On Oct 7, 2015, at 3:57 PM, Pushpasis Sarkar <psar...@juniper.net<mailto:psar...@juniper.net>> wrote:
Hi Robert, From: <rras...@gmail.com<mailto:rras...@gmail.com>> on behalf of Robert Raszuk Date: Wednesday, October 7, 2015 at 4:50 PM To: Pushpasis Sarkar Cc: "Stefano Previdi (sprevidi)", Hannes Gredler, "spring@ietf.org<mailto:spring@ietf.org>", Isis-wg, "Clarence Filsfils (cfilsfil)" Subject: Re: [spring] [Isis-wg] Handling same SID mapped to different prefixes and vice versa cases Hi Pushpasis , [Pushpasis] No. I seem to be repeating myself. But your suggestion of recursing will result in tunnelling traffic destined for all prefixes originated (not only loopback addresses) though MPLS. Operator may not like/want this. So associating the sid-index with the loopback address with a specific address provides flexibility in picking which prefixes should be tunneled and which ones need not. I do not quite agree with your above statement. Clearly you are thinking in terms how traditional MPLS router with perhaps specific implementation resolves paths and building forwarding tables in MPLS domain. Let's observe that Segment Routing imposition may but not necessarily must use the same imposition algorithm/mechanism. I can easily see a case at ingress where only carefully selected by operator subset of prefixes originated by exit X will be MPLS segment routed in the given domain. Rest may be native switched or IP tunneled/encapsulated. [Pushpasis] That is exactly what I was saying that such usecase will not be possible if we follow Stefano’s suggestion. As per his suggestion (at least that is what I understood so far)… The exit X will associate node-sid with only one loopback prefix.. And then ingress should automatically use the single node-sid for tunneling traffic for all prefixes originated by exit X.. this is not what I suggested. Again, sorry to repeat myself: You decide on a per-prefix base, which node-sid of the originator you want to use (if any). This gives you all the flexibility to: . group prefixes on different node-sids or . use the same node-sid or . use the prefix-prefix-sid whatever combination you need. s. And I am trying to say that will not provide the operator to pick a subset of prefixes.. If the ingress uses tunelling only for prefixes with which a index is associated, then the operators will have that choice.. As examples that can be accomplished by controller driven ingress map or use multiple next hops (the case where you propagate by BGP). [Pushpasis] I did not get this usecase much. But I think we maybe saying the same thing. Thanks -Pushpasis Cheers, R.
_______________________________________________ spring mailing list spring@ietf.org https://www.ietf.org/mailman/listinfo/spring
