[arin-ppml] FW: Creating a market for IPv4 address space inabsenceof routing table entry market

[michael.dillon at bt.com]

> > 1. Multilayered NAT to put more users behind each IPv4 address.
> > 
> > 2. Just pirate some addresses from another region. For 
> instance APNIC 
> > allocated all of 126.0.0.0/8 to Softbank, a Japanese cable provider.
> If
> > an American ISP decided to "borrow" these addresses, then 
> few of their 
> > customers would notice that they cannot contact consumer cable
> customers
> > in Japan. 130.0.0.0/8 is another interesting block as are 
> the various 
> > military allocations.
> 
> Is it just me, or do these first two alternatives not look so 
> great? ;-) More NAT-ing and a fragmented internet?

Being an ISP is a business. If it makes business sense to apply more NAT
or pirate IPv4 addresses from Japan, then that's what businesses will
do. Most people do not care about a fragmented Internet, unless the
fragmentation is close to home. 

The fact is that the Return on Investment for the above two scenarios is
limited, especially the NATing scenario, because you have to spend the
money but you really don't control how long you can use the investment.
If the market shifts strongly to IPv6, then you are boxed in and have to
invest a second time. This risk will tend to drive businesses to make
invest in IPv6 rather than stretching the lifetime of IPv4. Since the
IPv6 investment money will go towards v6-v4 interworking such as NAT-PT,
a really careful investor can put their money in equipment that supports
both the multi-layer NAT scenario as well as IPv6.

> > 4. Shop for an upstream provider that will provide the needed
> addresses.
> 
> This is reinforcing my point. The upstream will sense the 
> demand for more v4 addresses and might consider acquisitions 
> of smaller ISPs to get them.

It's much easier to shift your broadband products to IPv6, and then use
the recovered IPv4 addresses for leased-line customers. Nowadays, many
providers are doing this with their dialin infrastructure. This strategy
is what I would call cannibalisation, where an ISP reviews their
products and forcibly migrates certain lower margin products to IPv6 in
order to use the IPv4 addresses for more lucrative customers.
 
> Giving 
> > them to a customer who needs them is just business as usual so I
> expect
> > that the only way to access this surplus will be to buy 
> service from 
> > such companies.
> 
> And that might have something to do with their opposition to 
> transfer markets, eh? 

Why is that? In a transfer market, the ISP who will bundle free IPv4
addresses with Internet access service should always win a deal against
someone who just wants to sell some addresses. This makes it hard for a
market pricing mechnism to work. If very few address blocks are offered
for sale, it may seem as though the price should be $500,000 for a /24,
but if ISPs with spare addresses are still giving them away, then the
real value is close to zero.


> That's called "unbundling." There's a whole economic 
> literature on how unbundling components of a service with 
> different competitive conditions can improve consumer 
> welfare, and how bundling them together can increase the 
> market power of the bundler. 

This discussion does not involve consumers at all. They will 
never buy IPv4 addresses. In any case, the abundance of IPv6 
addresses will tend to improve the position of the consumer
because they can justify a /48 block if they want it.

--Michael Dillon