RE: Joining to SSM sources that are IPv6 mobile nodes

["Dave Thaler" <dthaler@windows.microsoft.com>]

Rolland Vida (Rolland.Vida@lip6.fr) writes:
[...]
> > Either option (1) would never be used for any multicast traffic, or
else
> > the host would
> > have to know whether a receiver might want to use an include-mode
list
> > with MLDv2 and tunnel to the home agent if so.
> 
> You can make the difference between an SSM and an SM connection by the
> group's address. There's a specific SSM address range, and it is
supposed
> that you do not do PIM-SM  in the SSM range.

It's not that simple.  Moving on to quote draft-ietf-ssm-overview-01.txt
section 1:
> Source-Filtered Multicast (SFM) : This is a variant of the multicast
> service model defined in RFC 1112. A source transmits IP datagrams to
> a host group address in the range of 224.0.0.0 to 239.255.255.255.
[for now ignore the fact that the above sentence needs to be reworded
to apply to IPv6 as well... hint to SSM WG...]
> However, each "upper layer protocol module" can now request data sent
> to a host group G by only a specific set of sources, or can request
> data sent to host group G from all BUT a specific set of sources.
> Such support for source filtering is provided by version 3 of the
> Internet Group Management Protocol (or IGMPv3) [IGMPv3] for IPv4, and
> version 2 of the Multicast Listener Discovery (or MLD) protocol for
> IPv6 [MLDv2]. 

The point here is that the problem is not limited to the SSM
address range for IPv6.  It really applies to include-mode joins
for any IPv6 multicast address.  Today the multicast model does not
require sources to know whether receivers might do this.  As a result,
without changing the model, you cannot use the address range to make any
decision here.

Based on the four people who have responded so far (Dupont, Haberman,
Luby, and Vida) it sounds like the consensus will be to always tunnel 
via the home agent if you need network-layer mobility.  If this
consensus holds, then I think the Mobile IPv6 spec should say that any
time the app sends multicast using the home address, the stack
SHOULD/MUST tunnel it via the home agent, and also say that the Home
Address Option SHOULD NOT/MUST NOT be used in multicast packets.  The
application can certainly choose whether to source from the Home Address
(in which case receivers will continue to get data if it moves), or from
the Care Of Address (where mobility must be handled by the application
as Mike Luby suggested).  There would then be no additional impact on
APIs, or the MAGMA or SSM WGs.

-Dave