Re: socket in use

Very helpful,  Scott. Thank you.


Is using the SO_REUSEADDR option bad practice?  Should a server pgm wait and
retry when its port is reported as in use?  Esp if the circumstances allow?

is there an instance nbr assigned to each socket to socket conversation and
is it transmitted along with the ip addr and port nbr in each packet of data
sent between sockets ?

Would this "conv instance nbr" enable the ack from the end of a closed conv
not to be mixed up with packets from a new conv?


Thanks,
Steve Richter



>Scott said:
>
>The TCP protocol is a 'reliable' protocol.  IT ensures that all packets
>are received on both sides of the connection.
>
>When a connection is closed, each side of the connection sends a 'FIN'
>(finished) packet to notify it's peer that the connection is to be closed.
>Being a reliable protocol, it has to ensure that this 'FIN' packet was
>received, so the side that receives the 'FIN' packet will send back a
>'ACK' (acknowledgement) packet.
>
>Okay, that's well and good -- but how does it know that the 'ACK' packet
>got received?
>
>The answer is...   it doesn't.   If the ACK is never received, the side
>that is expecting it will re-send it's packet after a timeout period.
>But if the ACK is received, it gets no response at all.
>
>Therefore, there is a period of time after that final ACK where the socket
>is waiting for a potential 're-send' of the FIN packet, just in case the
>ACK got lost.
>
>Forcing the socket to close before this timeout value has elapsed would
>violate the TCP protocol, and is generally not a good idea :)
>
>The 'timeout' value that determines when a packet is to be re-sent is
>called the MSL (Maximum Segment Lifetime) and it defines the length of
>time that a segment (a packet containing a piece of a TCP stream) can
>bounce around the network before being considered 'lost'.
>
>When waiting for a 're-send' because that final ack was lost, RFC 793
>states that you should wait for (2 x MSL) before giving up.   This way,
>you've taken into account both the MSL time, as well as the extra time
>that a slow network connection might add to it.
>
>So back to the question....  why can't you bind() to a port within a
>period of time after that port has been closed?
>
>The system doesn't allow two different programs/processes/etc to bind use
>the same port at the same time unless the 'SO_REUSEADDR' socket option
>has been set.
>
>Since that port is still in use (due to the wait for the 2 x MSL timeout)
>it does not let you re-bind to it until that timeout period has elapsed.
>
>I hope that makes sense...   It's a little difficult to put into words...
>
>
>
>+---
>| This is the Midrange System Mailing List!
>| To submit a new message, send your mail to MIDRANGE-L@midrange.com.
>| To subscribe to this list send email to MIDRANGE-L-SUB@midrange.com.
>| To unsubscribe from this list send email to
MIDRANGE-L-UNSUB@midrange.com.
>| Questions should be directed to the list owner/operator:
david@midrange.com
>+---
>

+---
| This is the Midrange System Mailing List!
| To submit a new message, send your mail to MIDRANGE-L@midrange.com.
| To subscribe to this list send email to MIDRANGE-L-SUB@midrange.com.
| To unsubscribe from this list send email to MIDRANGE-L-UNSUB@midrange.com.
| Questions should be directed to the list owner/operator: david@midrange.com
+---