Hi Zhiqiang Li,
Lots of thanks for your email.

Please see some comments inline below. Hopefully, they will be useful.

Regards,
Sasha

From: 李志强 <[email protected]>
Sent: Tuesday, July 28, 2026 5:09 PM
To: Alexander Vainshtein <[email protected]>
Cc: [email protected]
Subject: [EXTERNAL] Re: [bess] A short comment on 
draft-li-bess-evpn-ead-multipath

Hi Alexander, Thanks for your precise review and sharp observation on RFC 7432 
specifications. First, we fully agree with your core conclusion: when Type 1 
RDs are used (mandatory in spirit for per-ES, recommended for per-EVI), EAD 
routes from different PE/EVI form distinct NLRIs and cannot be supp
[Image removed by 
sender.]<https://report.mimecastcybergraph.com/?magiclink=https%3A%2F%2Fapi.services.mimecast.com%2Foauth2%2Fauthorize%3Fresponse_type%3Dcode%26client_id%3Do20nRkVXf7VUVnANkXhoOwGytEwGN0YAlyeDJn7oBTGNl2kN%26state%3DeyJhbGciOiJSU0EtT0FFUC0yNTYiLCJlbmMiOiJBMjU2R0NNIn0.iir_1BmTm5cdZeItrTwxRLccExJ0vcsRqX8a1ghKY2EahmNaru-hKmQLm4OEC0gHeijTF2n6mnppyE3GxZ9rNvO1SLGe09kMki-CcA8jxX9jRplw8DTbCFr_FzrlQKnyrZ7IEVwDpIP74dnPn3vt9gZOrubdxDK6RKaJmLf2sqkXuDd118yWWkoq4xkdiVuGfjQWVJyczaNTKFL5twBOK3r6f9cKkus3EnR8MZ4ezoum8hy6Cevj-9hPvfZmcFyd2WaS7UfgVsP1cQo185mnWZx7U8U7ZbI_XC5PYHE4B8MZUIuF5oGXstEkdo5_e8Dag9bqDD8A6rCWSDYG8J7QUQ.1qQwNUdEV15270L5.KPdujUDqWvuAv8r3aFlJYlewFEhRMrVcS-u-3XAEDgZET2oMC0YNXu1ksh0WSkGmAkBBf_K_37RKtz-qc9nWLZ25Gyz0Gu-LzQbJ0OyY2gyTaO-2ypNc31lGIks5l3_s-xh_062HEvoF2iPBko2v8pIDInFEw2Hnbz0T81ffVMn3Z3mY_43SrYYPdlM8QBCHUebwCvvufLkZYIlbEysmV4bCwIZ7d6ELFSZ8vBBa4U_ODWRE1WlaHZJSUVckcEaZsRz_p__b1elehJy7YJWuU2TzA-PJZsqGYfkcbZwGs0UuAxnEKlFX8WuEv9u8SiB2w8n2NnspwCLUi_klX61HvLpW8vgMyf6ylMR4dUfBmvVnLoLsbOqtl5CcZ2aPdP8sKVtvOs3hpmQkpztfPB6hbH0vKBm0X1B83MlewX3uk7tfLQlXMtiPUeMI3Fsn0a9eMl0YJYO5w5J_Og6vQkUtUfRxZUTcAVqenS1HTL85UgFvDlGHksDvT_TQkxYbVWU8sT6dzLO9w-qq5gO-qJ57Rok74DSp1ttdVWsV0MCsb_UoNuu4VwQfP5QHThzY7tYi5_yKlQva_FS4PMEWvADpV2gKOJ4Gj4ERNfiBVWHy2K_NKT2IJzVa3mvihzAeaD8RjI4zEpk54OuqwoPSQpvc6vsoUf_O_opE4wfvHBd51c_2jmTkCuKOd7bwI_-co4Im1eiEN2fPPt5cRXmVGvXx5HGxmxVuZmGWMObAX8Hc1nClz46VIy00ro481jy6aOAOjpFcPt--Xxf9rxaTFNB_QNMp73_G7_MrGP8kF8exHyrR3ytiTqId9rjbX0M16K0moLQfFDO_9cYfXtZGodzzpYz70pVz86AtaITusAI.tK7q_BOAJlK48tg-KlKKcA%26redirect_uri%3Dhttps%3A%2F%2Freport.mimecastcybergraph.com%2Fcallback>
CGBANNERINDICATOR

Hi  Alexander,



Thanks for your precise review and sharp observation on RFC 7432 specifications.



First, we fully agree with your core conclusion: when Type 1 RDs are used 
(mandatory in spirit for per-ES, recommended for per-EVI), EAD routes from 
different PE/EVI form distinct NLRIs and cannot be suppressed by standard BGP 
Best Path selection. Your reading of RFC 7432 is completely correct.



The confusion comes from ambiguous wording in the current draft.



We have no intention to "twist the BGP Path Selection algorithm" itself. The 
draft targets two separate practical scenarios:



1. Shared-RD deployments (more common for per-EVI cases, where Type 1 RD is 
only RECOMMENDED, not mandatory): With identical RD across PEs, EAD routes of 
the same ESI/Ethernet Tag form identical NLRIs, and default BGP 
single-best-path behavior suppresses redundant paths. The draft mandates BGP 
multipath for EAD routes to retain all valid paths — it applies an existing BGP 
capability to EVPN EAD, not rewrites the path selection logic. [[Sasha]] 
Quoting from RFC 2119 (the relevant text is highlighted):

SHOULD   This word, or the adjective "RECOMMENDED", mean that there may exist 
valid reasons in particular circumstances to ignore a particular item, but the 
full implications must be understood and carefully weighed before choosing a 
different course

From my POV the operator that decides to ignore the RECOMMENDED usage of the 
MAC-VRF RD deserves all the consequences of this decision.



2. Standard unique-Type-1-RD deployments: The draft works at the EVPN 
application layer, not BGP layer. It requires the ingress PE to logically 
aggregate EAD routes of the same ESI (despite different NLRIs) and pre-install 
all valid paths into the forwarding plane for fast failover, without touching 
BGP route selection at all.



For the case of multiple per-ES EAD routes from one PE (due to RT overflow): 
these are treated as a single logical path on the receiver side and do not 
count as independent multipath entries. [[Sasha]] This looks to me as a local 
issue within a specific implementation in the ingress PE and does not require 
any standardization.



We will revise the next version of the draft to clearly separate BGP 
control-plane behavior and EVPN forwarding-plane processing, to eliminate this 
ambiguity.



Thanks again for catching this critical wording issue. Further comments are 
very welcome.


Best regards

Disclaimer

This e-mail together with any attachments may contain information of Ribbon 
Communications Inc. and its Affiliates that is confidential and/or proprietary 
for the sole use of the intended recipient. Any review, disclosure, reliance or 
distribution by others or forwarding without express permission is strictly 
prohibited. If you are not the intended recipient, please notify the sender 
immediately and then delete all copies, including any attachments.
_______________________________________________
BESS mailing list -- [email protected]
To unsubscribe send an email to [email protected]

Reply via email to