Re: [netext] Update on flow mobility following discussion with ADs

[Julien Laganier <[e-mail]>]

Juan Carlos,

On Fri, Mar 4, 2011 at 11:59 AM, Zuniga, Juan Carlos
<[e-mail]> wrote:
> Julien,
>
> Capability discovery can be achieved by layer-2, 3 or even 7. The
> current charter restriction is specifically about no MN-ntwk PMIP
> signalling, and layer-2 signalling is outside the scope.

We assume that a L2 construct (logical interface) isolates the
unmodified IP stack from the underlying diversity of physical
interfaces. From this point of view it seems fair that the L2
signaling is in charge of signaling the capabilities of the MN (and
this has been done in 3GPP.) Of course the layer 2 signaling
definition is out-of-scope of this group, but as it's been done in
5213 where reasonable assumptions were made on what the L2 signaling
has to provide (attach triggers, MNID, and possibly handoff
indicator), for this flow mobility work item we're certainly in a
position to make the assumption that flow mobility requires capability
support indication and intent to do flow mobility vs. inter-access
handoff. Actually, we have no choice if we want a working system (see
point at the end of this note.)

The IP layer is unmodified and the charter forbids this group to
define a layer 3 MN-network signaling.

I don't know how a layer 7 would know anything about the capability of
the network stack to hide to the IP layer the underlying physical
interfaces...

> Hence, I think it is fair to design a PMIP flow mobility solution with
> the following assumptions:
>
> 1) MN capabilities are known,

by magic?

> 2) possible existence of layer-2 signalling to provide hints (HI=Flow
> Mobility),
> 3) and possible non-existence of layer-2 signalling to provide hints
> (HI=Unknown)

in 5213 HI=unknown leads the network to not provide inter-access
handoffs because it could break connectivity at a legacy host. I don'
t see how we can depart from that and thus if HI= unknown no flow
mobility should be offered to the host, only basic 5213 model where
each interface of host has its own PMIPv6 session. If you offer flow
mobility when the network does not reliably know the capability of the
host you risk breaking that host thus the protocol extension would
have no correctness...

--julien