RE: automatic tunneling and v6 interoperation

["Christian Huitema" <huitema@windows.microsoft.com>]

> [on the traffic from Teredo host to {native,6to4}]
> > The hypothesis is that the relay is close to the "native host".
During
> > the transition period, the native ISP have an interest in providing
a
> > relay service for use by their native subscribers. Their subscribers
> > will enjoy better connectivity, i.e. will be happier. Providing the
> > service does not result in extra bandwidth requirement: the packets
are
> > exchanged between your subscribers and the Internet; they are simply
> > using a v6-v4 path instead of a v6-v6 path.
> 
> Could you describe how the Teredo host finds this relay (close to the
> native host)?  Note that I was not describing the traffic from the
native
> host to a Teredo host.

This is actually documented in section 4.1.5 of the Teredo draft. The
Teredo client sends an ICMP request to the IPv6 node, through the Teredo
server. The native host receives the request over the native interface,
and sends an ICMP reply. The ICMP reply is routed to the nearest Teredo
relay over native IPv6. (Not routed very far if the gateway is on the
host, but that is not necessary.) The Teredo client receives the reply
from the relay, and associates the native IPv6 address to the IPv4 & UDP
port of the relay. Further packets to that address are sent directly to
this relay.

There is a minimum of security in that exchange: the echo request
carries a nonce, and the Teredo host checks that the reply carries the
same nonce.

> > > 2) the implication of "no relays" deployment to v6
interoperability
> > >
> > > If there are no relays, note that every node a Teredo host needs
to
> > > communicate with has to implement and enable Teredo, as well as
> > publicize
> > > the Teredo addresses in the DNS in addition to the others,
correct?
> >
> > They don't actually publish a Teredo address. I showed you Tuesday
the
> > result of "ipconfig" on my XP laptop: I have public addresses
associated
> > to the wireless interface, using the 2001:468:19ff:80::/64 prefix
> > announced on the IETF network; the Teredo interface is available for
use
> > as a local relay, but only documents a link-local address on that
> > interface. (And we do not advertise link local addresses in the
DNS).
> 
> Again, the local relay onyl addresses the communication from native ->
> Teredo direction -- what about the other direction?  Where are the
relays,
> and if not, the address needs to be in the DNS, right?

No. See above.

> > > So, I don't think Teredo (or 6to4 for that matter, but it's
slightly
> > > better from the second perspective) really solves the "economics
of
> > > providing IPv6 service" -problem.  The only thing it seems to
solve,
> > to an
> > > extent, is a relatively smooth and direct IPv6 connectivity
between
> > Teredo
> > > hosts.  On the other hand, speaking to any other kind of nodes
(e.g.,
> > > 6to4, native, ...) is riddled with identical problems as the IPv6
> > > deployment at ISPs.
> >
> > See above. The trick is that the relays are deployed by other ISPS,
for
> > the benefits of their own customers.
> 
> No, this addresses (to some degree) only the other direction.  What
about
> those who sit behind an ISP who doesn't care about v6?

They will use the relay provided by the ISP who cares, and they will
only use that relay for the clients of this caring ISP.
 
> > An IPv6 ISP that really wants to isolate its customers from the
Teredo
> > technology can do that by providing native connectivity and a Teredo
> > relay (not a server). The ISP's customers will not need to implement
> > their own relay.
> 
> Again, what about the Teredo hosts which have no such relays?

They have to rely on a Teredo relay in the network.

-- Christian Huitema