powered by Jive Software

Network Error Behavior

Hi, I was wondering if there’s any potential development being done on detecting and acting on a network error between the gateway and any of the legacy IM networks. For example. If you have a connection to AIM and you forcibly block the port that you use to connect to aim (normally 5190), the tcp connection persists, and persists, and persists for a very long time. When you unblock the port, all messages that were queued are sent/received. Wouldn’t be better to let the user know they have been disconnected after a couple minutes of being unable to keep the established connection essentially alive?


In theory, this ought to be being handled in the form of keepalive packets that will filter an error up the chain. Do you see anything in your debug logs when the connection is blocked? BTW, are packets dropped in the form of just dropping the packets, or are they rejected and an error sent back?

What I’ve noticed using 1.1.4 is that there is no information in the debug logs (or my tcpdump I believe) saying that the packets have been rejected. They are simply queued on the tcp connection until such time that the connection is working again. I believe the messages all get sent through the newly unblocked port when there is any attempt to send or receive a message on the connection. It’s like the connection is spooling the messages until the resource (the port) is available. When there is a new attempt to use the port after unblocking, all messages go through.

Again, the packets are not dropped or rejected, they are just queued up.