Question about 6to4

Martin List-Petersen martin at
Fri May 15 00:04:31 CEST 2009

Steve Wilcox wrote:
> Hey Ted,
>  you sort of answered your own question - the aggregate is announced as an
> anycast and the v6 routing table doesn't know where the individual v6
> packets are destined.
> So the v6 packets find their way to the nearest 6to4 relay, that then
> converts to v4 and its routed out as v4. As you say HE is one major sink for
> that, as a result of them being so well connected.
> Thats one of the downsides with 6to4 - the packet may go in the wrong
> direction in v6 before passing through the relay and then heading in the
> opposite direction to find the v4 endpoint.

And just to add to that, the v4 side works exactly the same. The anycast
prefix there is

In our case (we're in Ireland), the relays in our path were in Sweden,
Germany and Italy, which never is a good result, so we set our own
gateway up.

Now, talking about 6to4, 6to4 is never to be preferred. If you can get
people to use native IPv6, you should at any time prefer that solution,
or 6in4 tunnels for that sake.

The issue with 6to4 is exactly it's anycast nature.

Let's say, I'm using 6to4 and I'm connecting t a native IPv6 host. My
traffic goes to the nearest gateway announcing and is
translated to IPv6 there. Now the IPv6 host answers, and the answer goes
to it's nearest 6to4 host that announces 2002::/16.

This gateway does not have to be the same as the one I used and if any
one of them is but, nothing works. The issue is, that without having
access to both ends of that connection, I can't even troubleshoot it, so
it's often very difficult to diagnose where things went wrong with 6to4
to solve it.

It is a migration mechanism and nothing more than that. It should be
treated like that.

Kind regards,
Martin List-Petersen

> On Thu, May 14, 2009 at 10:32 PM, Ted Mittelstaedt <tedm at> wrote:
>> Hi All,
>>  I apologize in advance if this has been asked before a million times but
>> I have what is probably a stupid question about 6to4.
>>  We are in process of connecting to native IPv6, I am currently getting
>> the IPv6 BGP table from our upstream.
>>  I see that Linksys is supporting IPv6 out-of-the-box in some of it's
>> routers, via 6to4
>>  I figured this might be useful for some of our customers, to setup
>> a 6to4 relay router for these Linksys devices to use.  I have found plenty
>> of info on the Internet to setup a router as a 6to4 relay.
>>  My question concerns how exactly 6to4 -works-
>>  My understanding is that RFC3068 defines 2002:c058:6301:: as the
>> anycast for the (in this case) customer router to find the 6to4 relay I
>> create.  And that a 6to4 relay then uses a manufactured 2002:: IPv6
>> address formed by using it's IPv4 address with the 2002:: prefix, and
>> that any 6to4 routers tunneled into it are using their manufactured
>> 2002:xxxx:: addresses.
>>  So, if this is the case then wouldn't every 6to4 relay that's advertising
>> on the Intenet be present in the IPv6 BGP table?
>>  In looking at the various IPv6 looking glasses on the Internet, I see
>> a handful of 2002:: routes out there.
>>  But there seems to be no consistency anywhere.  Most of the looking
>> glasses
>> seem to show Hurricaine Electric's 2002:: advertisement, including my
>> own table.
>>  But beyond that, the advertisements for other relays seem to exist in
>> one router, but not in others.
>>  Advertisements for native routes do seem to be consistent.
>>  How exactly do routers know where to forward IPv6 packets destined for
>> a given 6to4   2002:: address if there is no route in their table?  And if
>> they are just sending 6to4 traffic to HE then how does HE know what to
>> do with it if it's not for their network?
>> Thanks,
>> Ted

Airwire - Ag Nascadh Pobail an Iarthair
Phone: 091-865 968

More information about the ipv6-ops mailing list