Sean Doran wrote:
As tli pointed out the top of the hierarchy is not arbitrary, it must be default free.
Sorry for the nitpicking, but this definition has at least two flaws: a) there's a bi-partite backbone configuration, where each half has default pointing to the other half. Both do not have to carry full routes. (Of course, this scheme has problems with packets destined to the blue, or can be extended to more than two partitions). Actually, there's a very simple way to fix the problem with packets to nowhere. Simply have routers at exchange points to drop packets routed back to the interface from which they came from. b) multihomed non-transit networks may want to be default-free and carry full routing to improve load sharing of outgoing traffic. Since they are non-transit, they cannot be considered "top of hierarchy".
Consequently, the number of things in the top, default-free hierarchy is always going to be limited, no matter what "type" of hierarchical allocation scheme is proposed.
Bingo. Faster boxes, anyone? :)
The further requirement that any given area be fully contiguous means that the "top" of the hierarchy must be self-repairing.
Note that IXPs are not "top" of the hierarchy, but just aggregators for essentially point-to-point links between tier-1 backbones.
One could propose to implement this as a big bridged network. The original DGIX proposal was along these lines. Operational experience with much smaller but still big bridged exchange points has demonstrated pretty much conclusively that this is a Really Really Bad Idea.
Not only technically -- politically that was a suicide, as it assumed a signle operator (consortium, or pork money funded). --vadim