Re: [Idr] draft-dong-idr-node-target-ext-comm-05.txt - WG Adoption and IPR call (9/27 to 10/11/2022) - Extended an Additional week to 10/18/2022

[Jeffrey Haas <[e-mail]>]

Robert,

On Thu, Oct 20, 2022 at 12:20:59AM +0200, Robert Raszuk wrote:
> >
> > The way I understand it by reading Jie's use case that this VRF scoped
> > too. It could be nodes scoped and VRF scoped.
> >
> > My understanding was that this was a logical AND case in such
> > circumstances.  Thus, a specific node for a specific VPN and would have
> > route targets in addition to the node-target.
> >
> 
> As you know for many years we have been addressing such cases by allocating
> a new RT. My point also already shared is that if the target node has 1000s
> VRFs your granularity of logical AND of RT and NT is simply not sufficient.

I won't dispute this point.  This is simply to queue the desired
funtionality as part of the discussion for the solution.

> In any case what was not discussed at all here is how that information on
> what BGP_IDs to apply is communicated to sourcing nodes. It may very well
> turn orders of magnitude easier to simply configure target nodes with a
> required configuration as opposed to sending policy to source nodes and
> modify the entire network to accommodate new auto filtering rules.

You mean you're not already using BGP-LS as a general discovery layer
interacting with your provisioning systems?

Hm.

-- Jeff