World IPv6 Day: Time to packet in?

Around 35 years ago, Vint Cerf at and Bob Kahn at Arpa were designing a set of networking standards that were to become TCP/IP, or Transmission Control Protocol/Internet Protocol, the underlying communications protocol of the internet.

One of their basic principles was that information that moved across the face of the network could be of any length in total, but was broken down into packets with a maximum of a thousand bytes or so that the network fabric could be guaranteed to handle it efficiently. To help move those packets across, each had various bits of information — including the address of the destination computer — bundled with the actual data being copied.

Vint Cerf was one of the designers of TCP/IP, the underlying protocol of the internet. Photo credit: David Meyer

The form of that address was crucial to the efficiency and utility of the network. If a megabyte of data had a thousand packets, then every chunk of extra processing needed to analyse and act on each packet's address would be multiplied a thousand fold. Conversely, if the addresses were too short, the total number of internet connections would be too small to be interesting, and Cerf and Kahn were thinking big.

For efficiency's sake, the address was made fixed-length, enormously simplifying the data structures and processing required in routers. The length was also limited to four bytes — 32 bits — which was a sensible compromise at the time; it reduced the amount of memory needed as overhead to the actual message data, while allowing for over four billion total addresses.

In the mid-1970s, that was greater than the global population — and for a network with no particular plans to break out of research, industry and academia, that seemed reasonably future-proof. IPv4, as the standard was known, was ready to go.

But the mid-1970s had a couple of surprises, such as 1975's Altair, the first personal computer to catch the public imagination, and a small company set up to create software for it — called Microsoft.

Space runs out

Fifteen years later, IPv4 had escaped from the lab and was well on the way to devouring the world. The simplicity and efficiency that Kahn and Cerf had adopted out of necessity made the networking protocols ideal for each new wave of cheap, connected hardware that spread around the world; suddenly, those four billion addresses seemed limited.

It wasn't so much that every person in the world would have a computer; more that for each human, tens or hundreds of computers would exist in gadgets, mobile phones, cars and smart infrastructure. And all those devices would want to talk. In 1992, work began on IPv6 (don't ask about v5) to fix this address exhaustion, as well as various security and management problems that had started to show.

IPv6 had a troubled birth, with lots of competing ideas and a fair amount of acrimony. However, the final specification included one key decision — the address field was quadrupled to 128 bits, giving a total number of possible connected devices of 340,282,366,920,938,463,463,374, 607,431,768,211,456. Which is enough for all the stars in the universe. Problem solved.

IPv4 strikes back

Or rather, it wasn't. In an ideal world, once IPv6 had been finalised, it would have been adopted by each new device coming onto the market, gradually edging out the old system until it achieved dominance, then exclusivity.

Even though the old address space is now full, you can still buy new internet devices, plug them in and have them work, which is all most people care about.

But another decision in IPv6 — to make the new addressing scheme incompatible with the old — meant that couldn't happen. There were very good engineering reasons to avoid the inefficiency and mess of trying to extend the old idea and start afresh with everything network scientists had learned, but practically, it was a disaster.

The old IPv4 system has fought back, too, by sprouting patches such as NAT (Network Address Translation), which effectively multiplies the total number of addressable devices on the old internet by 65536. This means that even though the old address space is now full, you can still buy new internet devices, plug them in and have them work, which is all most people care about.

The other things that IPv6 is good at — security, streaming data to multiple places, management — have also been addressed by a series of new ideas in IPv4.

IPv4 or IPv6?

The designers of each new device had a choice: use the old system and connect to everything on the internet, or use the new system and connected to practically nothing. The result, in 2011, is the need for a World IPv6 Day, when lots of people highlight the existence of the IPv6 parallel universe and try to connect to each other, and as many users as possible are encouraged to set up IPv6 and see how well it works. Which, if your ISP doesn't have IPv6 enabled, is not at all. There are ways to gateway between IPv4 and IPv6, but it doesn't get you very much.

And there are plenty of reasons not to bother: if you are going to run IPv6 as a user, rather than an internet infrastructure company, you'll have a brand new set of bits of software connected to the network with all their security concerns, configuration issues, patches to maintain, performance to tune and so on, without being able to ditch your old IP and without any particular advantage. Also, if you think the average domestic router is flaky with IPv4, you ain't seen nothing yet.

It's no more palatable if you're a company with a large network presence, as you'll have a large set of tools built around that old IP standard: they'll need to be replicated and the new ones tested and kept updated, for a user base you could keep in a matchbox. In short, you need a really good reason to go to IPv6, and those really good reasons — even with address exhaustion — are in short supply.

Some people do have really good reasons. IPv6 has been widely deployed by network providers in the backbone of the internet and in mobile communications systems. It will, in time, become the standard. It's just that, as World IPv6 Day demonstrates by its very existence, that time is not yet.