IPv6: No Sticks, Just Carrots

As I relayed last month, I think that the time for IPv6 has come and your organization will have to adopt IPv6 by no later than early 2011, since IPv4 addresses are slated to entirely run out by then. After re-reading the piece, though, I realize that my "move to IPv6" argument consisted solely of one big "stick," as in "carrot and stick"—and, well, sorry about that, folks. I mean, with the economy in the state it's in and in particular with the state that the IT biz is in, who the heck wanted to read that? Many of you are probably getting more than your share of sticks waved at you nowadays already a la "you guys better shape up or we'll move our IT out to the cloud," so you certainly don't need to hear that kind of stuff from me. (Actually, the answer to the boss about the cloud thing might be, "Yeah, great idea, boss. You go ahead and do that: centralize a mission-critical bit of computing and networking to a service company with a thousand other clients and see what kind of response you get when something goes wrong. I mean, your cable and phone companies always respond 'how high' when you tell them to jump, right?" Anyway, that's another column.)

This month, I decided to step back and start talking about IPv6 from a more positive point of view, with an eye to quickly dispelling some of the bad (and incorrect) stuff that I've run across on the web and from reader letters.

Carrot #1: Adopt IPv6 and you never have to subnet again.
C'mon, is there anyone out there who enjoys figuring out how many hosts can fit into a 255.255.255.252 subnet? Didn't think so. We only indulge in that particular bit of self-abuse because scarcity makes us do it. In contrast, not only do you not need to subnet in IPv6, but the rules forbid it. IPv4's addressing is 32-bit and the concomitant four-billion-address limit makes us work hard to get around it. In contrast, IPv6 is a 128-bit addressing scheme—yes, you heard that right, 128 bits. Why so many bits? Simple: so we can waste them. More bits means that the Internet authorities can hand out network blocks that are so big that they pre-define 65,536 subnets for all but the very smallest organizations (and 18 quintillion host systems for each of those subnets). Furthermore—and here's the part that'll hurt your head—there are enough IPv6 bits that they can hand out that many IPv6 addresses to virtually any organization that wants them, be it General Motors or General Wally's Army Surplus Bonanzarama. IPv6 is deliberately built with so many addresses that by design, most of them will never be used. General Motors may well have 65,536 different locations, but General Wally's probably only got one downtown store and a couple out in the 'burbs, which means—get ready, this is the part that will hurt your head—Wally will leave fallow and forever unused 65,533 of his 65,636 subnets… and that's perfectly okay. (Always remember: why so many bits? So we can waste them!)

Carrot #2: The price of IP addresses has to fall with IPv6.
Currently, I pay a monthly fee to my ISP of a few dollars per static IP address per month. It irritates me that despite the fact that the only real difference between what the ISP must do to serve me (a business customer) and what the ISP must do to serve a no-static-addresses home customer is a tiny bit of accounting, they get to charge me several times higher fees than they do to home customers. One of the reasons that they get to do that is, again, because IPv4 addresses are scarce.

So while I have no proof of this yet, consider this: The current plans for allocating blocks of IPv6 networks call for offering, again, address blocks of 64K subnets each containing 18 quintillion addresses, and IPv6 has enough space to be able to offer blocks of that size to 137 billion organizations. And oh, by the way, if your first reaction is, "Things change and we'll eventually run out of the 137 billion organizations," then I actually have some good news: The Internet authorities left enough bits unused in the 128 that they could, if necessary, crank up that 137 billion number by a factor of a thousand or so. Sure, there are fixed costs to running a network of any kind and I'm not suggesting that the cost of Internet access for businesses would drop to zero or anything like that, but when there's that much available, the scarcity argument for high prices sort of goes out the window, right?

There are lots more IPv6 carrots; so many that we could make more IPv6 coleslaw than we could eat in a millennium! For that, see you next month. Or, come join me at Windows Connections this March, where I'll be doing 75 minutes on IPv6. You don't want to miss it!