[AusNOG] Quigley announces architectural "stake in the ground"

[Mark Smith]

Brad Gould wrote:
> 
> Adrian Chadd wrote:
> 
>> So what about direct customer<->customer communication? Does that have
>> to pass through your aggregation layer(s) ?
> 

I think that should only be an option. Mandatory hair-pinning of traffic 
is terribly inefficient when the nodes are adjacent at layer two and 
therefore could take to each other directly (at layer 2 and 3).

> Yes. For a whole heap of reasons.  Its a L2 network - so whats going to 
> route between customers? 

Layer 2 switching, if you want to.

- there cannot be one huge broadcast domain for
> people to play in.

Why not? (and I'm not asking this because I don't understand the risks 
associated with a single broadcast domain)

   Each provider is going to provide a customer with
> probably a /64 from a different network - something needs to allocate 
> and route that traffic.
> 

So how are we going to run IPv4 over it, with our own address space, if 
the service only provides us ISPs with a routed IPv6 service interface? 
What if we want to run a MPLS LER at a SME's site - do we have to 
encapsulate MPLS in GRE in IPv6?

Regards,
Mark.