Re: [lwip-users] LWIP Timeout(?) 2 second to ACKnowledge the end of the transmission. log file TC_IP trace2.pcap

[address@hidden]

 Kieran Mansley wrote:
> On Tue, 2010-10-05 at 09:48 -0400, Damon, Antoine wrote:
> > Have you got an idea about this timeout(?).
> Sounds like the server is just waiting before closing the socket,
> perhaps because it's giving the client a chance to send another request
> but doesn't want to leave sockets open indefinitely.
Well, that's a bit odd. An HTTP/1.1 server SHOULD use the 
"Content-Length" field to indicate the length of the body (RFC 2616, 
section 14.13). If this field is not given, the only means to tell the 
client that the body is finished is closing the connection (section 
4.4). This is especially used when the webserver does not know the 
content length when sending the header (see below).

Also, if "Content-Length" is missing, the connection cannot be reused 
since the client does not know when the server has finished the previous 
body.

Looks like a bug in the Go-Ahead webserver (well, it's not violating the 
RFC, it's a bit slow only). But it could well be produced by the 
application running the webserver as the data returned seems to be 
generated rather than compiled-in. In this case, the server might not 
know the body length in advance and gets informed too late that the end 
is rechached.

Maybe you might want to check your application code in that direction...

Simon