sure, that this will work? Wouldn't it hurt the retry
mechanism of the stack?

Harald