[bess] Comment at the mic

[Robert Raszuk <[e-mail]>]

Hey Keyur,

I would like to share my perspective on your comment made at the BESS
yesterday.

What you pointed out that VPN demux should be removed or renewed when we
rewrite bgp next hop is very true in 4364 world or even in EVPN world where
VPN label is of local significance.

But in Srihari's proposal VPN demux may be of domain wide significance
hence mandating its removal at each EBGP boundary in the spec would be a
bad hint.

See with SID indicating to which VRF packet belongs applied consistently by
say NMS your option C becomes seamless without any "next-hop-unchanged"
hacks nor making PE-PE for multihomed sites failover requires any fancy
shadow LFIBs or control plane "taps" into updates which just pass by.

So while in general my personal opinion is that we do not need yet one more
way of  protocol encoding to accomplish L3VPNs which already is shipping in
the form of 4364 and EVPN standards if WG decides to proceed let's learn
from the past experience here.

Cheers,
R.