On Thu, 21 Jun 2012 22:07:46 +0200, Maurice Janssen wrote:
On 06/16/2012 10:05 PM, Markku Miettinen wrote:
I would expect all kinds of packet loss with any tunneled IPv6
connectivity.
Why is that? I understand that with a tunnel IPv6 connectivity
depends on a working IPv4 connection, but with a stable IPv4
connection to a good tunnel provider, I think there shouldn't be any
more random packet loss than with native IPv6.
That's exactly the problem. Native IPv4 is dependent on IPv4 infra and
IPv6 likewise of IPv6 infra. Any kind of tunneled connectivity is
dependent on both.
In worst case scenario with IPv6 the traffic goes first into 6to4
tunnel (or any other tunnel), and tunnel end possibly is not even
nearby, which is then possibly routed back near to point of origin where
another tunnel end is located. From there it possibly is routed back to
close to 6to4 proxy. In this imaginary case we have 3xRTT of optimal,
with 2 of them going in possibly different IPv4 routes and one IPv6
route. If there is a X% chance of getting packets lost with any native
connection, you should pretty easily get how big the probability of loss
is in this case.
So in principal I wouldn't run any services on any tunneled
connections. In most of the cases you don't see anything, but chances of
having trouble are much larger than with native connectivity.
--
Markku Miettinen
_______________________________________________
pool mailing list
[email protected]
http://lists.ntp.org/listinfo/pool
