Maybe I am missing something basic, but there seems to be a bit of a
disconnect on our assumption about what we need to have in in-home rotuers.
On the one hand, earlier discussion emphasized that we need to use
protocols and implementations that are well tested, robust, and simple
to use.
On the other hand, we have an expectation that these same boxes are
going to perform source address based forwarding selection, something
with which there is very little experience anywhere?
Yours,
Joel
On 11/14/2011 7:41 PM, Michael Richardson wrote:
"Guillaume" == Guillaume Habault<guillaume.haba...@gmail.com> writes:
Guillaume> the host select the appropriate address, the routing
Guillaume> protocol may route the
Guillaume> message to the wrong Customer Edge Router (especially if
Guillaume> we are in the case
Guillaume> where there is more than one Customer Edge Router in the
Guillaume> home network.
We have assumed that the home gateways will cooperate and provide source
address routing, making sure that packets are routed out the link which
can accept that source address.
Guillaume> However, it could work if Customer Edge Router can talk to
each other)...
Guillaume> And in the section 3.4, it is mentionned that :
Guillaume> To Support multihoming within a homenet, a routing protocol
that
Guillaume> can make routing decisions based on source and destination
Guillaume> addresses is desirable, to avoid upstream ISP ingress filtering
Guillaume> problems.
Guillaume> Do you plan to use both solution? or does it depend on
Guillaume> the topology? Or
Guillaume> does the aim is to be generic and let implementer decide
Guillaume> about the solution
Guillaume> to use ?
You need both pieces.
(Please consider a quoting mechanism that is more robust)
_______________________________________________
homenet mailing list
homenet@ietf.org
https://www.ietf.org/mailman/listinfo/homenet