Re: [netext] Needs of traffic spec. on MAG?

["Youn-Hee Han" <[e-mail]>]

Hi all,

> Carlos Jesús Bernardos Cano [mailto:[e-mail]] wrote:
> >
> > On Thu, 2010-08-19 at 08:27 -0700, Laganier, Julien wrote:
> > > Tran Minh Trung wrote:
> > > >
> > > > I agree that the prefix p3, according to RFC 5213 , does not have to
> > > > be simultaneously assigned to MAG1.
> > >
> > > I think talking about prefix assigned to a specific MAG is misleading.
> > > The prefix is assigned to the mobile node thru the NETLMM fabric,
> > > irrespective of which MAG(s) the mobile node is attached too.
> > >
> > > When flow mobility is enabled, the prefixes assigned to the mobile
> > > node can be used thru more than one interfaces, attached to more than
> > > one MAG.
> > >
> > > >                                          However when we use logical
> > > > interface at the MN, the prefix p3 should be shareable to allow flow
> > > > mobility.  In addition, the logical interface can receive packets
> > > > sent to any of its sub-interfaces as described in
> > > > draft-melia-netext-logical-interface-support-01 (property #2).
> > > >
> > > > The question here is how to change RFC 5213 to support shared-prefix
> > > > model?
> >
> > We should be careful about the wording, "shared-prefix model" is
> > commonly referred to the assignment of the same prefix to multiple MNs,
> > whereas here we are talking about assigning the same prefix to multiple
> > interfaces of the same MN. It is not exactly the same thing, IMHO.
> 
> Fair enough.
> 
> How about "per-MN HNP set"?
> 
> --julien

IMHO, we don't need any more to care about the prefix model when designing
a network-based flow mobility. We already have the per-MN (more exactly 
per-interface) prefix model which is the basic principle of RFC 5213.

Even when we support our network-based flow mobility, we should keep 
the principle if any flow is not needed to be moved at the time of a new MN 
attachment. The new MAG will be assigned a new prefix according to the
principle.

If a flow is needed to be moved to the new MAG at the time of a new MN
attachment, 
the existing prefix which the flow uses should be assigned to the MAG.
This strategy has been already supported in the handover case of RFC 5213.
However, we did not mentioned "shared-prefix model" in order to support the
same 
prefix allocation at the case of handover when designing RFC 5213.

Of course, I know well the same prefix allocation can happen frequently 
if we support flow mobility in PMIPv6. For all that, we do not need say that

PMIPv6 should support two strategies of prefix allocation.

One more thing.

As others mentioned in ML, I also think LMA does not need to know the
logical 
interface configuration on each MN.
However, IMHO, if LMA can know it, the reason to keep the per-MN
(per-interface) 
prefix model naturally disappear. LMA can allocate just one set of prefixes 
to any interfaces of MNs, at any cases and at any  times, 
because the number of MN's interface is just one in the LMA's perspective. 
This strategy will make the flow mobility extremely easy.

Youn-Hee Han