[arin-ppml] Q1 - ARIN address transfer policy: why the trigger date?

Tom Vest tvest at pch.net
Fri Jun 20 16:21:47 EDT 2008

Comments inline below.

On Jun 20, 2008, at 1:38 PM, Leo Bicknell wrote:

> In a message written on Fri, Jun 20, 2008 at 04:49:34PM +0000, Paul  
> Vixie wrote:
>> isn't this inevitable in times of shortage, regardless of the shade  
>> (regulated
>> or unregulated) of the market that inevitably occurs when humans face
>> shortages?  if it is, then our choice is not whether a market  
>> appears, but
>> rather, whether that inevitable market is regulated or unregulated.
> Without getting into the details of implementation (and by extension,
> how various countries have implemented them over time)...
> The two extremes are capitalism, which generally leads to the rich
> getting what they need and the poor getting nothing, and socialism,
> which generally leads to everyone getting a proportional amount
> (possibly less than they need).
> However, I think "shortage" doesn't encompass the entire situation
> in this case.  The really interesting question is one of fungibility.
> If your worry is that the poor eat then a freeze in Florida zapping
> the orange juice crop is not a big concern.  Yes, the price of
> orange juice will rise, and only the rich will have orange juice;
> but it's not the case that the poor have nothing.  They can have
> apple juice, or grape juice, or soda, or water, or lemonaid.  To
> those on a budget beverage choice is highly fungible.  This also
> has a great tempering effect on the orange juice market.  When the
> price rises a small amount demand naturally eases, keeping the price
> from spiking to absurd levels.
> Compare with say, oil.  For most consumers there's nothing they can
> put in their car tank besides gasoline, and swapping out the vehicle
> for a non-petroleum based vehicle is very difficult.  A relatively
> modest supply issue causes huge spikes in the price as a result.
> So the debate hinges on how much you believe IPv4 and IPv6 are
> <mutually substitutable>.  Are they like orange juice?  If so it  
> would take only a
> minor increase in IPv4 costs to drive people to IPv6, which would
> automatically prevent an IPv4 market from ever becoming a large,
> widespread thing.  If you believe IPv4 is more like oil, and IPv6
> is like hydrogen technology, promising but not yet ready then the
> lack of IPv4 will create a huge, volatile market as people panic
> since there is no real way they can move to IPv6.

Hi Leo,

I think that, purely from a protocol/software standpoint, an  
individual IPv6 address might be pretty close to substitutable for an  
individual IPv4 address, if not today then pretty soon.

However, that narrow equivalence is dwarfed by other huge contrasts  
between IPv4 and IPv6 -- gaps that might easily be exacerbated or  
locked into place forever if IPv4 is privatized, or a decentralized  
"transfer market" is legitimized, if you prefer that terminology.

Consider: Even if individual IPv6 and IPv4 addresses are technically  
equivalent to each other tomorrow, IPv4 could continue to be non- 
substitutable, and to have substantial "exchange value" (i.e., command  
a high price) for as long as most or all of the universe of "real  
Internet resources" (i.e., users, content, media, services, etc.) are  
attached to the Internet via IPv4. Randy has taught us all well: "IPv6  
is not compatible on the wire". So, for long as that remains true,  
every operator that wishes to have even the potential to exchange  
traffic with that universe of legacy resources will need at least a  
few IPv4 addresses -- either PA from an incumbent ISP, or PI from an  
incumbent ISP (or maybe just an independent IPv4 "dealership").

