how to aggregate multiple similar tcp flows?

[Christoph Badura]

Hey Carter,

On Mon, Mar 12, 2007 at 10:19:37PM -0400, Carter Bullard wrote:
> Basically a  problem with retransmissions
> that occur way out of the window.  Needed to account for winshift,  
> and the possibility that
> this was finally the segment we were waiting for (very large window).

what do you mean "way out of the window"? I'm certain that the sender
never sent outside the window.  And the receiver did advertise a window
shift of 0.

But it is fixed now. Many thanks.

> ESP issue was really a fragmentation issue, and that is now fixed.

I'm afraid that isn't fixed yet.  I still get output like:

$ ra -n -s +sloss +dloss -r trace9f.argus|head
   16:49:21.850751       F     esp      1.2.3.4          <->      4.3.2.1.0x0bf*      377       64        57990        51184   CON      67882          0
   16:49:21.853440             esp      4.3.2.1           ->      1.2.3.4.0x17e*      368        0        56400            0   INT          0          0
   16:49:22.858836             udp      1.2.3.4.500      <->      4.3.2.1.500           1        1          110          110   CON          0          0
   16:49:26.863293       F     esp      4.3.2.1           ->      1.2.3.4.0x17e*      145        0        47886            0   INT         63          0
   16:49:26.857781       F     esp      1.2.3.4          <->      4.3.2.1.0x0bf*       97       80        38766        64000   CON      45025          0
   16:49:28.055817             udp      1.2.3.4.500      <->      4.3.2.1.500           1        1          110          110   CON          0          0
   16:49:31.960600       F     esp      4.3.2.1           ->      1.2.3.4.0x17e*     7392        0      5598116            0   INT     944016          0
   16:49:31.974272       F     esp      1.2.3.4          <->      4.3.2.1.0x0bf*     2015      818       289482       654344   CON    3606147          0

I was wondering if this was a byte order issue and dug a bit into the code.
It is not clear to me how the byte order thing works.  See my next mail.

--chris