Re: You asked about multicast scope

[Brian Haberman <[e-mail]>]

On 3/29/12 7:16 AM, Stig Venaas wrote:
> On 3/29/2012 3:38 AM, Kerry Lynn wrote:
>> On Thu, Mar 29, 2012 at 12:02 PM, Iljitsch van Beijnum
>> <[e-mail] <mailto:[e-mail]>> wrote:
>>
>> On 28 Mar 2012, at 12:08 , Fred Baker wrote:
>>
>> >> I haven't read the spec yet, but isn't PCP supposed to work in
>> the service provider run NAT64/CGN case, too? In that case, the
>> multicasts need to escape out of the site or even organization to
>> reach the service provider at least in the SOHO case. So this would
>> be a scope just shy of global, maybe a new "service provider" scope?
>>
>> > I personally rarely use "zero configuration" and "service
>> provider" in the same sentence...
>>
>> Am I understanding you correctly when I take that to mean that the
>> admin (value 4) scope is appropriate because then the people running
>> the multicast routing can determine exactly how far these packets
>> travel?
>>
>> That makes sense, but there is one potential issue, that I think
>> some people who are well-steeped in IPv6 multicast should look at:
>> in this situation, the scope value 4 may need wider distribution
>> than side-wide, which is scope 5. I can't find any documentation on
>
> IMO you cannot do that. The higher scope value, the wider distribution.
> If you have scopes X and Y where X < Y, then the distribution of X
> should be a subset (can be the same) of the distribution of Y.
>
> There are implementations making such assumptions.

It is not an assumption, it is stately quite clearly in the Scoped 
Addressing Architecture (RFC 4007).  From Section 5 :

       A zone of a given scope (less than global) falls completely within
       zones of larger scope.  That is, a smaller scope zone cannot
       include more topology than would any larger scope zone with which
       it shares any links or interfaces.

>
>> whether that's ok or not between sessions right now, but I'm
>> reluctant to assume that a lower scope value can have wider
>> distribution than a higher one without having a spec that explicitly
>> says so or hearing from some implementers.
>>
>> Correct me if I'm wrong, but isn't Admin Scoped multicast defined on a
>> per-
>> port, per-application basis? That would suggest that some multicast group
>> addresses can (selectively) be forwarded through the CER uplink to the
>> ISP.
>
> There is nothing magical with scope 4. It is wider than link-local, and
> you need to configure routers to limit the distribution according to
> your policy.

RFC 4007 lays out all the rules you need to adhere to for scoped 
addressing boundaries.

Regards,
Brian