Re: [Roll] [6lo] New Version Notification for draft-thubert-6lo-unicast-lookup-01.txt

["Pascal Thubert (pthubert)" <[e-mail]>]

Hello Dario





Le 6 oct. 2021 à 21:50, Dario Tedeschi <[e-mail]> a écrit :

 Hi Pascal,

I think the 2nd and 4th cases can be merged, by allowing a root node to automatically propagate the following multicast messages, using MPL:

  *   All scope 3 (Realm-Local) multicast messages it either originates or receives on an MPL interface.
  *   All Unicast-Prefix-based IPv6 Multicast Addresses (RFC 3306) higher than scope 3, where the network prefix (given in the mulitcast destination address), matches the prefix of the DODAG ID (the RPL network's subnet).


WFM, conforms https://www.rfc-editor.org/rfc/rfc7346.html#section-5

Automatic forwarding of the 2nd address type could be optional and administratively configured.

If nodes are interested in other multicasts higher than scope 3, they must explicitly inform the root by sending DAO messages with appropriate Target Options.

---------

The 3rd case, I think, needs its own mop code (i.e. "Non-storing mode with source-routed multicast").


Yes, but we have to look at backwards compatibility / brown field and allow though not recommend to use mop 1; same issue as already discussed with mop 3 in the dread

For the 1st and 3rd cases, how do you envision multicasts propagating up the DODAG (towards the root)? Would a node simply L2 unicast to its preferred parent?


The 6 LN does. It know about RPL and registers the multicast address. Upon. The first registration The 6LR sends a unicast DAO to the Root as we do for RUL. This time though the address in the target is multicast. The root makes a copy per 6LR that shows as transit and sends along the unicast SR path to the 6 LR as it would for a RUL unicast. Less efficient that mop 3 that builds a real multicast tree, but backward compatible for for nodes on path…

Works?

Pascal



Regards
Dario


On 10/6/21 6:00 AM, Pascal Thubert (pthubert) wrote:
OK Dario, so we’d have 4 optional combinations:

unicast = mop 1 multicast = mop 3 (what the draft does today)
unicast = mop 1 multicast = MPL (that I believe the draft allows today but should clarify); in that mode, not message to the Root, the root floods all multicast messages with the idea that there’s always a listener somewhere
unicast = mop 1 multicast = mop 1 (to be added) in that mode the 6LR sends a DAO to the root for a multicast target, and the Root sends n messages that are unicast source routed to the n 6LR that have listeners, only the last address in the SRH is multicast
unicast = mop 1 multicast = MPL (that I believe the draft allows today but should clarify); in that mode, in that mode the 6LR sends a DAO to the root for a multicast target, and the root uses MPL only when there’s known listeners

Do we describe them all? Should we consume RPL MOPs?

I suggested that AODV RPL reuses MOP 4 to leave room…

Pascal

From: Dario Tedeschi <[e-mail]><mailto:[e-mail]>
Sent: mercredi 6 octobre 2021 0:05
To: Pascal Thubert (pthubert) <[e-mail]><mailto:[e-mail]>
Cc: [e-mail]<mailto:[e-mail]>
Subject: Re: [6lo] New Version Notification for draft-thubert-6lo-unicast-lookup-01.txt

Hi Pascal

See my comment  below.
On 10/5/21 12:40 PM, Pascal Thubert (pthubert) wrote:
Hello Dario

Please see below;


Le 5 oct. 2021 à 20:15, Dario Tedeschi <[e-mail]><mailto:[e-mail]> a écrit :
 Hi Pascal,

Thank you for new draft. However I do have some comments/questions.

What benefit does the ‘M’ bit provide over simply detecting a multicast address in the Target Address field?

The IPv6 multicast address type is clearly defined in RFC 4291 (section 2.4)<https://datatracker.ietf.org/doc/html/rfc4291#section-2.4>, and the detection of such an address is trivial. Most (if not all) Stacks have a simple function/macro to do that job and many existing protocols already use this mechanism to distinguish between unicast and multicast addresses.  It seems to me that a special bit to indicate multicast registration would be redundant and require handling for 4 different cases, 2 of which would be errors:


  *   M = 1, Target = multicast addr
  *   M = 1, Target= unicast addr  — ERROR
  *   M = 0, Target = multicast addr — ERROR
  *   M = 0, Target= unicast addr



True enough. Dario.

I’ve been pondering that too. On the one hand it seems cleaner to announce the service that the 6LN expects. Otoh as you point out it can be inferred from the address.

