[lwip-users] TCP_MSS questions

[ThomasJ]

Hi all

It is probably a logical explanation for the behavior I see with different
TCP_MSS settings.
Hope some one can explain me why.

Our device is communicating with our server with TCP.
It is a classic server/client. Server sends a request and waits for a reply.
Protocol data send from server and clint are 14 bytes.
In some cases the server will send a big chunk of data.
In that case, the server sends the data in 128 bytes chunks.

With a TCP_MSS higher then 128 (256, 1024, 1460 etc), the client (lwip)
takes longer time to receive the complete data transfer. Sometimes it takes
the client 250 ms to receive the data. With TCP_MSS set to the 128 bytes
chunks, it never takes more then 14-16 ms!!! (For the same data)

Nagle algorithmic is disables in both ends!

I'm using the "netbuf_copy_partial" to get the data chunks with len
parameter set to 128

My own guess is, that when the TCP_MSS is set higher then 128, then either
does  "netbuf_copy_partial" wait a bit longer to see if it receive more data
before it returns? (Event I have specified 128 bytes length in the function
call).
Or is the TCP/IP delaying the netconn otification, in case more data is
coming due to the bigger TCP_MSS setting???

Thomas





--
View this message in context: 
http://lwip.100.n7.nabble.com/TCP-MSS-questions-tp21215.html
Sent from the lwip-users mailing list archive at Nabble.com.