Hi Rajesh, The allocation of End.DT/DX SIDs are generally determined by the configuration of allocation mode on egress PE, and are stable after allocation, i.e., without affection by the BGP next-hop resolve processing. And, the forwarding behavior of END.DT/DX is also different, so that they must be instantiated separately. However, the difference between REPLACE and REPLACEB6 is negligible, they are both SID swap behavior. Let me put it another way, would you plan to define a new endpoint behavior for each case when the resolved outer tunnel to BGP next-hop is SRv6 policy, or GRE, or MPLS, ... etc ?
Regards, PSF ------------------原始邮件------------------ 发件人：RajeshM 收件人：彭少富10053815;draft-salih-spring-srv6-inter-domain-s...@ietf.org; 抄送人：spring@ietf.org; 日 期 ：2021年11月15日 11:27 主 题 ：Re: [spring] Questions about REPLACEB6 of draft-salih-spring-srv6-inter-domain-sids Hi Peng, Thanks for the draft acknowledgement. Regarding your question about REPLACE/ REPLACEB6 Why we need DT4/DT6/DT46/ DX6/DX4 sids, they are also " local behavior of the BGP egress speaker". End.DX6 - Decapsulation and IPv6 Cross-Connect. End.DX4: Decapsulation and IPv4 Cross-Connect End.DT6: Decapsulation and Specific IPv6 Table Lookup End.DT4: Decapsulation and Specific IPv4 Table Lookup End.DT46: Decapsulation and Specific IP Table Lookup By having separate SID at egress for each Endpoint Behavior, it will be well defined. On similar lines we have well defined Endpoint Behavior REPLACE and REPLACEB6 sids. Hope this clarifies. Thanks Rajesh Juniper Business Use Only -----Original Message----- From: peng.sha...@zte.com.cn <peng.sha...@zte.com.cn> Sent: Tuesday, November 9, 2021 9:02 AM To: draft-salih-spring-srv6-inter-domain-s...@ietf.org Cc: spring@ietf.org Subject: Questions about REPLACEB6 of draft-salih-spring-srv6-inter-domain-sids [External Email. Be cautious of content] Hi authors, Thanks for bringing SRv6 SID swaping idea in the distributed scenario. I think there is valid case when the target prefix originated from remote domain is invisible or unreachable to the intermediate nodes of the local domain, and the intermediate nodes of the local domain have routes to the REPLACE SID allocated by the border node of the local domain. However, the question as I raised at the meeting, REPLACEB6 may be not necessary, because it is the local behavior of the BGP speaker to swap a single SID or a segment list. It may be not appropriate to advertise prefix with REPLACE behavior to the upstream neighbor at time-1, and re-advertise the same prefix with REPLACEB6 behavior at time-2 just because its forwarding information is changed from a single outgoing SID to a SID list. Otherwise, it will cause excessive advertisement overhead in the network. Just like MPLS BGP-LU, the difference of a single outgoing label or an outgoing label stack at the local node is not aware by upstream node. There may be some reasons for re-advertisement, such as the need to announce a new metric, but these reasons do not support the need for REPLACEB6. Regards, PSF _______________________________________________ spring mailing list spring@ietf.org https://www.ietf.org/mailman/listinfo/spring _______________________________________________ spring mailing list spring@ietf.org https://www.ietf.org/mailman/listinfo/spring
