Is your protocol upgrade really necessary?
Published: 08 Mar 2002 13:09 GMT
It is a truism of computing that if something's old, it must be updated. With nearly 30 years under its belt, the IP protocol has provoked a genuine global revolution -- and some think the cracks are starting to show. Like a bright new housing estate on the edges of a haphazard industrial town, IP version 6 has been carefully constructed to avoid all the problems of the old and make the most of the new. But like so many new estates, nobody seems to want to live there.
While the meltdown of IP version 4 -- the one we all use today -- has been long predicted, it has like many cities the ability to reinvent bits of itself. Its habit of fixing problems as they occur has seen IPv4 avoid catastrophe. Now, even as governments and standards bodies around the world are funding initiatives to move everyone over to the new standard, critics are increasingly coming to doubt whether IPv6 will ever justify such a change.
The history of IP is one of practicality and pragmatism mixed with idealism. First mooted in 1974, IP gradually accreted functions and standardisation and by the early 80s was standard in Unix. Unix systems made up the bulk of computers in the universities and research establishments that comprised the early Internet, and because IP was open and easy to port across to other operating systems it soon became the protocol of choice for anyone wanting to join in. One of the early design decisions was the addressing -- each packet has a 32-bit address in it that specifies its destination, and that gives 4.3 billion unique destinations. As this was bigger than the world population when the standard was fixed, it seemed like comfortable over-engineering at its best.
In the early 90s, the Internet started to grow alarmingly. IPv4 had some problems with this expansion -- in particular, the 4.3 billion addresses it could support suddenly seemed rather parsimonious. The IETF predicted that the world would run out of addresses at a 'Date of Doom': March 1994. The expected huge growth in home networks, car automation, mobile devices and connected equipment would only hasten the end, with every person on the planet potentially needing ten, 20 or more addresses in their everyday life. IP version 6 has a 128-bit address field, enough to individually address not only every person and every trouser pocket, but every proton on Earth. The last big boost to IPv6's stock price came with 3G, which was to give everyone a roaming Internet terminal -- but that excitement hasn't held up.
There are other things IPv6 does better than basic IPv4: security, broadcasting, routing and maintainability. But in each case, additions to IPv4 have covered the cracks and made it a viable alternative to deploying IPv6 -- if the address crunch doesn't happen, then IPv6 is going to be a complication, not a solution.
Reject kings, presidents and voting
Now two researchers from Japan are reacting to the Japanese government's announcement in 2000 that IPv6 was a matter of national priority. Nobuo Ikeda and HajimeYamada have produced an analysis showing that the lack of free addresses is as much a matter of hamfisted management as anything else. While they note that ICANN -- the Internet Corporation for Assigned Names and Numbers, which handles IP address allocation -- says that IPv4 addresses will be exhausted by 2008, this assumes a continued exponential growth in demand that hasn't materialised. On the other hand, the American Registry of Internet Numbers says that 1.9 billion addresses have been used, leaving 2.4 billion, with a consumption steady at 50 million a year -- based on best data, the researchers say, IPv4 will remain viable for 26 years.
But the real problem, according to the researchers, is that the allocated numbers have been barely touched. Of the 1.9 billion allocated addresses, only 130 million -- or around 3 percent -- have been used. A lot of this is down to the way IPv4 is broken down to classes of address, whereby large numbers are grouped together for ease of addressing. But now routing improvements such as the Classless Inter-Domain Routing (CIDR) protocol have increased the precision with which IP addresses can be allocated. Even with the maximum wastage caused by the limitations in CIDR, however, Ikeda and Yamada say that at least another 11 years of grace remain.
That would be hugely increased, they say, if users were encouraged to return addresses they've been allocated and are not using. That this is not happening is because the normal economic forces don't apply: if people were charged even a token amount per IP address, there would be much more care taken in the use of what is a limited resource.
Furthermore, Network Address Translation (NAT) multiplies each Internet address by around 65,000 -- already, most home network users are given one IP address and share it among multiple devices. There are limitations to this, as every NAT user knows: it's hard to manage servers from behind NAT and complex systems such as videoconferencing don't cope well. But for most uses, it's fine.
The researchers conclude that the Internet has come a long way on the famous maxim "Reject kings, presidents and voting. Believe in rough consensus and running code," and that by the time IPv6 actually becomes necessary the future will have changed so much that it may itself be inappropriate to requirements. Leave it to the users to decide, Ikeda and Yamada say.
On current form there isn't a user in the world who cares tuppence for IPv6, and that's something which should make governments, corporations and network planners pause for thought.
To have your say online click on TalkBack and go to the ZDNet forums.
Did you find this article useful?
19 out of 56 people found this useful
- Share this article:
