On Sun, Jul 3, 2011 at 5:49 AM, Mark Smith <
ipng(_at_)69706e6720323030352d30312d31340a(_dot_)nosense(_dot_)org> wrote:
Declaring 6to4 to be historic might encourage native IPv6 deployment,
but I think it will also make trialing IPv6 much harder.
We don't need to trial the IPv6 protocol. There are hundreds of thousands of
native users accessing production-grade IPv6 services, like Google's, every
day. We know the IPv6 protocol works. ISPs do need to trial IPv6 deployment.
But 6to4 does not help there, because 6to4 is not deployed by ISPs. They
will need to trial native deployments, or 6rd. If *users* want to trial IPv6
until native IPv6 is available, then they can use configured tunnels.
The involvement in World IPv6 day by large content providers and the
apparent lack of significant problems would be suggest the opposite is
now the case. Google continuing to provide youtube video content to 6to4
tunnel users (such as myself), nearly a month later, suggests that any
problems with it are tractable.
I would assert that the problems with 6to4 are not tractable without
disabling 6to4. Our IPv6 brokenness statistics for before and after World
IPv6 Day are very similar.
6to4 users are easy to spot by the 2002::/16 prefix, so if Google needed to
they could probably quite easily limit their IPv6 content delivery to native
only IPv6 users.
No, that's not how it works. There is no problem with 6to4 when it works as
well as IPv4. The problem is that 6to4 only works *at all* (never mind works
"as well as IPv4") 80% of the time.
Unfortunately, in the 20% of the time that it's not working, Google has no
idea that the user has a 2002::/16 address. Google only knows, after the
fact, that the user suffered a 20 or 75-second timeout and was not happy. So
it would serve no purpose to avoid serving users that successfully connect
from 2002::/16 addresses; once the AAAA record is handed out, the damage is
done. What Google could do, however, is stop handing out AAAA records to
networks that have significant number of 6to4 users in the future. We're
considering this.
An update on that data would be useful.
As I said before, our data on IPv6 brokenness did not change significantly
after World IPv6 Day. Can you take that as a proxy that 6to4 has not
improved?
_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www.ietf.org/mailman/listinfo/ietf