On Sat, Dec 08, 2007 at 04:22:40PM +0100, Robert Hoffmann wrote:
>
> 3) It waits until all ACKs for the segements sent in 2.) have been received
> (often that's 250 ms doing nothing at all!).
> 4) It sends 6-20 segments and so on until the file has been sent completely.
This reminds me of problems between delayed ack's on one side and nagle
on the other side. Is the last packet before the "pause" a segment that
does not use the full MSS?
A sample capture of two of these pauses and spikes might shed some more
light on this issue.
Cheers,
Sake