Bogus bogon?

Mon Jul 8 23:20:05 UTC 2002

Tony Tauber wrote:
> On Mon, 8 Jul 2002, Rob Thomas wrote:
> 
> > Hi, John.
> >
> > > 192.88.99.0/24 which is the 6to4 anycast network.  Do we 
> really want 
> > > to be filtering that prefix?
> >
> > Good question.  I'm re-reading RFC 3068 now, and the RFC appears to 
> > allow for the advertisement of this prefix into the global 
> table. I'm 
> > wondering if this is wise, however.  It seems this prefix 
> would best 
> > be used internally.  What do others think?
> >
> > Thanks,
> > Rob.
> 
> I believe the idea is that it typically comes from outside 
> one's domain because it's supposed to allow edge islands of 
> IPv6 to find the mainland.  If you had a connection to the 
> mainland under your control, you wouldn't need this trick to 
> get there.
> 
> At least that's my understanding.
> 
> Tony
> 

Close. The prefix should be advertised to the customers of the network
which has the interconnect. If those customers are other SPs, the
downstream will receive it from outside, and should advertise it to
their customers. If a SP chooses to provide the interconnect service, it
may still want to receive that prefix from outside as a backup. 

The point is that it is a well-known prefix that allows end customers to
deploy IPv6 even if their direct SP chooses not to support it on their
timeframe. It is designed on the assumption that the first hop SP is
doing nothing about IPv6 deployment, and that includes not filtering the
prefix. It would be wise to continue announcing it even after the SP
starts announcing native IPv6 to customers because there will be some
islands out there still, and it will be more efficient for the endpoints
to directly tunnel over IPv4 than to run it through a central gateway
service in each SP.

Clearly each origin AS announcing the prefix will want to limit their
exposure to their customers, but SPs without an interconnect should not
have a problem accepting  it. 

Tony