On (2013-04-02 10:57 -0700), Pedro Roque Marques wrote:

> The document describes the behavior of the end-system and the external
> Forwarder. In my mind, the scenario where Q-in-Q (or other L2 encap) is
> used between the end-system and the Forwarder because of additional L2
> hops should not need to be explicitly described.

I don't understand. Are you saying even if host would double tag, that
would not be relevant to Forwarder? Looking at page 12 I don't know how to
signal two tags to Forwarder.

> Searching thought the document /32 seems to appear in examples. If you
> don't mind pointing out explicitly where you believe the text should be

I'm probably reading it wrong, page 13, request 2, associate node is one
that caught my eye.

> The route-server is not on the forwarding path.

I was thinking native MPLS-to-the-edge in DC, where I only have VPN
Forwarder between my host and core MPLS PE (preferably right in the host as
'virtual router'). Then if I'd use OptB, I'd need to collapse route-server
and VPN forwarder into same box?

> The data-center gateway typically implements option B (or IBGP plus
> next-hop self) and stitches the MPLS-over-GRE LSP segment (internal to
> the DC) with an MPLS over L2 LSP segment outside the data-center, in
> scenarios where the service is delivered to a WAN L3VPN.

Would this be true for green-field as well? I'd really love to see solution
where I can plug host anywhere in MPLS network.

-- 
  ++ytti

Reply via email to