[Guy Harris <gharris@xxxxxxxxx>]

Matt Smith wrote:

> The client (*.20) is telling us that at the application level, he is 
> getting a timeout, never recieving the data.  We don't have control over 
> the client application, but the server application we do (*.12).  We are 
> trying to determine where the problem resides. If the clients is infact 
> recieving the data, and maybee a problem at the application level.

It *might* be a problem at that level - but I suspect not.  I suspect 
the client machine's TCP might not be handling segments with FIN and data.

I suggest you ask the supplier of the operating system on the client 
whether its TCP stack can handle that.  If it can't, and they have no 
fix or other workaround they can supply, ask the supplier of the 
operating system on the server whether there's a way to get it not to 
send data with a FIN.

> Out 
> of 10,000 transactions a day, only about 5 have this problem.  This is a 
> problem because the server thinks the transaction was sucessful,

The server's TCP stack is getting an RST from the client, so, at least 
at the TCP layer, it knows there's something wrong.  Is the server 
application checking whether the send it's doing *AND* the close of the 
connection (or a shutdown of the send side) are getting errors?  If not, 
you should add those checks - the RST might be getting reported as 
errors to networking calls.

(What OSes are the client and server running?)