RE: on-link and off-link

Indeed

Which I read as "classical ND is not suited to these configurations" whereas RFC 8505 works just fine.

PVLANs are an illustration that the Woodstock thinking is not the safest thing and people may want more privacy and protection / isolation. This is true in SP / cable with PVLAN isolating modems, this is also true in an enterprise where zerotrust and SAVI are getting mind share. 

Keep safe,

Pascal

> -----Original Message-----
> From: Vasilenko Eduard <vasilenko.eduard@huawei.com>
> Sent: lundi 12 juillet 2021 18:15
> To: Pascal Thubert (pthubert) <pthubert@cisco.com>; Philip Homburg <pch-
> ipv6-ietf-7@u-1.phicoh.com>; ipv6@ietf.org
> Cc: Alexandre Petrescu <alexandre.petrescu@gmail.com>
> Subject: RE: on-link and off-link
> 
> By the way,
> It blocks not just native P2MP technology, It does break any filtering,
> like "PVLAN" or "WiFi isolation", Because if it is not possible to send a
> multicast to other hosts to check DAD even for LLA.
> Ed/
> -----Original Message-----
> From: Pascal Thubert (pthubert) [mailto:pthubert@cisco.com]
> Sent: Monday, July 12, 2021 7:04 PM
> To: Vasilenko Eduard <vasilenko.eduard@huawei.com>; Philip Homburg <pch-
> ipv6-ietf-7@u-1.phicoh.com>; ipv6@ietf.org
> Cc: Alexandre Petrescu <alexandre.petrescu@gmail.com>
> Subject: RE: on-link and off-link
> 
> Exactly; the model at the IP layer has to be multilink subnet.
> Your link local gets you to a router but not to all in the subnet.
> 
> With IPv6 all is abstracted with multicast operations and link local
> addresses.
> So what IP considers a link and who belongs to a subnet (is constrained by
> but) does not need to match what the L2 sees, e.g., the L2 links (e.g., an
> ethernet wire or a radio range) and the L2 broadcast domains (e.g., a
> 802.1 bridged domain or a Wi-Fi ESS).
> IPv4 gave the wrong habit of thinking they have to be congruent. They
> don't.
> 
> Pascal
> 
> > -----Original Message-----
> > From: ipv6 <ipv6-bounces@ietf.org> On Behalf Of Vasilenko Eduard
> > Sent: lundi 12 juillet 2021 17:35
> > To: Philip Homburg <pch-ipv6-ietf-7@u-1.phicoh.com>; ipv6@ietf.org
> > Cc: Alexandre Petrescu <alexandre.petrescu@gmail.com>
> > Subject: RE: on-link and off-link
> >
> > NBMA or P2MP are not supported by IPv6 anyway because it is not
> > possible to check LLA uniqueness - multicast does not work.
> > LLA proxy is needed on the hub site.
> > Ed/
> > -----Original Message-----
> > From: ipv6 [mailto:ipv6-bounces@ietf.org] On Behalf Of Philip Homburg
> > Sent: Monday, July 12, 2021 6:20 PM
> > To: ipv6@ietf.org
> > Cc: Alexandre Petrescu <alexandre.petrescu@gmail.com>
> > Subject: Re: on-link and off-link
> >
> > > I think one question
> > > might be: do we agree that each IP address is assigned on an
> > > interface, and thus is on-link on a particular link?
> >
> > This is not correct. On a NBMA link it is possible that all GUA
> > addresses are not on-link.
> >
> > The conceptual model of on-link is that a host can use neighbor
> > discovery to find an onlink neighbor. If the neighbor is not on-link
> > then traffic has to go (at least initially) through a default router.
> >
> > --------------------------------------------------------------------
> > IETF IPv6 working group mailing list
> > ipv6@ietf.org
> > Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
> > --------------------------------------------------------------------
> >
> > --------------------------------------------------------------------
> > IETF IPv6 working group mailing list
> > ipv6@ietf.org
> > Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
> > --------------------------------------------------------------------