RE: Open issues list? [Re: New (-02) version of IPv6 CPE Router draft is available for review]

[Mikael Abrahamsson <swmike@swm.pp.se>]

On Wed, 30 Jul 2008, Francois-Xavier Le Bail wrote:

> Significant bandwidth can be wasted by flooding.
I would just like to highlight a scenario that everybody should be aware 
of, sorry if this is obvious to everybody here, I know not enough people 
in other places are aware of it.
With most ATA/VOIP boxes, the uplink interface is 10/half. A multicast 
stream SDTV might be 6-8 megabit/s, so when channel zapping, there might 
be a short while (100ms) where two streams are being sent out on the local 
segment. If the L2 switch connecting the devices doesn't do IGMP/MLD 
snooping, the ATA box port will be take more than 10 meg of potential 
traffic for the duration of the burst, which hurts voice quality (since 
this will most likely cause packet loss). Since most cheaper L2 devices 
also do not have per-port buffers, but instead have shared buffers, it 
means that this problem might hurt traffic on other ports as well, 
incurring packet loss there also.
So, what I like about IPv6 and DHCPv6-PD is that these kind of low-speed 
devices that should never get multicast traffic, can be put on another L2 
domain and have L3 separation from the devices that actually need to 
listen to the multicast.
I don't know how this might fit into the CPE router draft, but my 
recommendation for CPE router is that it should have at least 4 LAN ports 
and it should be possible for these 4 LAN ports to be 4 separate routed 
interfaces to enable L3 separation between different devices.
This can also be solved by MLD snooping and personally I'd like to see 
both, but it might be cheaper for the CPE manufacturer to implement routed 
ports than to get a built in L2 switch that actually does MLD.
--
Mikael Abrahamsson    email: swmike@swm.pp.se