Re: [Idr] draft-vinod-lavallee-bgp-optimal-route-reflection-00.txt

[Robert Raszuk <[e-mail]>]

Vinod,

Let me also describe how all goals of your draft are met by IDR WG doc 
as well as by associated with it Next-Hop SAFI draft 
(draft-varlashkin-bgp-nh-cost-01) without any need for using SPF on the 
RRs.

Imagine you have number of PEs which per operator's logic know their 
exit preference. Using new NH-SAFI they can idicate to RR their own 
local metric to each of available next hop. If one or N of those paths 
with previously advertised next hops goes away RR will immediately 
replace it with new set of available paths.

This is simplest form of local operator decision on what is the order of 
paths in the network yet number of paths received by PE is decoupled 
from the preference (unlike in your draft). It is also robust as at any 
point it will not result in the customer's outage.

Here I would like to thank you for highlighting within our discussion 
that PE driven provider's optimal path selection is requested by some 
operators. This will clearly help to progress 
draft-varlashkin-bgp-nh-cost-01 and it's implementations further and it 
offers a very good tool to accomplish such requirement.

Best regards,
R.