We've tripped over an oddity in our routing set up and I'm baffled. I hope this is a suitable list to ask for assistance on. We use BGP on our metropolitan network with private ASs. At the core are two Cat 6500s running 12.2(33)SXI1. They run IBGP between them with no sync and, obviously, no IGP on a dedicated point-to-point 2x10GE link. That has worked beautifully for IPv4 for years. With IPv6 the routing continually bounces. Routing on the EBGP peerings to JANET and to other routers in the MAN are stable. What seems to happen is that each router advertises itself as the next hop for prefixes in the IBGP. Each router has a connected route for the /126 between them and a local /128 for its own interface address. Each router then installs an IBGP route for the /128 of its neighbour routed via that same address. That presumably overrides the /126 of the connected interface. Turning on ipv6 routing debugging shows the route to the remote /128 being invalidated, all routes via it being removed and then after a suitable pause - about a minute - being reinstalled and the process beginning again. The remote neighbour address (the remote /128) is pingable during the entire process. Anyone got any ideas about what we've missed here? Sam Sam Wilson Network Team, IT Infrastructure Information Services, The University of Edinburgh Edinburgh, Scotland, UK -- The University of Edinburgh is a charitable body, registered in Scotland, with registration number SC005336.