From...

September 21, 1999
Web posted at: 12:29 p.m. EDT (1629 GMT)

by Mark Frauenfelder

ALSO: The next best thing to IPv6?

(IDG) -- Bank panics. blackouts. food riots. doomsayers haven't had so much fun since, well, since Y1K. So what will they do when Y2K has come and gone? Most likely they'll set their sights on technology's next day of judgment: the Great IP Crunch of 2010. That's the year the Internet Engineering Task Force predicts that every available Internet Protocol address will be taken, making new connections to the Net impossible.

How could such a thing happen? Before you accuse the Internet's creators of shortsightedness, recall that the Arpanet (the predecessor of the Internet) was designed simply to help researchers at U.S. universities share precious computing resources, not to provide every PC, cell phone and toaster oven on the planet with its own node. How were Vint Cerf, Jon Postel and Danny Cohen supposed to know that the system they sketched out on a piece of cardboard would one day be carrying everything from international currency transactions to Marilyn Manson videos? It's a testament to their genius that the Internet Protocol they designed 20 years ago still holds up under the billions of bits sent today.

But there's only so much the Internet can take. Experts say that by 2010 the current Internet Protocol, IPv4, will reach the end of its tether.

First, some background. IP is the set of rules the Internet uses to send data from one computer to another. IPv4, the current Internet Protocol, uses 32-bit addresses. Problem is, you can juggle 32 bits into only 4 billion or so combinations. That means only 4 billion or so addresses, which means only 4 billion or so computers on the Net. Period.

Of course, back in the old days, 4 billion addresses seemed like an inexhaustible supply. When a company needed addresses for its network, the Internet Assigned Numbers Authority often gave it far more than it actually needed. Companies like IBM (IBM) , AT&T; and MIT each received 17 million IP addresses. Only in the past several years has it become clear that the addresses are running out – about half of them have already been allocated. The Internet Assigned Numbers Authority is now a lot more careful in assigning IP addresses but the end is in sight. What happens then?

MORE COMPUTING INTELLIGENCE
	IDG.net home page
Industry Standard home page
Forecast 1999 | The next 10 years
QuickStudy: Internet Protocol Version 6 links
Windows 98 looms, and Y2K too, but the question is: What happened to IPv6?
Reviews & in-depth info at IDG.net
Industry Standard email newsletters
Industry Standard daily Media Grok
Industry Standard financial news
IDG.net's personal news page
Year 2000 World
Questions about computers? Let IDG.net's editors help you
Subscribe to IDG.net's free daily newsletter for computer industry cognoscenti
Search IDG.net in 12 languages
News Radio
	Fusion audio primers
	Computerworld Minute

Enter IPv6. (IPv5 was a short-lived experimental protocol.) IPv6 is the proposed replacement for IPv4. It has been in development by the Internet Engineering Task Force for about four years. IPv6 offers many improvements over IPv4, most notably its much larger number of potential addresses.

How many addresses can you get from IPv6's 128-bit header? Here's one way to look at it: Imagine 1 trillion Bill Gateses standing in a circle (not a pretty picture, but play along for a moment). Now ask each one to convert his fortune into pennies and toss them in a collective pile. If each penny contained 1 trillion tiny computers and each computer had its own IP address, you'd still have used only a fraction of IPv6's potential space.

There's more – IPv6 offers things like quality of service, encryption and authentication, all of which were welcome improvements to IPv4 four years ago. However, these days a number of IPv4 add-on applications do a fine job of making up for many of the protocol's shortcomings. Essentially, there are now only three reasons to switch to IPv6: more addresses, better routing and autoconfiguration.

First, autoconfiguration. Whenever a company adds a new computer to its network, that computer has to be set up with an IP address, DNS server, default router and so forth. All of this information has to be manually entered either by the user, by a system administrator or through the use of DHCP (Dynamic Host Configuration Protocol), which automates much of the assignment procedure but requires its own maintenance. In these days of massive company mergers, it's not uncommon for a company to switch ISPs, which requires that every node be reassigned a new IP address. With IPv6, getting on the network is as simple as plugging a cable into your computer's Ethernet port.

The second real advantage IPv6 has over IPv4 is the way it handles routing – how data packets hop from node to node as they travel to their final destinations. A router controls traffic on a network or at an ISP, and it must maintain a table of all available routes so it can send data packets on their way. As more organizations come online, the routing tables grow.

