6to4 clarification needed

[Tony Hain]

John,

Part of your confusion appears to be separating the destinations with
IPv4-derived prefixes from those with other global prefixes (frequently
called native). Sorting through the terminology, the 6to4-relay is a router
that ties the environment using native prefixes to the IPv4 network so that
6to4-routers at sites can access the native environment. When 6to4-routers
tunnel with each other they extract the IPv4 address, but when the
destination prefix has no IPv4 component they simply 'default' to the
6to4-relay.

Another part appears to be deciding when a host would act as its own
6to4-router vs. having the IPv4 router also provide that function. The
simple answer is this is a local decision, but if the router is providing
the service it should do so in both directions. It appears from your
description that you have a fully routed public IPv4 network; so all hosts
could (and in my opinion should) act as their own 6to4-router. If my
interpretation is true, they would talk over local tunnels rather than
through the 6bone. The implementation where the difference is easy to
describe is when the IPv4 environment is broken by NAT. In those cases the
NAT should act as the 6to4-router since it is the one that knows the current
public IPv4 address. If it is a dumb NAT, one of the hosts behind it could
be configured to act as the IPv6 router for the segment by providing it with
knowledge of the public IPv4 address, while configuring the NAT to pass
protocol 41 traffic to it (of course manual configuration and dynamically
allocated IPv4 addresses create an interesting operational concern). Either
way there is one device acting as the IPv6 router, sending RAs to the
segment so the other hosts believe they are on an IPv6 network. While it
would be possible to configure each host behind a NAT to act as its own
router for the outbound traffic, the inbound would have to go through the
one the NAT knows about, so the result is operational awkward and more
difficult to debug than necessary. It would work without any hack to
existing code necessary, but would require significant configuration
maintenance.

Hope this helps,
Tony

-----Original Message-----
From: owner-6bone@ISI.EDU [mailto:owner-6bone@ISI.EDU]On Behalf Of John
Comeau
Sent: Tuesday, June 12, 2001 9:42 AM
To: 6bone@ISI.EDU
Subject: 6to4 clarification needed

I've read all I could find on the web regarding 6to4, and am somewhat
confused. Since the IPV4 address of the target router is embedded in the
IPV6
address, shouldn't my IPV6 stack be smart enough to follow ipv4 routing for
all 6to4 traffic? So, my relay router would only need to be used for inbound
traffic and for non-6to4 IPV6 outbound traffic?

I was thinking of hacking the sit.c source to make it treat 6to4 traffic in
this manner. Or has someone already done this for linux? Is the freebsd stf
device what I'm looking for?

Of course I could, as some 6to4 descriptions say, route all my IPV6 traffic
to
the relay router, but isn't that somehow defeating the beauty of the 6to4
mechanism, being able to use all my available links to route the traffic?
Not
to mention, if another host on my network starts using 6to4, it doesn't make
much sense having to talk to him via the 6bone when he's reachable on my
LAN.
--
John Comeau - Chief Technology Officer
Dialtone Internet - Extremely Fast Web Systems
phone://954-581-0097x113  fax://954-581-7629
mailto://jcomeau@dialtoneinternet.net  http://www.dialtoneinternet.net