Re: [Idr] IDR WG LC on draft-ietf-idr-ix-bgp-route-server (2/2/2015 - 2/16/2015)

[Randy Bush <[e-mail]>]

>> while A may not announce anything to B, B may have gifts for A
> That's fine, there is still a valid forwarding path here, why would we
> not set up a (BFD) session?

you would.  the point is that B can not count on receiving a route to A
to tell B to set up bfd to A

>>>>> which provides some assurance that you know your potential bfd peer
>>>>> set.
>>> Is't this the point of the RS, to facilitate in the face of the
>>> unknown?  with the above approach, assuming I filter against the RS,
>>> if I wanted to, would I not still reach the same outcome?
> 
> I was just commenting on how I feel that this doesn't violate the RS
> model.

cool.  the exercise started with arnold asking what would make me more
comfortable with RSs given that i like my data plane to be congruent
with my control plane.

> Assuming you filter properly (including asking the RS not to spread
> your prefixes to those you don't want it to), you shouldn't get any
> unsolicited BFD setup requests.

makes sense to me.  at bknix with the gang who are figuring out how to
tell members to use ixp-mangler to only configure a bird peering if both
parties agree.

randy