[Idr] BGP routes with color, Question 2: Route origination and propagation

[Jeffrey Haas <[e-mail]>]

As promised during the recent IDR interim, we want to start a set of threads
to the IDR mailing list to clarify points discussed during that interim
about the BGP routes with color proposals.  These questions are directed to
the authors of those proposals.  The responses to those questions will be
used to drive Working Group discussion about the solution space.


Question 2: Route origination and propagation.

A point of strong discussion during the recent IDR interim covering BGP
routes with color was related to where routes within a given proposal were
originated, and what their expected propagation was within the proposal.

Two examples during the discussion:
1. Origination of a route with color at an Autonomous System Border Router
(ASBR) for a domain containing an endpoint address.

2. Origination at the Provider Edge device hosting the endpoint.

There may be other scenarios.

A route-selection "pinch point" is any place in the BGP topology wherein
multiple paths for the same NLRI may be received, only one is selected, and
is propagated.  There is a loss of forwarding-diversity at such points under
some circumstances.  Mechanisms that permit additional BGP routes to be
propagated can address such loss of forwarding diversity.

With regard to propagation through route-selection pinch-points, the interim
discussion illustrated that there were differences of opinion as to when it
made sense at an ASBR to use next-hop-self or propagate more than one path
across the peering session.

The above inputs to the scenario(s) culminate in an ingress device having
one or more than one instances of the received path.

What is the default scenario you expect your proposal to use?

Are there situations where you may want to have origination or propagation
behave differently than your default scenario?  If so, how does that work?

Please illustrate the scenarios with a topology for discussion.

-- Jeff