Re: [IPv6] Clarifications: Architectural comments on draft-ietf-6man-ipv6-over-wireless-

[Vasilenko Eduard <[e-mail]>]

The big ecosystem that was created in 6lo is not needed for the majority of use cases. Just draft-chakrabarti-nordmark-6man-efficient-nd is enough to fix the majority of ND problems (root causes are in draft-ietf-v6ops-nd-considerations). Statefulness is enough.

I fail to see how MLSN would help in the migration on the one link.

Yet, WiND has its use case for IoT. There are places where MLSN is needed (802.15.4). It is not WiFi or wired Ethernet.
Eduard
-----Original Message-----
From: Pascal Thubert (pthubert) [mailto:pthubert=[e-mail]] 
Sent: Wednesday, August 16, 2023 1:05 PM
To: Vasilenko Eduard <[e-mail]>
Cc: Brian E Carpenter <[e-mail]>; IETF IPv6 Mailing List <[e-mail]>
Subject: RE: [IPv6] Clarifications: Architectural comments on draft-ietf-6man-ipv6-over-wireless-


> MLSN is too heavy for WiFi. The draft that was predisessor of WiND (I heferenced it many times) is more then enough (because it is statefull). Nothing additional is needed for wired.

> Ops people would not have a choice if MLSN would be incorporated into the ND basement. They would have to deal with complexity. It is not a choice.

> I am just asking to step back to 2015 – that proposal was the best (just stateful ND, no MLSN). You were a co-author.

1) draft-chakrabarti-nordmark-6man-efficient-nd was based on RFC 6775. RFC 8505 the very much like RFC 6775, with improvements on link local operation (simpler now), use if TLLAO (a bugfix), and the TID (for movements, much needed in wireless, from MANET experience). 

2) There's no going back to RFC 6775, you're missing many years of 6lo history, but you can find the rationale summary in https://www.rfc-editor.org/rfc/rfc8505.html#appendix-B. 

3) We could revive/enhance draft-chakrabarti-nordmark-6man-efficient-nd if something in there is still std track missing, but I fail to see which if we consider the multicast and prefix registration drafts as part of the RFC 8505 ecosystem. What we really need is a framework doc that compiles how things work and which RFC defines what.

4) https://datatracker.ietf.org/doc/html/draft-chakrabarti-nordmark-6man-efficient-nd-07#section-5.1 requires a proxy for mixed mode. RFC 8929 defined that proxy and is part of the RFC 8505 ecosystem as well. RFC 8929 allows both routing proxy and bridging proxy; that's the choice that the ops will have. Routing Proxy is new, and probably the thing you're opposed to. Bridging Proxy is still translational bridging, the classic thing, with a data path exactly as it is today in an ESS.

4) Stateful ND *is* MLSN at the IP layer. By the definition of IP link in the draft. It does not mean that the L2 broadcast domain has to be split, just that the ops may decide to do so on their choosing. At L3, the AP has a wireless P2MP or a collection of P2P IP links, which ever model ops choose. Then a bridging proxy may be used to connect the wireless links to the wired. 

I think your opposition is mostly due to a partial view on the whole architecture, but we have the same bottom line.

All the best

Pascal