> Hi Behcet,
>
> This is a known feature of a route-over LoWPAN, we have been
> discussing that already a lot the last two years, and have converged on
> the current model. This model actually is not a multi-link subnet as
> such.
[behcet] I think it is multi-link subnet but a different type.
> Instead we have a single non-transitive (radio) link, where exact-match
>routing is used
>
>
[behcet] What do you mean by exact-match routing? Why does it make prefix
sharing OK?
> to overcome the fact that not all interfaces can hear each other at all
> times. We are referencing the Autoconf addressing model which
> makes similar assumptions about radio links.
[behcet] I read that document. It says IPv6 address configured on this
interface
should be unique within the routing domain. This is also valid with non-shared
prefixes.
>
> On Aug 4, 2010, at 10:05 PM, Behcet Sarikaya wrote:
>
> > Hi Zach,
> > Prefix sharing assumption stated in Sec. 1.3 I think creates multi-link
>subnet
>
> > issues described in RFC 4903 (not referenced in your ID) especially if
> > route-over is used. Similar concerns were raised in IETF 78 session.
>
> This was from a person in the audience not familiar with the 6LoWPAN WG if I
>remember right.
>
> > Of course if mesh-under is used then there is no problem but then there is
> > probably not much case for the Roll protocol.
> >
> > It seems that we have this dilemma of choosing between the two evils.
> >
> > I suggest that we avoid this problem by assigning a unique prefix for each
>IPv6
>
> > host and then freely use the roll protocol for route-over and use prefix
>sharing
>
> > only if mesh-under is used. It is clear that you can not have it both ways.
>
> The suggestion you make below simply does not work with 6LoWPAN > header
>compression, nor does it make sense with mobility.
>
[behcet] I guess you mean if different prefixes are used. Also your concept of
mobility is different, you don't consider moving to a different network which
is
probably OK for 6LOWPAN nodes.
> You may do that with a formal IPv6 network of course (see the Autoconf WG).
>
Yes.
Regards,
Behcet
_______________________________________________
6lowpan mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/6lowpan
