[ns2]: TCP-connections get pruned ?!

["Christoph Haenle" <christoph@starburstcom.com>]

Hi everybody,

I just setup a large network with a couple of TCP-connections and an SRM
connection in ns-2.0b18.
First, the TCP-connections start and later (>3sec) the SRM agents join the
multicast-group 0x8001.

I looks to me that for some strange reason, the traffic originating from
TCP-connections is treated and routed like multicast traffic, i.e. it is
broadcast along some spanning tree and pruned back:

Here is the relevant part of my tracefile:

======================================
[...]
+ 0.0370431 71 69 tcp 1000 ------ 0 71.12 198.1 0 1
- 0.0370431 71 69 tcp 1000 ------ 0 71.12 198.1 0 1
+ 0.0378431 69 70 tcp 1000 ------ 0 71.12 198.1 0 1
- 0.0378431 69 70 tcp 1000 ------ 0 71.12 198.1 0 1
+ 0.0378431 69 72 tcp 1000 ------ 0 71.12 198.1 0 1
- 0.0378431 69 72 tcp 1000 ------ 0 71.12 198.1 0 1
+ 0.0378431 69 182 tcp 1000 ------ 0 71.12 198.1 0 1
- 0.0378431 69 182 tcp 1000 ------ 0 71.12 198.1 0 1
+ 0.0386431 70 69 graft 80 ------ 30 70.0 69.0 -1 2
- 0.0386431 70 69 graft 80 ------ 30 70.0 69.0 -1 2
+ 0.0386431 72 69 graft 80 ------ 30 72.0 69.0 -1 3
- 0.0386431 72 69 graft 80 ------ 30 72.0 69.0 -1 3
+ 0.0430244 182 141 tcp 1000 ------ 0 71.12 198.1 0 1
- 0.0430244 182 141 tcp 1000 ------ 0 71.12 198.1 0 1
[...]

The packet with sequence number 1 originates at node 71 and has destination
198. As shown, it gets routed first to node 69, but then node 69 routes it
to node 70, 72 and 182 simultaneously for some reason. Furthermore, node 70
and 72 send back a prune (which appears as a "graft" in the logfile due to
a minor bug, but it's actually a prune).

Can anybody tell me what's going on here?

    TX,
       
       -Chris.