> -----Original Message-----
> From: Uma Chunduri [mailto:[email protected]]
> Sent: Tuesday, April 12, 2016 5:47 AM
> To: Eric C Rosen; Xuxiaohu; [email protected]
> Cc: [email protected]
> Subject: RE: [mpls] Clarification on the motivation of
> draft-xu-spring-islands-connection-over-ip-05
> 
> Hi Eric,
> 
> Quick  response  In-line [Uma]:
> 
> --
> Uma C.
> 
> 
> -----Original Message-----
> From: mpls [mailto:[email protected]] On Behalf Of Eric C Rosen
> Sent: Monday, April 11, 2016 11:51 AM
> To: Xuxiaohu; [email protected]
> Cc: [email protected]
> Subject: Re: [mpls] Clarification on the motivation of
> draft-xu-spring-islands-connection-over-ip-05
> 
> On 4/7/2016 6:39 PM, Xuxiaohu wrote:
> > [Xiaohu] The FEC associated the above label L is the /32 or 128/ prefix of 
> > node
> N. When the IGP next-hop towards that FEC is a non-MPLS node, the LSR
> receiving the above MPLS packet with top label of L is desired to forward that
> MPLS packet towards node N via an IP-based tunnel. In this case, the node N is
> the remote peer for that FEC.
> 
> Yes, in this case, the tunnel's receive endpoint can be regarded as a remote 
> label
> distribution peer of the tunnel's  transmit endpoint.
> 
> However, in the above case I don't think it is obvious that you want to 
> IP-tunnel
> the packet to N.
> You might just want to tunnel it around the
> non-MPLS node.   In that case, the tunnel's receive endpoint can still
> be regarded as a remote label distribution peer of the transmit endpoint, but 
> the
> receive endpoint is not N.
> 
> [Uma]:  Sure, this can be done; if at all if there is a shortest path node 
> towards
> N (supporting MPLS/SR) and if we can tunnel to that node yes, packet can be
> delivered to N eventually.
>                However, this involves (additional computation/adjustment)
> computation by all the boarder nodes to deliver it to the shortest SR node
> towards N.
>                In that case, actually I would have expected operator to put
> that additional node Label as the top label instead of N. If the former has to
> happen operator has to
>                exercise how this should happen through a knob ..(apart from
> the additional work on the boarder node).

Agree with Uma. There is no need to introduce any complexity to the SPF 
algorithm by taking N as the remote peer.

Best regards,
Xiaohu

> 
> 
> 
> 
> _______________________________________________
> mpls mailing list
> [email protected]
> https://www.ietf.org/mailman/listinfo/mpls

_______________________________________________
spring mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/spring

Reply via email to