Will IPv6 ever be a reality?

February 19 2010

Update: 6:40am pst Feb. 3, 2011 — The central pool of IPv4 addresses is depleted! We’re not really out of addresses though, not all IPv4 are used, but we’re getting close.

I’ve recently re-read The IPv6 mess by D. J. Bernstein. The essay was probably written in 2002 or 2003. It predicted remarkably well the current state of IPv6. IPv6 can’t access an IPv4 network. Because all the good stuffs are on IPv4, nobody uses IPv6 and everybody sticks to the old protocol. There are tunnel mechanisms to bridge the two, but they are extensions; not part of the core protocol.

The pool of available IPv4 addresses is shrinking and will probably be exhausted by 2012. Without any IPv4 addresses left will the world starts using IPv6? I suspect this will not happen. The problem with IPv4 is the exhaustion of public addresses, but most people don’t need a public address. They are using Internet to read their mails and surf on the web. They don’t want to have their own server at home.

Consider the following scenario: Late 2012, the pool of IPv4 addresses was exhausted a few weeks ago. You’re an Internet Service Provider (ISP) and Internet is still mostly IPv4. New clients want to have access to Internet, all of it; not just the part that was migrated to IPv6.

You could get more IPv4 addresses nonetheless. It would be relatively simple to do: pay more. When demand for a resource is high and supply is short, prices go up. In 2010, hosting providers charge around $1 per month for an additional IPv4 address. Assuming it’s about the same price for an ISP; $1 when clients are paying $20 or more each month is reasonable. But as Internet keeps growing, the price will be driven further up and eventually we’ll have to find another solution.

IPv6 proponents argue that it will be the time when the new protocol will take off, and all our problems will be solved. But IPv6 is not an easy or cheap solution. You need to change or update network gear, older software might not be compatible. And many people will have to be re-trained.

Oh no! NAT to the rescue!

NAT used at a large scale would be cheaper and easier. Instead of giving all clients a public address, multiple clients share one public address.

Suppose an ISP is charging $20 per month, and the price of an IPv4 address went up to $5 per month. If it uses 1 address for 10 clients That’s a saving of $45 over a revenue of $200 each month: almost a quarter of it.

NAT is not free, routers will have to be beefed up and reconfigured. But overall it is cheaper and easier than IPv6:

You don’t need to upgrade all you infrastructure. You can gradually switch your clients to NAT; freeing up IPv4 addresses for new clients. The cost can be spread over a long period of time and the benefit is immediate. With IPv6 you have to pay the full price upfront for a distant benefit.
NAT is already widely used. Almost all system administrators had to setup a NAT at some point. The cost of re-training them will be lower.

I’m not arguing that using NAT is better than switching to IPv6. NAT is not a good solution, it’s a bad short-term solution aggravating the problem in the long run. NAT is like the fast food restaurant around the corner and IPv6 is the healthy meal you cook at home. You know you should go for the healthy meal, but you’ll go to the fast food, because it’s cheap and easy. You’ll feel bad, but forget about it quickly, because it doesn’t have immediate consequences.

Like the OSI model, IPv6’s technical brilliance wont make it successful. Regular Internet users want to surf the web and read emails. They don’t care about how things work, as long as it works. If IPv6 could communicate natively with IPv4, it would have been widely deployed and used today.

In the future we will drive IPv6 connected flying cars on Mars

That’s what —I think— will happen:

2012 or later, no more IPv4 address left
ISP’s start using NAT for their cheaper plan
After a while they supply clients who need a public address —typically BitTorrent users— with IPv6 addresses, but keep using NAT for regular v4 access
IPv6 gains momentum in some area, like gaming or file-sharing. But other domains like the web —HTTP— or email —SMTP, POP, and IMAP— stick to the old IP protocol because it’s good enough

I don’t know if there will be a complete switch to IPv6 one day but I am sure nobody will switch overnight just because there is no IPv4 addresses left.