The MPLS WG Archive

[Chris Flores <chris.flores@onfiber.com>]

Eric - 

I have not missed the point, although I may not have explained my point
adequately. I agree, packets traversing the MPLS VPN do not provide the
appropriate information to route via the core if the VPN failed. Obviously,
packets traversing the MPLS VPN originate from the customer's network and
would most likely use private address space (i.e. RFC 1918). 

My point is this. BGP (specifically iBGP) is needed in the core for packets
not utilizing a MPLS VPN or if MPLS fails. It may be the case that there is
a full mesh of MPLS LSPs (PE-PE) that serve all traffic flows within the
network - customers, external peers, etc. What if MPLS fails? Yes, VPN
customers are out of luck till the MPLS VPN can be re-established. However,
external peer traffic, etc should still route as if the network did not
implement MPLS.

c

  

-----Original Message-----
From: Eric Rosen [mailto:erosen@cisco.com]
Sent: Thursday, September 28, 2000 3:11 PM
To: Chris Flores
Cc: 'Eric Osborne'; 'Javier Antich'; Michel Redondo Ferrero; mpls@UU.NET
Subject: Re: MPLS/BGP routing question 



Chris> that was my  point. in this scenario, you would not  want to turn BGP
Chris> off :) 

I'm afraid you've missed the point.   In the VPN scenario, it doesn't do any
good to  run BGP in the  core, because the IP  header of the  packets do not
contain the information you need to match the packet to its route. 

Besides which you could never hope to hold all routes from all VPNs in every
core router anyway.