On June 30, U.S. federal government officials expect to declare an early victory on the IPv6 front. But they admit that meeting their much-heralded June 30 deadline for IPv6 compatibility is just the opening salvo of a long-term battle to get their networks ready for the Internet of the future.

Opening salvo. And the standards track RFC for IPv6 (RFC2460) is nearly ten years old.

Sean Michael Kerner, writing at internetnews.com, has some interesting insights on the slow IPv6 take-up outside of US government agencies.

Whenever I step outside the US Government bubble and ask vendors or regular enterprises about IPv6 I typically get very little enthusiasm. In fact, at the recent Interop and NXTcomm shows in Las Vegas, I barely got more than a shrug from any vendor I asked about IPv6. Many will claim to support it in their products, but it's not a factor that is driving demand.

The reality is that the current IPv4 addressing system is working just fine for American enterprises and consumers. Though address exhaustion is predicted globally within a few years, it's not a pain that anyone in the US is really feeling today.

No pain, no gain?

It is unquestionably the case that most users of IP (and, by that, I mean both consumers and providers of content) are either unaware of IPv6 or simply don't see any compelling need for it. They are feeling no pain and, without that pain, there's little incentive to do anything.

The problem is that, by the time anyone is feeling any pain, it may be too late.

So. How might that pain manifest itself?

Here are some possible scenarios.

IPv4 rationing

As end users of Internet services so vastly outnumber providers of those services, it is they who will feel the pain first.

Consider an ISP that has, say, a million customers but an IPv4 assignment of only 524,288 addresses (a /19 network). Clearly there are insufficient IP addresses available to allow every one of those million customers to be on-line simultaneously but, as addresses are allocated dynamically and only for short periods, the ISP already has a means - DHCP - by which he can ration addresses.

In order to prevent this from becoming a simple lottery, the ISP can introduce a tariff structure under which customers would pay a premium to be guaranteed an IP address on demand or could elect to pay less and take their chances.

For an additional, possibly very large premium, customers could be permanently allocated a static address.

NAT becomes the norm

There's now no shortage of companies here in the UK offering mobile broadband, delivered over 3G networks.

If you've ever used one of these networks, you'll know that you are generally not allocated a public IP address when you connect but rather a private one in the 10.0.0.0/8 network reserved by RFC1918. This address is translated at the network border, so that you can gain access to some Internet resources.

Yes, some but not all.

In the past, we have occasionally allocated 3G networking devices to mobile users and the most common use to which these have been put is for replicating Lotus Notes mail and applications while travelling.

Recently, these just stopped working. We were able to load web pages, using http and https, as well using as a variety of other protocols, but Lotus Notes connectivity was lost. The ISP had locked down all but a small range of commonly used ports, probably as a file sharing countermeasure (not a spam countermeasure - port 25 was still open). Port 1352 was a casualty.

When we probed the ISP's technical support people about it, they asked what we were trying to do.

"Open a connection on TCP port 1352," we said.

"Why?"

"Lotus Notes replication."

"We don't support Lotus Notes."

I predict that, as IPv4 becomes more scarce, we will see more providers - and not just wireless ones - electing to provide service using RFC1918 addresses and NAT. These providers will similarly restrict connectivity, partially as an abuse countermeasure and partially because implementing NAT on a very large scale will be expensive.

Companies holding very large network assignments recognise their value

Some early adopters of IPv4 still hold permanent assignments of very large networks. For example, IBM has a /8 (nearly 17 million addresses) and HP has a /7 (over 33 million addresses).

While logistically complex, it is theoretically very simple for these companies to renumber their internal networks so that they either use a near contiguous range of addresses within these huge assignments or migrate to RFC1918 address space. This is primarily a project management challenge, not a technical one.

If the project were properly resourced and executed, these companies would then be in a position to sell or, more probably, lease very large networks to interested parties and this would have the effect of extending the life of IPv4 and create a nice little revenue stream for the companies concerned.

Or, some federal agency or other regulatory body may try to enforce the handing back of some of this wasted IPv4 space but that will only lead to court, and will thus not extend the life of IPv4 at all.

Category: IPv6
Technorati: IPv4 IPv6

Unable to post a comment? Please read this for a possible explanation...