"[IPv4] allocates and manages routing tables in ways that don't scale well with the Internet's growth," says Richard Draves, who leads the Systems and Networking research group at Microsoft (MSFT) Research. IPv6, on the other hand, hierarchically assigns addresses, which makes routing more scalable.

But the most urgent reason to move to IPv6 is still the address problem. With addresses in short supply, many companies can't get enough to assign to all the computers on their network. When a firm has more computers than IP addresses, it uses a NAT (Network Address Translation) box to create a temporary connection between a computer and the Internet. NAT systems are currently in widespread use around the world. According to Martin McNealis, IP product-line manager at Cisco (CSCO) 's IOS Product Marketing division, NAT boxes are actually preferred over IPv6 by most Cisco customers. Why? For one thing, everyone's still using IPv4. "There's an old maxim that says, 'Things that work persist,' which is why there's still Cobol floating around," says Vint Cerf, who created TCP/IP with Robert Kahn in the 1970s and is now a senior VP at MCI WorldCom.

Clearly, switching over to IPv6 is not simple, especially for large sites with custom Internet software written for 32-bit IP addresses. Steve Deering, lead designer of IPv6, compares the problem of upgrading that software to the Y2K problem. But instead of changing two-digit years into four-digit years, programmers have to sniff out 32-bit address fields and change them to 128-bit fields. It's a messy job, and another reason why many companies are dragging their feet in the switch to IPv6.

But all things considered, Cerf and Deering say it's well worth the trouble. Cerf is almost religious in his belief: "Some of us feel NAT boxes are sort of an abomination because they really do mess about with the basic protocol architecture of the Internet."

The more NAT boxes that are patched into the Internet, the less the Internet behaves as it was designed. NAT boxes tremendously complicate the system. "They introduce all sorts of problems into the Net," says Deering, who also cochairs the IETF's Internet Protocol Next Generation working group and is a technical leader at Cisco Systems. Imagine, Deering says, having to go down to the post office to borrow a street address each time you wanted to send or receive mail. NAT should be viewed only as a stopgap, Draves says. "If the Internet came to rely on NAT fundamentally, the Internet would end up being much less maintainable overall and much more fragile."

To help push IPv6 forward, the Internet Assigned Numbers Authority on July 19 gave regional registries around the world the go-ahead to begin assigning numbers based on the new standard. So far, the majority of the organizations that requested IPv6 numbers are research departments and universities. The only commercial ISPs to ask for such address blocks are two Japanese firms, Internet Initiative Japan and Nippon Telegraph & Telephone. ISPs outside the U.S. may be more keen on switching to IPv6 than those in the U.S. because they've had a tough time getting enough IPv4 addresses. But if the number of mobile Internet devices takes off, more U.S. commercial companies may start clamoring for IPv6 addresses. NAT doesn't work well with mobile devices, and if the world is flooded with hundreds of millions of handheld Web browsers, there'll be little choice but to switch to IPv6.

Of course, if everyone waits until the last minute, it could end up costing much more not just to engineer the transition, but in the cost of the disruption to what has become a crucial part of our economic and social infrastructure. At a meeting in July to launch the IPv6 Forum, a pro-IPv6 consortium of service providers and carriers, President Latif Ladid issued a warning: "Y2K will be peanuts compared with moving from IPv4," he said. "If nothing is done, in 10 years' time it will be 10 times more expensive."

Asking the Experts

Steve Deering
Lead designer of IPv6 (and self-professed pessimist)

A: "It's quite possible it won't happen. It's conceivable that we will just continue to do short-term hacks and band-aid whatever is required to keep living with IPv4. I think that would be very unfortunate for the health of the Internet."

Vint Cerf
Cocreator of TCP/IP

A: "At some point there will be a critical mass of v6 users, and there will be a lot of pressure for everybody to become compatible with that. "I imagine we may not ever be entirely free of v4 but what may happen is that the NAT boxes will be pressed into service to allow something to use a v4 address over a v6 protocol, as opposed to the inverse."

Martin McNealis
IP product line manager, Cisco Systems

A: "At this point it is still very much 'to be decided.' It depends on who you talk to. It's by no means a given that IPv6 will be the next-generation Internet infrastructure anytime soon."

Rich Draves
Systems and Networking research group leader, Microsoft Research

A: "I would say I'm pretty confident, but not extremely confident [that we'll migrate to IPv6]. There's a possibility that we won't. I think that would be an unfortunate outcome; the Net would end up becoming more fragile and less functional."