Packet duplication on "Places..." SAP announcement

[Mark Handley <]

I'm seeing large packet duplication on the Places All Over the World
session announcements:

Waiting to accumulate statistics... Results after 9 seconds:
 
  Source        Response Dest    Overall     Packet Statistics For Traffic From
   * * *        224.0.1.32       Packet      130.240.64.20 To 224.2.127.254
     v       __/  rtt  211 ms     Rate       Lost/Sent = Pct  Rate
192.36.145.3    UPPER-GW.Sunet.SE Reached RP/Core
     v     ^      ttl    1        31 pps        0/99   =  0%  11 pps
192.36.145.5   
198.67.142.126  icm-kth-11-P0/0/0.icp.net 
     v     ^      ttl    2        97 pps       17/99   = 17%  11 pps
198.67.142.125 
198.67.142.134  icm-tl2-12-P4/0-STM1.icp.net 
     v     ^      ttl    3       103 pps      -17/82   =-20%   9 pps
198.67.142.133 
198.67.142.78   icm-tl2-10-P1/0-STM1.icp.net 
     v     ^      ttl    4        78 pps       99/99   =100%  11 pps
198.67.142.77  
198.67.142.74   icm-pen-11-P4/0-OC3C.icp.net 
     v     ^      ttl    5         0 pps      -96/0    = --%   0 pps
198.67.142.73  
192.36.137.1    schnell.ebone.net 
     v     ^      ttl   38        91 pps       66/96   = 69%  10 pps
207.41.200.99   pen-mbone-1.sprintlink.net 
     v     ^      ttl   71       101 pps       11/30   = 37%   3 pps
204.70.64.77   
204.70.64.61    dec3800-2-fddi-0.WestOrange.mci.net 
     v     ^      ttl   71       193 pps        0/19   =  0%   2 pps
204.70.74.77    dec3800-2-fddi-1.Washington.mci.net 
     v     ^      ttl   71       430 pps        0/19   =  0%   2 pps
204.70.176.23  
204.70.131.68   mbone.Reston.mci.net 
     v     ^      ttl   71       346 pps        ?/19           2 pps
204.147.128.189
204.147.128.1   cs.res.vbns.net 
     v     ^      ttl   71       369 pps
204.147.128.2  
140.173.4.29    isiepc-mitpc.cairn.net 
     v     ^      ttl   71       259 pps        0/19   =  0%   2 pps
140.173.4.30   
18.26.64.2      dartnet.lcs.mit.edu 
     v     ^      ttl   71       317 pps        ?/19           2 pps
18.26.64.1     
18.26.0.165     video-room-mrouter.lcs.mit.edu 
     v      \__   ttl   71         2 pps
18.26.0.4       18.26.0.4
  Receiver      Query Source


Attempting to read this, I think the duplication is happenning very
near to the source resulting in at at least 100 copies of each
announcement packet, that the ICM routers are particularly bad at
keeping reasonable stats, and that many of the packets were then
dropped at a TTL threshold from Ebone to Sprintlink.  Tcpdump tells me
that the IP id on all the packets is the same, so it's not likely to
be the fault of the announcing software.  Each duplicate has a TTL
three less that the previous one indicating that the loop causing this
is probably three hops long.

I'm not seeing similar duplication on the video, so perhaps it's got
something to do with the Cisco at Sunet joining the SAP announcement
group (I gather that ciscos do this so they can give you named
sessions in their stats rather than just IP addresses).

Cheers,
	Mark