Game is way too picky when it comes to a single packet loss

[Claviusnex]

snip

You still can't say the problems are caused by one lost packet. Just because Ping shows a lost packet doesn't mean the TCP connections that are between the client and server will. Also if the TCP connection does detect a lost packet it will go into retransmission processing. This is all handled by the OS. If these don't work the connection will be closed with an error being returned to the application. Again the only way to determine if there really is a missing packet and more importantly what was the data in that packet is to run a sniffer. It shows everything going back and forth between the client and server and is what is needed to know where to start looking in the code. Look at it this way, not providing a trace is like calling your mechanic to have him fix a rattle in your dashboard without taking the car to them so they can look at it.

I did some poking around the other day with netstat. The client opens two connections to the server that remain open while the client is running. These are TCP connections so they are reliable transport. I have not looked at a trace but it seems logical to me that these connections are set up to allow full duplexing of client/server communications. It could also explain why server based information continues to be seen at the client but client actions stop working on the server while in limbo. I also noticed an occasional connection to a Google site. Since I use Google for my DNS my wild guess is these were DNS requests. This implies to me that the client can also open other connections when needed. Any one of these could be the cause of the problem. Without a sniffer there is no way to tell which. I rarely get disconnected and on the rare occasions I have I can usually attribute them to something like a DDOS attack. I might run some traces if I get bored just to see what goes back and forth between the client and server. Hopefully any text isn't in Japanese but I doubt I will be that lucky.

I understand it is not feasible to run a trace for something that occurs rarely but when someone can come close to reproducing a condition at will, as this thread implies, that is the time to do it. It is the only way to be able to perform problem determination and be taken seriously by the tech support personnel. By suggesting the sniffer I am just trying to help the affected folks progress beyond the simplistic blow off known as "Contact your ISP". And yes I hedge on the dropped packets as having close to 40 years in the business I know the last thing a tech support person or developer wants is to be told what's wrong with nothing to back it up. It tends to put them in a box and makes it harder to explain to the customer when they determine the problem is something else altogether.

Now back to painting the house.