NT strangness

[Richard Draves]

>     The question is, what is the usefulness of the 
> Carpenter/Jung draft
> compatible address? does it allow the v4 only hosts to speak 
> to v6, or what?

The Carpenter/Jung 6-over-4 draft does not allow v4 only hosts to speak to
v6. (That's what translators do. Several groups are also working on
translators. For example see
http://www.cs.washington.edu/research/networking/napt/.)

6-over-4 doesn't use any special type of address, that's one of the great
things about it. Like configured & automatic tunnels, 6-over-4 is a way of
transmitting v6 packets over a v4 network by encapsulating them in v4
packets. Unlike configured & automatic tunnels, 6-over-4 treats the v4
network as a real link layer, with link-local addresses, Neighbor Discovery,
etc.

Here's a scenario where 6-over-4 is incredibly useful: Suppose you have a
campus network with many v4-only routers. The campus network supports v4
multicast across the campus. Sprinkled around the campus are v6 machines
(which also support v4) that you want to connect to the 6bone, but for
whatever reason (availability, cost, time, stability) you haven't yet
upgraded the campus routers to support v6. What to do?

One solution is to manually setup a network of configured tunnels between
all the v6 machines on a campus and a campus gateway to the 6bone. There's a
lot of effort here to create and maintain all the configured tunnels. Every
new v6 machine on campus requires additional work.

A much better solution is to use 6-over-4 to connect all the campus v6
machines to the campus 6bone gateway. Because the campus v4 network is a
virtual link layer for v6, you can use stateless autoconfiguration. New v6
machines on campus will automatically discover the 6bone gateway and
configure themselves with global 6bone addresses! The only configured tunnel
that you need to create & maintain is between the campus 6bone gateway and
the 6bone backbone.

The only caveat is that the campus v4 network must support v4 multicast.

Rich