On Jun 14, 2011, at 2:16 PM, Tony Hain wrote:
Keith is correct, and the further issue is that the *-only-* reason the
'poorly managed' relays are in the path is that the content providers are
refusing to deploy the matching 6to4 router that would take a direct
connection from the customer.
Well, to be fair, they weren't supposed to have to get involved with 6to4 if
they didn't want to. To the extent that 6to4 causes pain for operators that
don't provide relay routers, this is a mostly-unanticipated and unintended
consequence. (Operators weren't expected to *like* 6to4, because at the time
6to4 was written there was a general dislike for tunnels in general. But
neither was it expected to cause them as much pain as it apparently does.)
6to4 direct between the content and consumer is still an 'unmanaged' tunnel
which takes exactly the same path as IPv4 would, so the 'badness' is not due
to managed vs. not. In the grand scheme of things, the last thing the
content providers want is for the network to wrest control over streams into
a walled-garden model. Unfortunately they are not thinking this through,
they are just whining because the deployment of a second prefix on their
content servers does not conform to their IPv4-think dream world.
Though it's clear that some operators want to impose the walled garden, I think
6to4 can cause some pain even for those that don't. I'm hopeful that this pain
will be short-lived and will be remedied in the short term by (a) relay router
operators acting more responsibly, (b) content-providers providing their own
6to4 addresses, their own 6to4 gateways for return traffic, and/or (c) (less
ideally) DNS tricks to minimize IPv6 exposure to networks known to have trouble
with 6to4. In the not-so-much-longer term I hope the pain is relieved by
widespread adoption of native IPv6 or at least "managed" transition schemes
such as 6rd.
But did we really expect there to not be growing pains associated with IPv6
transition and deployment? We certainly had them with IPv4 transition and
deployment. After all, 6to4 is pretty much the ARPAnet/MILnet to Internet
transition revisited; that's where I got the idea.
Ietf mailing list