[ppml] [narten at us.ibm.com: PI addressing in IPv6 advances in ARIN]

[Tony Hain]

Owen DeLong wrote:
> ...
> On the internet, there simply isn't a good mechanism for tying either
> end of the connection to who should pay for it.  To further complicate
> this matter, there is the issue of spoofed address traffic.  Should
> I really be billed for someone originating a terrabyte of traffic I
> didn't know existed, just because they picked one of my IP addresses
> at random?

In general I agree with your post, but this point seems like it need some
further thought. Say there were a settlement process in place. A transit
provider would look to the volume over the interconnect, likely not at the
source address. This would follow down the chain until the loop provider was
left to sort out which customer it came from. Their first inclination would
be to look at the address, but if spoofed would likely not even be one of
their customers. This would lead them to argue that the upstream was wrong,
then when pushed back to eat the bill, they would have to implement strict
RPF at their customer edges to block the nonsense. They might also move to
volume charging on each loop which would pass the cost along to the real
source. Assuming the spoofed traffic was from a zombie, it would provide the
cost incentive to get the machine cleaned up. 

In the grand scheme of things there is no fair way to judge who initiated
traffic flow in any particular direction, so fine-grained per-flow volume
based charging is not operationally deployable.

Tony