Monday Night Footbal -- on Google?

Saku Ytti saku at
Sun Jan 15 19:14:56 UTC 2012

On (2012-01-11 17:45 -0500), Justin M. Streiner wrote:

> >If multicast is used it shouldn't take 150pbps, it should be much lower.
> That could be one of the things that helps spur v6 adoption -
> multicast being somewhat less of an afterthought :)
> While v4 multicast works, and delivering video is one of the things
> it can do very well, some networks don't route v4 multicast or
> exchange v4 multicast prefixes, so its utility on a wide scale can
> be limited.

This is misguided, IPV6 does no magic to help scale multicast to Internet
scale compared to IPV4.

Scaling multicast to Internet scale  would make our core routers
essentially flow based routers. And as there is finite amount of how many
of these flows you could hold, we would need some way to globally regulate
how and who can push their content as multicast and save lot of money and
who will have to pay the full price.
Those who are left out, might feel like multicast is used to stop

Now maybe we could specify some sort of stateless 'manycast' in IPv6, where
you'd map destination AS numbers as source address. Needing to send only
one copy of traffic per destination ASN (or less if you can map multiple
ASN in source address), and then destination ASN would need to have Magic
Box to do stateful magic and could cherry-pick what they care about. But
that's lot of complexity for very incomplete solution, as it would only
remove states from transit.


More information about the NANOG mailing list