Under most conceivable/internally consistent scenarios, incumbent ISPs  
(and/or non-operator speculators) eventually will be the only source  
for IPv4 for all practical purposes. Incumbents that actually use IPv4  
may or may not be willing to part with any; in fact some will almost  
certainly continue buying IPv4 until the price becomes "really  
prohibitive". However, some incumbent IPv4 owners may be incentivized  
to sell some IPv4 to make money. If you are commercial operator (esp.  
one that is publicly traded), then your first job is to make as much  
money as possible -- maybe by adding customers, maybe by servicing  
growing customers, maybe just by commanding a higher margin across all  
customers, regardless of their growth rates. If you are adding  
customers, or your customers are growing, then (to repeat) maybe  
you're also an IPv4 buyer yourself -- and thus acutely conscious of  
the fact that anything you sell today you might need to buy back,  
perhaps at a much inflated price, in the future. However, if you think  
you can command a higher one-time price for selling off PI IPV4 today,  
compared to the recurring revenue from reserving/using that for  
current and future direct customers in the months/years to come, then  
maybe you consider selling some now.

Since your bottom line is revenue/profit, how will you maximize this  
particular, possibly transient opportunity? First of all, you won't  
dump all of your unassigned addresses all at once, at any price, any  
more than you'd sell 100% of your salable upstream bandwidth all at  
once if that obliged you to flood the market and undercut our own  
prices. You'll definitely have a "reservation price" below which  
you'll just use internally or save for a more eager buyer.

By the same logic, you won't be cannibalizing the IPv4 addresses that  
your customers are using directly -- not so much because they would  
get upset, but rather because as long as your edge resources (and  
those all of the other incumbent IPv4-based operators) remain attached  
via IPv4, then the IPv4 requirement will persist, and the IPv4 market  
will remain profitable. If IPv6 remains (at best) (just) an acceptable  
substitute for IPv4 -- i.e., offers no new features that are  
unavailable via IPv4 -- then those edge resources will be the absolute  
last IPv4 addresses to go... and even after they're gone too, maybe  
you'll continue running a reverse gateway (aka "IPv4 bottleneck  
server") to keep that IPv4 market alive just a little bit longer...

How many really large operators would have to "choose altruism" and  
forego this strategy in order to preclude this outcome? Perhaps that  
would be a good research question for a better economist than me...

Barring such altruistic defections, how long will the IPv4 era last --  
i.e., how long will it take for (/30 x total number of direct peerings  
between all ASNs) to consume 32 bits, assuming that growth and new  
entrant rates are sensitive to the rising cost of access? That would  
be an easier one to extrapolate -- even a poor economist like me can  
intuit that the number would be very, very large.

Bottom line: A "resource transfer" market will magically transform  
IPv4 into the new and dramatically improved "last mile bottleneck" --  
"dramatically" improved because the bottleneck will be between  
everybody and everybody else.

Apologies for taxing the patience of anyone who got this far; I'll do  
my best to spare the list another telling of this particular story, at  
least, hereafter.


> I personally happen to think they are relatively fungible.  The
> mere fact that IPv4 is unavailable will drive people to IPv6, and
> any small market that develops (black, white, or grey) will only
> accelerate that trend.  To that end any market is self-defeating;
> and I think very quickly so.  If you believe, as I do, that the mere
> fact that there is no more IPv4 available from ARIN will drive  
> people to
> IPv6 then I think it may well be the case that a "market" does not
> appear. (*)
> Others have different opinions.
> * Which is not to say I believe there will be zero transactions.
>  Someone will hijack space, someone will sell, someone will buy.
>  For there to be a real market, there must be multiple buyers, and
>  sellers and some reasonably well defined way for them to meet.
>  A few isolated individual transactions does not count, at least
>  to me.
> -- 
>       Leo Bicknell - bicknell at ufp.org - CCIE 3440
>        PGP keys at http://www.ufp.org/~bicknell/
> _______________________________________________
> You are receiving this message because you are subscribed to the  
> ARIN Public Policy
> Mailing List (ARIN-PPML at arin.net).
> Unsubscribe or manage your mailing list subscription at:
> http://lists.arin.net/mailman/listinfo/arin-ppml
> Please contact the ARIN Member Services Help Desk at info at arin.net  
> if you experience any issues.

More information about the ARIN-PPML mailing list