All things related to multicast

["Marshall Eubanks" <>]

If anyone knows of any set-top box that speaks IGMPv3, I would 
love to hear about it. It will go into my case for  multicast presentations.

On Mon, 17 Oct 2005 10:51:34 -0400
 Chuck Anderson 
<>
 wrote:
> On Mon, Oct 17, 2005 at 11:52:46AM +0200, Thomas Kernen wrote:
> > It's great to hear that vendors are finally understanding SSM and 
> > IGMPv3
> 
> There are still some vendors out there who don't think it is a
> priority to implement IGMPv3 properly.  I've mentioned one of them on
> here before.  Does anyone have any tips for convincing vendors of the
> importance of SSM and IGMPv3 so that proper implementations will be
> forthcoming before the decade is over?

Here  is what I sent Apple on this topics (heavily revised and updated). Feel 
free to use as
appropriate :

Inter-domain multicast deployment for broadcasting (except on the I2) is 
basically stalled,
as is deployment of globally routable multicast on the Internet
as a whole. See http://www.multicasttech.com/status/ for further details. 
Deployment is stalled
because of the complexity of inter-domain any source multicast, and also 
because of worries about
security vulnerabilities in inter-domain multicast. A new multicast service 
model, Source-Specific
Multicast (SSM) has been developed to address these concerns, and is 
especially appropriate in an
Internet video broadcasting  environment.

SSM deployment requires nothing from the broadcaster except the  selection of 
an appropriate
Multicast address, and little from the ISP. However, it does require support 
from both the
player and OS of IGMPv3 on the receiver side, and as a practical matter needs 
to be supported by
IGMP snooping switches. 

A new protocol, Automatic Multicast Tunneling (AMT), should allow for an 
automatic multicast
rollover to unicast. AMT is specifically tailored to provide SSM support for 
end users  who do not
have access to inter-domain native multicast.

SSM plus AMT offers a means of breaking the log-jam and providing usable 
service. 
Basically, this requires  IGMPv3 support in the kernel, SSM support in the 
player, IGMPv3 support in
the switch, and use of the AMT shim. Three ISP's have indicated that they 
would step up and provide
the multicast fan-out for AMT.

As multicasting increasingly becomes the standard for IP video distribution, 
video providers will
need SSM, both for protection of their content within the walled garden, and 
to allow for the
provision of video content from outside the walled garden. My analysis of the 
statistical
distribution of video content selection indicates that even the United States 
should be able to
support 50,000 to 100,000 channels of video content (plus, of course, a large 
amount of VOD, some of
which will need to be  pre-cached), and the global numbers will be even 
larger. Just as Amazon.com
is estimated to make some 40% of its sales from the several million titles 
that it carries that do
not sell well enough to be stocked in bookstores, my models predict that in a 
100,000 channel video
distribution universe as much as 50% of the audience (and thus the revenues) 
are likely to come
accrue to the ~ 99,000 channels that are not supported by the walled gardens. 
The only realistic way
I see to support this will be through the use of SSM, IGMPv3 and (at least in 
the near term) AMT. 

Regards
Marshall Eubanks