[arin-ppml] Routing table growth, was Re: IPv4 is depleted today

[Iljitsch van Beijnum]

On 3 sep 2008, at 18:08, John Curran wrote:

> Okay, I'm holding a very different set of assumptions...
> where am I mistaken in the list below?

No problems with 1 - 4. If v4 trading becomes possible, then 5 - 7. I  
don't think 8 - 9 are too relevant, other mechanisms will make ISPs  
route address space. (I.e., the ability to bill customers.)

I think the place where our views differ is that I have assumed that  
there are two classes of end-users: ones that require a very small  
amount of address space from their ISP (1 address or less) and ones  
bringing their own address space. (There are other classes as well but  
I'm not sure if those exist in large enough numbers to be relevant.)

Your assumption seems to be that for an individual customer, address  
space will be obtained somehow.

In my model, the number of people bringing their own address space (=  
the 90% of allocations and 10% of address space given out each year  
currently) won't change when we "run out" of IPv4 space because small  
blocks can presumably be obtained somehow, there's enough scraps  
laying around and/or being returned to the RIRs to make this continue.  
But the people that need to get space from their ISP aren't going to  
go out and find address space themselves (otherwise they'd be in the  
other category) and ISPs aren't going to do this on their behalf  
because it's too time consuming and too expensive, and is a dead end:  
every year it will be harder and more expensive. Might as well put in  
the big NAT boxes immediately.