[ppml] My view on IPv4 (was: Re: IPv4 wind-down)

[Iljitsch van Beijnum]

On 22-mrt-2007, at 9:04, Sean Reifschneider wrote:

> I've never tried to reach a service that I couldn't reach because  
> it only
> had a v6 address.  My network can handle v6 no problem, but there's  
> 100% v4
> coverage, as far as I can see, and so there's absolutely no reason  
> for me
> to go through the pain of setting up and maintaining v6.  Setting  
> up v6
> right now would, literally, gain me nothing.

(Except future-proofing.)

This is very true, and will largely remain true as IPv4 space runs  
out. People who have IPv4 space today won't have a problem, it's the  
people who need more address space when there no longer is any who  
will be in a bind, because at that point, the vast majority of all  
internet users will still only be reachable over IPv4. This means  
massive amounts of NAT. This is not the nice, friendly NAT where you  
have three PCs behind your Linksys DSL router and you get to forward  
ports so fussy applications like VoIP still work.

It will be the kind of NAT where a service provider puts 10, 100 or  
even 1000 customers behind a single IP address, and the number of  
usable TCP ports starts being a problem. Forget about port mappings  
and hence any applications that are more complex than client-server.  
At that point, those users may want to add IPv6 to their heavily  
NATed IPv4 so they can run peer-to-peer and server-to-client  
applications in addition to client-to-server applications. Only at  
THAT point, it will become interesting for people with enough IPv4  
space to also support IPv6 so they can talk to those of us who are  
behind several layers of NAT.

In other words: the running out of IPv4 space is a necessary  
requisite for wide scale IPv6 adoption. Without it, nothing is going  
to change. Therefore, any policy that seeks to artifically avoid  
running out is harmful because it perpetuates an address starvation  
model. We need the water to boil at some point so the frog jumps out.

A few data points: 90% of the IPv4 address space is used up by 10% of  
the requests, which are done by large ISPs. The last two years we  
used 170 million addresses per year, but this year so far 57 million  
addresses, with 1244 million still available, or 5 - 7 years worth  
(without considering additional increases in the numbers of addresses  
used per year).

> What I need are *USERS* who are on v6 who are trying to reach these  
> sites.

I've been running IPv6 at home for several years. When I vist one of  
your sites, how do you know that I would prefer to get at them over  
IPv6?

> These are the places we need to be providing incentives to to  
> switch to v6.
> We need to reach a tipping point.  There are two ways of going  
> about that:
> One is to try to convert tens of thousands of entities who have  
> smaller
> allocations (like /24s), the other is to go after dozens of  
> entities that
> have larger allocations (like /8s).

There are 43 /8s given out to end-users. Reclaiming all of those  
gives us 3 years extra at the rate we've been burning IPv4 addresses  
the past quarter. Although that could be useful, it doesn't change  
the fundamental issue, just the date at which the pain will make  
itself felt.

> I don't think that treating v4 like land, and giving people the  
> ability to
> sell it, is a good way to go.

I disagree, for the following reasons, in no particular order:

- the selling off of small blocks of IP space will further fragment  
the routing table and make our routing problems more urgent

- demand without supply leads to increasing prices, increasing prices  
lead to hoarding, we could very well see the number of available  
addresses go _down_ rather than up

- being able to sell addresses for money (or even the possibility of  
that in the future) will be a strong disincentive for returning  
address space for current /8 holders

- it's unfair that more than 50% of all IPv4 address space is held by  
US entities which then get to make a lot of money from them, while  
the developing world holds next to no address space and would have to  
buy it from richer countries

- markets are unpredictable, but a smooth IPv6 transition can only  
happen in a predictable environment. Without this, people will either  
transition too slow and run into reachability gaps when they're not  
on IPv6 fast enough, or too fast and spend more money than necessary