[ppml] Soft and hard landings, was: Re: Policy Proposal: IPv4 Soft Landing

[Iljitsch van Beijnum]

On 18-mei-2007, at 19:19, Scott Leibrand wrote:

>> The effect of a policy like this is that:

>> a. IPv4 becomes more painful to use because addresses are harder  
>> to get
>> b. the incentive to move to IPv6 is reduced because the moment  
>> that the
>>     IPv4 is completely depleted is postponed

> As written, those two statements appear to contradict each other.   
> To put it in economic terms,

You're right, they appear to contradict. But this isn't unheard of.  
For instance, any economist can tell you that raising prices can  
easily result in making less money.

> the effect of a policy like this is that:

>    * IPv4 becomes more expensive (though additional work required)
>      because addresses are harder to get
>    * as the cost of IPv4 growth goes up, IPv6 will become a
>      cheaper/easier alternative for a number of orgs (for some earlier
>      than others)
>    * as some orgs switch their growth over to IPv6, the moment that
>      IPv4 is completely depleted is postponed.  This delays the
>      increase in IPv4 cost (through slower triggering of IPv4 Soft
>      Landing thresholds), and allows more time for the cost of
>      deploying IPv6 services to come down (through natural replacement
>      of old IPv4-only gear with new IPv6-capable gear, and by giving
>      people time to learn IPv6)

You only look at the cost part without considering the benefit. Apart  
from less tangible things like being a technology leader or being  
prepared, the benefit of adopting IPv6 is negligible: you can only  
reach a tiny part of the internet over IPv6: something in the order  
of 0.1%. So the cost difference between IPv4 and IPv6 can never drive  
IPv6 adoption.

> To re-use a recently proposed analogy, let's say that 5 years from  
> now a major war starts in the Middle East, and a good majority of  
> the oil extraction infrastructure is destroyed..  The US has to  
> decide how to use the Strategic Petroleum Reserve (SPR) in order to  
> minimize economic disruption while consumers switch to alternatives

[...]

The problem with this analogy is that with oil, you need a constant  
supply to operate. With address space, once you have it, you're set,  
no need to come back for more.

So a better analogy would be that once oil is no longer available,  
the switch to other power sources is seamless, but it's no longer  
possible to build new roads for lack of asphalt. Fortunately, there's  
still plenty of steel so in the alternative, it's possible to lay  
down tracks. The car manufacturers anticipated all of this and by now  
all models come with both regular tires for driving on roads and  
steel tire-less wheels for use on the tracks.

The problem is that business etc are unwilling to spend the money to  
convert their parking lots to allow people to arrive over tracks and  
park their cars using the tracks system. This means that cities  
aren't going to install tracks because everyone keeps driving on  
roads, anyway.

But now we are running out of asphalt. We can either use the reserve  
and keep building the roads we need today, even though we know we can  
only keep doing that for a few more years, or we can artificially  
restrict the availability of asphalt.

In the latter case, it becomes increasingly difficult to build new  
roads, but since there is no track infrastructure to speak of, the  
only effect is that it costs more time and money to build less  
effective new roads. Although new roads now have fewer lanes than  
they really need, congestion is a problem, but you can still get  
everywhere by road and the remaining asphalt is going to last for a  
fairly long time so pretty much nobody bothers with tracks.

The alternative is that the remaining asphalt is used up as per  
current needs, and soon it becomes clear that the supply is going to  
run out in the near future, so some people start installing tracks to  
be ready. When asphalt runs out, people still keep driving on the  
existing roads, but all new infrastructure can only be tracks and the  
transition from roads to tracks starts happening in earnest.

> As you can tell, I think that a policy like Soft Landing that  
> attempts to provide incentives for people to gradually move their  
> growth away from IPv4 will help ease the disruption of the  
> transition.  As such, I believe that a proposal like Soft Landing  
> is much better than the alternative of continuing existing policy  
> until the unallocated IPv4 pool is exhausted.

The way I see it, the undeniable prospect of running out of IPv4  
address space within two to four years is a prerequisite for any  
meaningful IPv6 adoption. It could very well be that the prospect of  
running out isn't even enough, and we actually have to run out before  
large numbers of internet users take notice. (In my most cynical  
moments I'm thinking lots of people won't care that the IPv4 club  
isn't admitting new members and aren't even going to bother adopting  
v6 when we're out of v4 space.)

We're currently still in the situation where people can fool  
themselves into thinking that we're not going to run out of IPv4  
space within the next decade or so, hence they don't have to take any  
action. Implementing any measures that postpone the moment of running  
out only give people more excuses to ignore IPv6.