Another way of seeing this is that the error cases that you indicate can be detected if we have the bit otherwise they can’t.
[DT] I take your point about detecting the errors, assuming an implementation could do something useful with that knowledge, other than just discarding the message.



Then there’s anycast which is missing from both RPL and ND , which cannot be distinguished by the look of the address and thus requires a bit.
[DT] As for the anycast address, I suppose the question to ask is what would a router do differently knowing such information? I suspect we would have to define some new behavior along with the new bit.



Then there’s possibly the need of an IPv4 AF. All in all I tended to favor having the bit but that’s really not a strong position, happy to be convinced otherwise.
[DT] I presume you are talking of "IPv4-Compatible" and "IPv4-Mapped" IPv6 addresses. If my presumption is correct, aren't these still easily identifiable through their unique prefixes (::/96 and ::ffff/96, respectively)?



What do others think?
[DT] I have no strong opinion. The M bit just seemed redundant.





I also wonder about the requirement for non-storing RPL networks to propagate multicast membership up the DODAG. My understanding is that non-storing networks typically use MPL (RFC 7731)<https://www.rfc-editor.org/rfc/rfc7731.html>  which does not need multicast memberships to be propagated throughout the DODAG. It uses a flooding mechanism to forward multicast datagrams, and does not unicast at L2. Could the new document accommodate non-storing networks using MPL?

Sure;

Bottom line here is that for MPL all the multicast packets of interest for the LLN are flooded throughout so I suspect that there is no need for the 6LR to signal to the root.
[DT] Yes, that's my understanding as well.




If that’s the case then there’s nothing to standardize.  All I need to clarify is that the RPL behavior in the spec is the one expected in a RPL domain that supports mop 3 otherwise what is done is out of scope for this doc.


 Do you see it otherwise?
[DT] I agree that only RPL mode 3 needs to be defined and other modes are left out of scope.




I mean should the 6LR signal unicast to the root like for unicast traffic when serving a RPL unaware leaf?
[DT] That certainly could be an optimization for non-storing mode so that a border-router might know what multicast groups to forward from outside the network. Unfortunately though there is no MOP that is "Non-storing with multicast", although one could argue semantics and simply use MOP 1.

[DT] If we were to opt for such behavior, 6LR nodes could simply add RPL Target options to their DAO's, for the multicast groups they were interested in (including those requested by leaf nodes).



 If so wouldn’t it be expected that the Root makes n unicast to all 6LRs that have listeners?
[DT] I'm not sure that would make sense when MPL is being used, but it makes for an interesting alternative to MPL.




Should we describe that mode as well?
[DT] As an alternative to MPL? Sure.




Pascal

Regards
Dario



On Sep 27, 2021, at 6:32 AM, Pascal Thubert (pthubert) <pthubert=[e-mail]<mailto:pthubert=[e-mail]>> wrote:

Dear all:

This draft is a continuation of our work on RFC 8505, 8928, and 8929.

Comments welcome!

Pascal

-----Original Message-----
From: [e-mail]<mailto:[e-mail]> <[e-mail]<mailto:[e-mail]>>
Sent: lundi 27 septembre 2021 15:29
To: Eric Levy- Abegnoli (elevyabe) <[e-mail]<mailto:[e-mail]>>; Pascal Thubert (pthubert) <[e-mail]<mailto:[e-mail]>>
Subject: New Version Notification for draft-thubert-6lo-unicast-lookup-01.txt


A new version of I-D, draft-thubert-6lo-unicast-lookup-01.txt
has been successfully submitted by Pascal Thubert and posted to the IETF repository.

Name: draft-thubert-6lo-unicast-lookup
Revision: 01
Title: IPv6 Neighbor Discovery Unicast Lookup
Document date: 2021-09-27
Group: Individual Submission
Pages: 15
URL:            https://www.ietf.org/archive/id/draft-thubert-6lo-unicast-lookup-01.txt
Status:         https://datatracker.ietf.org/doc/draft-thubert-6lo-unicast-lookup/
Html:           https://www.ietf.org/archive/id/draft-thubert-6lo-unicast-lookup-01.html
Htmlized:       https://datatracker.ietf.org/doc/html/draft-thubert-6lo-unicast-lookup
Diff:           https://www.ietf.org/rfcdiff?url2=draft-thubert-6lo-unicast-lookup-01

Abstract:
  This document updates RFC 8505 in order to enable unicast address
  lookup from a 6LoWPAN Border Router acting as an Address Registrar.




The IETF Secretariat


_______________________________________________
6lo mailing list
[e-mail]<mailto:[e-mail]>
https://www.ietf.org/mailman/listinfo/6lo