I researched this some more, and now wish to withdraw what I proposed earlier.

The correct way to handle this is to do what golux (++) said.

Here is the sequence of events on failure:

* You attempt to SEND to the client
* Client has closed the socket
* The error is indicated to your program via a $SIG{PIPE}
* You need to trap that signal as golux suggested, and set the value of the global $b_err
* You check the value of $b_err before the next activity on the socket.
* IF triggered, close the client socket, and wait for the client to reconnect (Go back to "accept").