Re: [dnssd] multicast over wireless links

[Tom Pusateri <[e-mail]>]

> On Jul 24, 2014, at 1:25 PM, Ted Lemon <[e-mail]> wrote:
> 
> Asking IEEE to fix WiFi is not really a good approach to addressing the problem of inefficient multicast for existing deployments, and this work is intended to address existing deployments.
> 
> Therefore, I think that if we want to address this use case, the right way to do it is to come up with a mechanism for doing multicast distribution of DNSSD updates on broadcast-friendly networks, and not to just make this the default operational mode for DNSSD.
> 

Having spent a good portion of my career on IP multicast (from Token Ring to IP Multicast VPNs), it baffles me how a radio can have such a difficult time doing IP multicast. The 802.11 link layer clearly needs improvement in this area. Access points regularly fall over dead when bridging an ethernet wired connection to wireless when there is a high data rate multicast stream (actually 4Mbps isn't that high). This is just unacceptable.

With most radios, multicast is the most efficient form of transmission. But in 802.11, every receiver in range currently receives N copies of the multicast to their antenna before throwing all of them but one away. Surely, this can be improved. Multi-unicast over radio broadcast has to be the least efficient mechanism to achieve this.

With so many wireless devices now, streaming live sports events over multicast should be ultra efficient but this isn't the case on 802.11.

There is work to be done in the IEEE and I think that the PIM and MBONED WGs would be pleased to see us open a dialog with IEEE to improve this deficiency.

Optimizing for a single link layer is a rathole that we should avoid if you can take a step back and look at things from a long term perspective. Coming up with an optimization or work-around for a single link layer is fine as a short term solution.

Tom