RE: L=0 [was draft-pioxfolks-6man-pio-exclusive-bit-02.txt]

["Templin, Fred L" <[e-mail]>]

Hi, I have been making statements about NBMA links for a long time now
both on the lists and in drafts. AERO is based on the NBMA link model:

https://datatracker.ietf.org/doc/draft-templin-aerolink/

as was ISATAP [RFC5214]. But, please note well that AERO *is not*
the offspring of ISATAP, although they are distant cousins.

Thanks - Fred

From: ipv6 [mailto:[e-mail]] On Behalf Of Mark Smith
Sent: Tuesday, January 30, 2018 1:19 PM
To: Pascal Thubert (pthubert) <[e-mail]>
Cc: Michael Richardson <[e-mail]>; IPv6 List <[e-mail]>
Subject: Re: L=0 [was draft-pioxfolks-6man-pio-exclusive-bit-02.txt]



On 30 Jan. 2018 17:33, "Pascal Thubert (pthubert)" <[e-mail]<mailto:[e-mail]>> wrote:

Hello Brian:

Introducing NBMA topologies implies routing.

The trouble is this isn't an NBMA link layer, although it might seem like it.

The limitation of NBMA is that they don't support multicasting to some or all of the attached nodes. IPv6 support for NBMA goes to the effort of restoring that capability by emulating it.

On the links described here, that multicast restriction is seen to be a security advantage, and if it was restored via emulation, it is a security policy violation.

NBMA links do not intentionally prevent attached nodes directly sending packets sending to each other. That is a limitation of NBMA links that is to be overcome by either hair pinning via the default router (via L=0), or by setting up a direct virtual circuit or path of some sort between the nodes and ICMP redirects.

In other words, on an NBMA link, these constraints are natural ones to work around. On a Private VLAN, TR-101 etc. link, they're intentionally imposed ones that must not be worked around because they exist for security reasons, not because the link doesn't naturally support those capabilities.

Regards,
Mark.


RPL takes some of its origins in an effort to build ND for a general NBMA network(https://datatracker.ietf.org/doc/draft-thubert-nina/). Initially DIOs were RA-DIO, and DAO was an option of an NA. 6Lo ND and RPL were one.

The game of areas caused them to split and slightly diverge, till at some point ROLL decided for a separate ICMP. But we have a working solution for that problem.
All the best;

Pascal

Le 29 janv. 2018 à 20:12, Brian E Carpenter <[e-mail]<mailto:[e-mail]>> a écrit :
On 30/01/2018 03:40, Michael Richardson wrote:

<big snip>

...then I think that we
should accept that the "LAN" is actually no longer broadcast, and is a
star/hub-spoke, and we should architect on in that way.

To be future proof, you may have to stop using the phrase "LAN" at all,
and accept that the L2 topology may be a local mesh (of which hub & spoke
is a simple case). That means rethinking everything, not just details
like the L bit.

  Brian

--------------------------------------------------------------------
IETF IPv6 working group mailing list
[e-mail]<mailto:[e-mail]>
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------

--------------------------------------------------------------------
IETF IPv6 working group mailing list
[e-mail]<mailto:[e-mail]>
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------