JORDI PALET MARTINEZ wrote:
However, for what it matters here, 6rd is native after
exiting from the ISP, same as 6to4 is native after
exiting from the 6to4 relay. As we may not be able to
know how "much" of the "native" IPv6 traffic is 6rd in
the last mile, I think we should consider all 6rd traffic
as native for those measurements, otherwise, we will be
biasing the data. Even it may be the case of an ISP using
6rd for some part of its network, and native for the other.
We may need to state "IPv6 native as measured, may be
encapsulated in the last mile".
I disagree.
6RD is not a last mile solution. With the existing levels of IPv6 traffic, an
ISP would deploy a couple of 6RD relays at their main IX. In a country such as
France, it means that a 6RD customer in Nice would see their IPv6 traffic being
encapsulated all the way to Paris over IPv4 crossing the entire ISP's network
for some 400 miles. In Spain, the really native part of the traffic would
possibly start in Madrid.
Imagine 2 customers in Barcelona, who are with two different ISPs using 6RD.
The 6RD relays are in Madrid for both, likely not much more than a few miles
away or even possibly in the same IX (I don't know the details of IXes in
Madrid). The 6RD traffic from one goes over IPv4 all the way to Madrid, then
goes native for a few miles to the other ISP's 6RD relay, then back over IPv4
to Barcelona. This is not native.
On top of it, like any other tunnel solution, it stinks because the IPv6
traffic goes back all the way to Madrid while, if it was IPv4, there is a good
possibility that the two ISPs are peering in Barcelona in a small colo and the
traffic never leaves the town.
The obvious: the solution is, duh, native IPv6 in Barcelona and the reason ISPs
are deploying 6RD is, duh, because they don't have it.
As mentioned before by Christian and other, 6RD is very similar to a tunnel
broker deployment, and the number and location of relays change everything.
Philip Homburg wrote:
At the moment, I have 5 IPv6 connections:
1) native
2) tunnel to my ISP
3) tunnel to HE
4) tunnel to SixXS
5) 6to4
Obviously, the first one is native and the others aren't,
Then we agree.
Roger Jørgensen wrote:
It mean IPv4 running on top of IPv6, IPv6 running on top
of IPv4 is not native. But this easy way of seeing it
only affect IPv4 and IPv6, not all of the others.
I could go for that; native IPv6 mostly means "no IPv4" for the agreed scope.
But I was trying to be more generic, the idea being to prevent a clever
marketing droid finding a way to tunnel IPv6 over avian carriers and calling it
native.
I guess it all boils down to, are we talking about end
to end native, or the transportation of the L3 protocol
being native?
If I understand you correctly (I think I do) I would say it ends up being the
same thing. What we want is no reliance on IPv4, what we don't want is limit
implementation options.
How good is a native IPv6 network if it has to be redone when decommissioning
IPv4? That's where the oxymoron is. I think a good way to see it would be:
remove all IPv4 end-to-end. If it still works, it's native IPv6. If not, it's
not native.
Michel.
_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www.ietf.org/mailman/listinfo/ietf