[ppml] NANOG IPv4 Exhaustion BoF

[Iljitsch van Beijnum]

On 5 mrt 2008, at 13:58, Joe Maimon wrote:

>> Who the **** cares what happens when we're out of IPv4 space?  
>> Because  then we'll be out of IPv4 space, and the if THAT isn't  
>> enough to get  people to adopt IPv6, we might as well all go home  
>> and read up on the  X.25 specs because this internet thing will be  
>> over and done with very  quickly.

> Yes, that was the prevailing group-think after the whole "we will  
> build it and they will migrate" thing didnt work out.

So what are you saying here? That IPv6 is never going to fly, because  
if it would, it had done so immediately after its creation?

> ipv4 will "run out" and the whole world will wake up the next  
> morning and mass migrate to ipv6. This is your viewpoint, one you  
> have been consistent about.

> This is a pipe dream. Or a nightmare. Or a pipe nightmare.

If it happens when you're awake it's generally not a dream or a  
nightmare. People who have all the IPv4 address space they need won't  
be doing much around the time we run out. But those who need an influx  
of new address space to keep their business running (i.e., ISPs) and  
those who want to start something new will have to do something  
different than they've been doing so far. If that something is not  
IPv6, that means the something won't be sustainable in the long run.

> Here is another viewpoint.

> ipv4 wont run out. It will just become scarce (and scarcer), and  
> resultingly, valuable (and more valuable).

Sure, if you want to have a service hosted, you only need a few IP  
addresses, which you'll be able to get one way or another.

However, we just saw pretty pictures that indicate 83% of all ARIN  
IPv4 address space is held by 1% of the ARIN members. Those few dozen  
organizations use up millions of addresses each year. They pay less  
than a cent per address per year for that now. For them, the prospect  
of having to pay (say) $1/address the first year after the v4  
exhaustion will be an unattractive one, but then having to work harder  
and pay more each year is what makes this completely unworkable. So  
either they'll collapse their existing address space by putting more  
users behind a single address, or they'll go to IPv6.

> Nobody will mass migrate to ipv6 because:

> - it doesnt work well enough for their needs
> - they dont understand it
> - they cant afford the equipment forklifts
> - they cant support it
> - they cant afford to support it
> - its a network without anything compelling on it

> The last point is the one that makes or breaks.

These are all things that can only be attributed to IPv6 AS IT IS  
TODAY. They are not inherent properties of the protocol. At some  
point, we'll need to jump over our own shadow and stop demanding that  
IPv6 has the same usefulness as IPv4 from the very first day it's  
adopted. (That is not to say that ARIN shouldn't fix its IPv6 path MTU  
discovery and routing issues...)

I've written enough on why trading address space is a bad idea, but I  
also wrote a lot about why PI for IPv6 was a bad idea, and that went  
through anyway. So I'm not wasting my time trying to convince the  
unconvincable. But to the would-be IPv4 traders: depending on how  
exactly this happens, it can be merely a waste of time and money, or  
it can blow up in our collective faces. Please make sure you get it  
right.