Hello,

Can you help me to demystify a particular point regarding TCP/IP:

Here is the situation:
We have written a client/server application that is running over a LAN and we are now trying to make it work over a WAN (internet).
To do that, we use the "simple" way, ( without TCP punch through or any fancy stuff ):
Port forwarding.

Let's say Client C is trying to establish a TCP connection with Server S on port 4444, and both C and S are behind two differents NAT (NAT_C and NAT_S )
At first we thought that it will be necessary to write port forwarding rules (for port 4444) for both NAT_C and NAT_S.
Because of a technical problem we were not able to setup the port forwarding rule for NAT_C. (But port forwarding for NAT_S was ok )

Despite this, we decided to run a test.
And we achieve to establish a valid TCP connection: Data were succesfully transmitted from C to S AND from S to C.
The point is that there was NO port forwarding on NAT_C .
Here are my questions:

How does the TCP/IP connection was correctly established between the two peers ?
How data can flow from S to C WITHOUT port forwarding on the NAT_C ?

From what we understood, it seams that port forwarding is only necessary on the server side. WHY ?

Thanks for your help

Okay,

I thaught that NAT devices where recording outbound connexion mapping only in the case of an UDP connection (from what i read in an article concerning UDP hole punch through).
How does the NAT device know when a particular mapping is "out of date" ? Does it inspect TCP packet to know when the connexion is closed and so, to remove the corresponding mapping in it's internal table ?