On killing IPv6 transition mechanisms

[Martin Millnert]

On Thu, 2010-03-11 at 13:56 +0100, Jeroen Massar wrote:
> > Well, if providing 6to4 relays for the gigabits upon gigabits of 6to4
> > IPv6 traffic there is on the Internet is actually harmful for the
> > deployment of IPv6, we'd gladly stop to provide the service.
> 
> Wow, please provide those statistics somewhere. I am sure lots of people
> would love to see this, especially with a small analysis of what the
> traffic most likely is (the answer is most very likely NNTP).

We have our own stats at http://stats.csbnet.se/public/ipv6/ .
Tele2 has some at http://ipv6.tele2.net/6to4_stats.php .
Sweden is pretty much littered with 6to4 relays and I do not have access
to a complete set of statistics.  So I extrapolate. Our reach for our
own 6to4-relay into other people's IPv4 networks is very limited. So I'm
estimating something on the order of 1-5 gigabits at peak hour in
Sweden. 3301 run their own relays.  It is hard to get a good view of
this.  The best yield in result vs effort would probably be to measure
SFlow stats on the Netnod IX's, I guess.

> > Same for
> > the Teredo relay we run, which considering your stance on providing 6to4
> > relays, I'm sure you are ten times as eager to kill off. :)
> 
> Teredo has the same issues as 6to4: anycast in both IPv4 and IPv6 thus
> you never know the path that the packets will follow, thus it is
> horribly hard to debug; unless you have access to every single hop in
> the path of course.

Except Teredo isn't anycast in v4.

See http://technet.microsoft.com/en-us/library/bb457011.aspx

On v4 client-side, it's client-server, and client-relay, where both are
unicast UDP connections. A v6 connection over Teredo is in that regard
assymetric, but not in the same way as 6to4. A Teredo ipv6 connection
settles down at the relay closest (in BGP terms) to the non-Teredo v6
side.  Teredo-Teredo is a special case (see fig 15 and 16).

That the teredo server used is encoded in the Teredo address
unfortunately doesn't help that much in debugging network issues. This
falls on operators. 

> THAT is the reason why 6to4 and Teredo are a bad thing: debugging.
> Over all these years there have been a lot of problems with those setups
> and people keep on complaining as it is causing brokeness.
> And no, there is no magic way to fix this, if you have it though, please
> illuminate the rest of the world with your awesome idea.

I'm no magician, sadly.  RFC3068 does require good monitoring in section
4.4, of the relay by its operating party. That's the best you can do :/
Same with Teredo.

Cheers,
Martin
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part
Url : http://lists.cluenet.de/pipermail/ipv6-ops/attachments/20100311/7bafb7d0/attachment-0001.bin