Changes at CISCO's 6bone-router

[Ran Atkinson]

	There have been various changes at cisco's 6bone-router recently.
This note summarises these to the list.  Cisco's RIPE database entry was just
updated accordingly.

	MERIT has moved their IPv6 tunnel from CISCO to CICnet in order to
cleanup the 6bone topology a bit now that CICnet is up and running.  Both
MERIT and CICnet are located in Ann Arbor and their internal connectivity is
much shorter than the prior MERIT--CISCO path.

	Similarly, SUMITOMO-US will be relocating their tunnel from CISCO to
DIGITAL-CA in the near future because DIGITAL-CA is fewer hops and because
DIGITAL-CA is already the upstream node from SUMITOMO-JP.  This also will tend
to clean up the 6bone topology a bit.

	The tunnel between G6 and CISCO has never worked well and has always
had very high rates of packet loss (possibly related to congestion on the
US--France IP links).  Hence that tunnel's prior designation as
"experimental".  The CISCO-G6 tunnel has now been taken down and no longer
exists.  The preferred path between CISCO and G6 is to transit NRL, based on
experimental measurements indicating that path works very reliably and with a
fine RTT.
	
	Over the past couple of months I have observed that some of the RIPng
updates received by CISCO have contained advertisements that were incorrect
and tended to cause routing loops.  While there are many ways that routing
loops might occur, one common issue might be related to redistributing local
static routes into RIPng for sites that are more than one hop away.  I would
like to encourage each site running RIPng or IDRP or any other routing
protocol to double-check their configuration to make sure that they aren't
initiating (as different from the usual retransmission of received
advertisements) advertisements for nodes that they can't actually reach or
that are more than one hop away.  One could postulate that some of the
connectivity issues shown on the various 6bone statistics web pages might be
due to routing loops of one sort or another.

	I would also like to encourage folks to give thought to how the 6bone
topology might be cleaned up in their neck of the woods.  Dorian has observed
that it would be nice if the 6bone topology reflected the underlying IPv4
topology.  I concur with this assessment.

Ran
rja@cisco.com



--