Invisible Hands, was Re: Multihoming sites and ARIN

[Scott Huddle]

David,

Thanks for your comments, they were very thoughtful.

At 02:22 PM 2/21/97 -0800, davidk at ISI.EDU wrote:
>Note that IPv4 space in itself is limited and that alone can distort
>market forces. In many situations where supplies are limited, prices get
>set artificially high and there is no guarantee that this could not
>happen with IPv4 address space.

IPv4 space is bounded (though very large), note that technologies
such as NAT, V6, and clever use of private space expand the supply,
though.  Almost everything is limited ("buy land, God ain't makin'
any more dirt"). In general, markets are the most efficient mechanism
for the allocation of scarce resources. Prices aren't "set" in a market, 
they are determined by supply and demand.  

>Ooh, and what does Jon do with the money? Or do you want to give away the
>address space for free and then some lucky people can make huge profits
>or you divide it very democratically among the current Internet users
>which will be even a better recipe for disaster.

Use the inital money to fund root name servers, or the IETF, or the
fine work at ISI :)  What happens with the money from the auction of
radio frequencies?  

>> Note that the current allocation schemes can destroy the routing
>> system and there are no methods to fix it.
>> 
>Note that a free market system can do the same. And there are no methods
>to fix it ...

I disagree.  Market demand is shaped by aggregate utility functions
for the members in the market, obviously the utility of broken 
routing is zero, and the price for a routing slot in a broken system
is zero.  From the supply side, why would an ISP sell a route slot 
that breaks their systems? The price for that would be infinite.

>The current registry system is not perfect but there is no evidence that
>current allocation policies are destroying the routing system. In fact,
>the policies are defined by consultation with the users of the address
>space to avoid just that.

Yes, I agreee, but there are examples of market failure (e.g. little
sites wanting to multihome) that aren't met by the current procedural
scheme.  Also, note that not all of the users of address space are
consulted -- only certain users get space from the registries.  Indeed
most of the "heat" on the list has been generated by those that
feel have a stake but don't have a voice.  In a market model, they
all vote.

>> Clearly, a market based scheme insists on making things work
>> (else why buy the slot or pay for the space) 
>
>And what happens if a free market registry sells one address block too
>much?

I'm not sure I follow.  The routing slot market is controlled 
by individual ISPs, IP space by a global market.  Can you clarify your 
point?  

>> I assert that a free market would determine the demand and supply
>> curves and an equilibrium point for a default free routing table.
>
>And what if demand is bigger then supply ?

Prices go up.  What happens in a procedural based allocation?  Do
you tighten the allocation and make it harder to get space?  This
is effictively raising the price as well.

>And what if the equilibrium point for the number of route entries is much
>higher ?

You have an effective damper on route explosion, you have a market to
encourage renumbering, and you have a market to encourage deployment 
of additional technology such as NAT.  Sounds like A Good Thing to me.

How does the current procedural allocation have *any* direct control on 
route entries?  The current processes are indirect at best.

>The free market will certainly solve these problems but it could turn out
>to be more expensive then our current system.

Why?  A free market typically drives costs *down*.  I'm interested
in your examples here.  OPEC controls the supply of oil, how do
Evil Greedy Bastards control the supply of new IP space?  

As a counter example, witness the Bass brothers debacle in trying to 
control the silver market.

>Don't get me wrong, I am not a supporter of bureacratic solutions over a
>free market economy and would prefer any free market solution. However,
>there is no proof whatsoever, that a free market will indeed work better
>then the current system, and even worse then that, if this turn out to be
>the case, we cannot go back anymore. Therefore, it might be the time for
>an experiment with a limited amount of address space to test (part of)
>the free market theories because a free market solution indeed has a lot
>of potential and could indeed turn out to work better, as you claim it
>will, then the current approach. But then, this is not the topic of the
>ARIN list but <piara at apnic.net>, so please reply on that list only,

Thanks for these comments.  

best regards,
-scott