April 2003  Volume 29, Issue 3
April 2003 Issue
   Format for Printing        Send Feedback
Technology / Innovating
e-Definitions: All Jammed Up and No Place To Go
by Mark Tamminga
Searching for a hardier protocol to handle Internet traffic.

Communications Protocol
There is probably no other business that competes with the same loony intensity as the computer industry. It's so nasty and relentless that Sun Microsystems even has a "vice president of strategic litigation." Yet out of this snarling frenzy has come one of the most remarkable communal efforts in human history: the Internet.

Nearly 40 years ago, the Pentagon's Advanced Research and Projects Agency decided that the computers at key research institutions needed to talk to one another. In the process, the agency realized that the only path to success was to invent a reliable communications standard that was both simple and resilient. The scientists given the job of concocting that standard understood that, at a minimum, the disparate computers on the network would have to come to agreement on two major fronts. There would have to be a common format for the data, and there would have to be a surefire way of getting that data from one machine to another.

The Network Control Protocol that came out of that effort begat ARPANET. And ARPANET, of course, begat the Internet. Wildly innovative for its day, NCP showed the power of packet switching, which snipped the data stream into small packets, or datagrams as they were originally called. The actual data snippet was the most important part of each packet's payload, but packets also contained the data's source and target address, just as an envelope contains not only a letter (the data) but the destination and return address as well. These packets were then left to find their own way to their destination, where they were reassembled in the right order.

NCP worked just fine in the beginning. Indeed, it ran the show from 1970 until 1982. But as early as 1972, planners realized that NCP was destined to run into trouble. Other networks had begun to spring up and NCP provided no easy way for traffic to move among these new networks. NCP also turned out to be inherently slow and stood a good chance of being overwhelmed by the increasing size of ARPANET. So in 1973, Vint Cerf and Bob Kahn proposed a faster, more scalable and hardier communications protocol that would allow for the internetworking of these burgeoning and unrelated networks.

Cerf and Kahn called their idea the Transmission Control Protocol/Internet Protocol, a standard that took final form in 1978. Notice that TCP/IP is really two protocols. Building on the packet switching ideas of NCP, Cerf, Kahn and others decided it was best to split up the network communications job: TCP would take care of dicing the data into packets at the source and safely stitching them together again at the destination. IP would be responsible for addressing and moving those packets to where they were supposed to go.

Armed with this powerful new set of protocols, Cerf and other ARPANET administrators then had to convince the world to play along. On that front, it always helps to have the military on your side. Those folks don't much cotton to footdraggers. Once convinced that TCP/IP was the way to the internetworking Promised Land, the Pentagon simply forced all 400 or so ARPANET research sites and universities around the world to switch over to TCP/IP from NCP on January 1, 1983. Simple as that.

And that's when things started getting interesting. So interesting, in fact, that many mark that date as the true birth date of the Internet. TCP/IP was decentralized, public domain software. Anyone could glom on if they had the right code. These were perfect conditions for what then happened: the geometric growth and amazing emergence of the Internet we know today.

There was just one problem. IP, which takes care of moving packets around, assigns an address to all the machines attached to the Net. As is usually the case in computing, the wildest speculations in the early '80s as to how many addresses would be required are now looking woefully inadequate. The Internet, it seems, is filling up fast.

The current version of IP-IPv4-has been in operation for more than 20 years, one of the most long-lived standards in computerdom. But long-lived isn't necessarily a good thing; IP would have run out of free addresses back in 1996 if some major retooling and patchwork hadn't saved the day. But you can do only so much with duct tape. The current guessing is that we'll run out of IP addresses as early as 2008-particularly if pervasive Internet access through cell phones, fridges and toasters takes off as predicted.

The pending logjam was seen as early as 1990 and steady work has gone into creating a replacement for IPv4, dubbed IPv6. (IPv5 was used up on an obscure protocol that never really went anywhere.) IPv6 provides for a vastly expanded IP address space-enough for every proton on the planet by one account. It also pays attention to the kind of data each packet contains. Audio and video data can thus be given a higher priority over less time-sensitive data like e-mail traffic.

IPv6 is a much-needed improvement to the infrastructure of the Internet. But here's the rub: Almost all the equipment now used to run the Net would have to be replaced or reconfigured to deal with the new protocol. With millions and millions of devices now chattering happily away in IPv4, this won't be easy. Moving to IPv6 will be a tortured process that will drag on for at least a decade, with IPv4 and IPv6 operating in tandem the whole time.

As the old joke in software circles puts it: "In six days God created the world. But that's because there was no installed base."

Mark Tamminga ( mark.tamminga@gowlings.com) practices law and fiddles with software at Gowling Lafleur Henderson LLP in Toronto. He is the coauthor of the new ABA book The Lawyer's Guide to Extranets.

Excerpted from an upcoming book to be published by the ABA Law Practice Management Section.