[karp] karp-ops-model: VRFs

[Sam Hartman <[e-mail]>]

Hi.
dacheng and I are working through karp-ops-model updates.
Section 3.4 of draft-ietf-karp-ops-model discusses VRFs and the key
table.
It says that probably the right answer is that each VRF should have a
key table but more analysis is required.

When i wrote that more analysis might be required I was mostly thinking
about inter-VRF routing.
I've thought more about this, but it seems like most of the use cases I
can think of fall into two categories:

1) A routing peering relationship within a VRF. There you want the VRF's
key table

2) A service provider routing peering that exchanges routes for multiple
VRFs. There you want a key from the non-virtual VRF used by the service
provider.

I've convinced myself if that in the vast majority of cases it's fine to
model the key table as per-VRF.
So, unless there are objections I'll update the text to make this as a
conceptual recommendation.
Obviously the specific details are left up to the implementation and
this text will not be stating any normative requirements.

--Sam