Experimental Physics and
Industrial Control System

Ø  ASYN already provides a timeout mechanism with the option of tearing down the TCP connection on timeouts.

 

But that requires an additional delay in detecting that the socket has indeed disconnected by waiting for a write or read operation to time out.  And what if the remote end is sending unsolicited messages at some unknown interval?  It would then be up to the remote end to determine that a write failed and tear down the socket and reopen it? Could there be state information lost when a socket is closed and then re-opened?  It seems cleaner to let the underlying OS implementation optionally take care of sending keepalive packets to prevent the types of problems Pedro is seeing?

 

Mark

 

 

 

What’s use case for SO_KEEPALIVE?  ASYN already provides a timeout mechanism with the option of tearing down the TCP connection on timeouts.