[ih] Ingrid Burrington on North Virginia

[Joe Touch]

On 1/12/2016 6:21 AM, Vint Cerf wrote:
> on the assumption that not much traffic is black holed (could be a bad
> assumption, I know), then the total traffic injected into any network
> that is part of the Internet could be considered "Internet traffic." The
> low black hole assumption means that the injected traffic comes out
> somewhere on a network of the Internet, so you only need the one
> measurement to avoid double counting.

Yes, but even without persistent black-holes, the Internet also drops
packets occasionally (e.g., routers that reboot or die while packets are
in transit, as well as more common buffer overloads).

So, as Ted suggested, we're back to definitions. Is "traffic" a packet
that used the Internet at some point or one that successfully traversed it?

There's also other traffic that isn't "injected" into a net but uses it
anyway, e.g., signaling traffic - i.e., do ICMPs count? How about ARPs?
and routing protocols, not typically visible at the edge?

IMO, these "dark packets" are hard to determine.

Joe