Re: New draft on embedding the RP address in IPv6 multicast address

[Brian Haberman <bkhabs@nc.rr.com>]

Hi Rami,

Rami Lehtonen wrote:
> 
> Hi,
> 
> I also see another problem in embedding the RP address in IPv6 multicast
> address. Now the sources have to know the address of the RP they are going
> to use, when selecting the multicast group address. Is there need for RP
> address discovery/announcement?

I don't think so.  Generally, I see applications being assigned
addresses
by network administrators.  That allows the admins to setup their group
to RP mappings.

Other opinions?

> 
> Another problem that was identified already is that we have only one RP per
> group. This means that sources from the another domain than the RP can't use
> their own domain RP without RP-RP signalling (e.g. RP-RP Register messages).

Right.  That is one of the open issues I pointed out.

> 
> Also the receiver domain's RP does not join SPT (S, G) but (RP, G) which
> might be suboptimal. It could change afterwards to (S, G) when it learns the
> source address but this needs additional state creation signalling
> throughout the domains in between.

Actually, most PIM routers switch on the first packet anyway.  So this
approach gives you the same level of support as using MSDP.

> 
> And last point is that if there are no active sources for that group, this
> model allows Internet wide multicast state creation (from receiver(s) in
> another domain to the multicast group RP in another domain) compared to the
> domain wide state creation in IPv4 MSDP model. This is because we don't have
> any source active information available.

Good point.  That issue needs to be investigated.

Thanks,
Brian