General

The Closing of the Internet Frontier —Contined

An unsatisfactory situation
However, two large problems remain. Firstly, the technique of combining NAT and private addressing, while minimally functional, is deficient in a number of respects. There are many classes of application that simply can't function over NAT: some kinds of content streaming, one kind of FTP, anything involving end-to-end security like IPsec—the list goes on. There are workarounds for some of these, but we have lost something important—the notion that an IP address is a routing identifier of global significance—and this loss means that all solutions involving NAT will be inherently inefficient, suffering from manageability and scalability problems.

Secondly, some mobile phone companies are engaging in third generation (3G) [definition] deployments already. Some of them are making a virtue out of the fact that they are startup companies with an exclusive focus on 3G, Hutchison 3G UK being perhaps the most well-known example. These companies, being startups, have no IP address allocations and precious little prospect of obtaining globally valid ones for all of their customers' terminals.

The answer is to upgrade
There is a way out. IPv6, the successor to IPv4, has a greatly expanded address range, enough to fit in the current Internet many times over. Furthermore, it's likely that organisations applying will be assigned more space than they could ever possibly use. You could even encode the phone number in the address, providing you didn't have too many digits.

You still need to talk to the IPv4 Internet, of course. One way of doing this would be to have dual-homed proxy servers, operating on both IPv6 and IPv4 addresses. When a request comes in from an IPv6-addressed handset, an IPv4 HTTP request goes out to the greater Internet, and the result is returned to the IPv4 proxy server and then to the IPv6 handset.

It's NAT by another name, certainly, but there are many interoperability mechanisms defined currently, and a high confidence that as isolated islands of IPv6 begin to crop up, full IPv6 connectivity will soon become a possibility. (There are also some organisations providing commercial IPv6 service today, although I would advise having someone with serious IPv6 experience on board before investigating this.)

It might seem like a long way to go, but consider the statistic that the top five web sites generate the vast majority of HTTP traffic; as soon as they switch, will not the rest of us follow?

IPv6 assignments are being made as we speak, and could cost you approximately the same as IPv4 for a 'small' block ($2,500), rising to $25,000 for a 'large' block. (Note: I believe that the charges ARIN used to levy for IPv6 allocations are no longer in force.) There are however some stringent requirements for IPv6 requests in the initial phase. You may want to talk to a qualified consultant (such as us) or your own staff to see if IPv6 can solve your addressing nightmares.

Summary
If you are an ordinary ISP looking to jump up a level, plan carefully and success will be yours. But stick with IPv4 for the moment—many UNIXes come with IPv6 stacks, which will be relatively easy to upgrade when the time comes along. Mobile phone companies, however, ought to look very seriously at what IPv6 can do for them, while not forgetting their IPv4 allocation.

—End

Back to page 1