[rrg] Fundamental objections to a host-based scalable routing solution

[Dino Farinacci]

> Dino Farinacci wrote:
>
>> The hosts do not have visibility of the path. They only know the  
>> other
>> end is sending them packets. And you get this as well from a
>> router-based approach.
>
> Dino,
>
> in both host-based and network based solutions, the availability  
> status
> of a path needs to be measured through the delivery, or the lack of
> delivery, of packets.  My point is that the faithfulness of such
> measurements differs between host-based and network-based solutions:

Agree, but thats much different than having "end-to-end path  
visibility". Visibility of the path is knowing each node along the  
path which either approach doesn't have by default.

> For host-based solutions, delivery of a packet is reliable evidence
> that a path is available because the packet has traveled the entire
> end-to-end path.  For network-based solutions, delivery of a packet is
> less reliable evidence, because it doesn't account for path failures
> downstream of the router receiving the packet.

Yes, but there is typical packet loss in the core than the edges. So  
it's a probability game.

Dino