Keynote: Geoff Huston

Hi, thank you and good morning. Um it's good of all of you to come out here on Australia Day. Um and I hope the sausage sizzle goes really, really well. I hear Vince Serve came and talked to you yesterday and is a really good guy and an excellent speaker and a professional optimist. I'm not. So, what I want to talk to you today about is the slightly uglier side to all of this. Um you know, when you look at open systems, I'm blowed if I know why this thing doesn't quite display. In the 1970s, there were really two foundational technologies that I think started a movement around openness. Um both were in some ways, I think, almost accidental. UNIX, as you well aware, was actually a byproduct of an antitrust regime at AT&T. But literally having invented this really cool operating system in their labs, it was just impossible for that company to commercialize it. So, in some ways, they just simply gave it to the community and look what happened. Astonishing success. Similarly, another technology appeared in the 1970s and that was packet switching. Packet switching really was revolutionary because previous to that, we'd thought about communication in really sort of two ways. One was writing it on a piece of paper and sending it through the postal system and the other way was talking to people, which required required real time. So, when we built electronic networks to handle people talking, oddly enough, what we tried to do was to keep time constant so that this entire switching system that we built called the telephone network preserved time all the way through. And it was this revolutionary cut through that happened with packet switching which sort of thought, you know, I actually don't need to preserve time when it's computers. I don't care when you get the next packet as long as you get it at some point, that's cool. And up came a revolutionary approach to networking that instead of switching time switch packets. It was cheap. It worked but uh fundamentally there was something very strange about that technology. Although a lot of commercial companies tried hard to make things like that work what really made a difference what fundamentally changed networking, oddly enough, was a contract from the US Defense Advanced Research Project Agency, DARPA who contracted the good folk at Berkeley to write an open source implementation of this newfangled networking protocol TCP/IP. Revolutionized the world because all of a sudden that single open technology just wiped out everything else on the planet. We're now actually in a monoculture in networking. There is only TCP/IP. And even the residuals of the voice switching network, you keep on hearing Telcos all the day going, "Oh, we're giving up the voice network. We're going to an all IP network. Aren't we wonderful and sexy and everything else." You know, but openness is really good. Interestingly from a world of vendor technologies in the 1970s and 80s, where if you had stuff from digital, you had everything from digital. Digital cables, digital plugs, digital computers, digital, you know, digital technology. If you had stuff from IBM, it was all from IBM. And companies, enterprises, government agencies were locked against vendor. That all fell apart in the and it fell apart predominantly because of this pressure from openness. Folk wanted to compete. Folk wanted to bring in their product and plug replace stuff from vendors, other vendors. And open technology served that brilliantly, don't they? I'm like they are accessible. They are neutral in terms of competition. Anyone can build stuff against it. Because it's open, you can make it better. It's functionally extensible. And of course commercials can use it too. It is exploitable. So, you know, the internet itself didn't become what it is today because IP was good. As a technology, IPv4 was about as mediocre as any other technology at the time. Except AppleTalk, that sucked. You know, anyone who goes could have done it, right? And and it wasn't that it was IP, it was the fact that there was this brilliant open implementation that you could put even on a toaster, as Simon Hackett did from Adelaide, you know. It was so good and so widely used that even the bugs in that implementation are in current stacks today. You know, everyone uses that one block of code. So, you know, why are we here? So, what, you know? I'm not here to sing from the open book. I'm like you can all do that and you all know why you're here and you'll know why openness is good and that's fine. But my real proposition to you is that it's not that open technologies are good and you know, that's fine. It's staying open that's the problem. Because when you get people building on it, when you get extensibility, their first reaction is that bit's mine. You can't use it. And all of a sudden you start putting ring fences around that technology. So, staying open is really hard. And it's really, really hard for the internet. And you're about to see another fight of titans. Because we're not sure that an open internet will still be around in 5 years time. What is going on behind that proposition? Why is this such a problem? Well, you know, very few useful technologies are static. Only the old ones, right? The ones you don't care about and get, you know, get into sort of stasis. Because technologies do evolve, uses change, exploitation models change. No one when they thought about shipping packets ever thought it would happen in the air. You know, no one ever dreamt that the iPhone was going to be part of some what some weird future. No one ever thought, by the way, that there'd be more mobile users of the internet than not in a couple of years time. That was never part of what we ever conceived of the internet. And if anyone's ever tried to make TCP/IP work fast over wireless, you'll know what I'm talking about. It's crap. Um speak to the NBN folk. Um So, the issue is that in trying to make things stay open, there's this constant sort of war of demarcation. Because folk who invest money, the competitive interests, want to pull that into their realm of private ownership. And there's this other part, the openness bit, the common interest and public good. And we've seen those particular debates happen recently in a whole bunch of places. In the US in particular, that first one, net neutrality, is a big thing. This idea that, you know, well, I'm in control of the packets. TV on on IPTV, good. BitTorrent, evil. BitTorrent packets won't get through. My packets will. And even yes, I like Google, I don't like being. I like my you know, trying to bias the network and no longer being neutral for all players is certainly part of the problem here. Carriers have the technology to basically respond to different kinds of packets in different kind of ways. What should they do? Trying to make sure that our networks are neutral. Trying to make sure that every application, even yours, still works on tomorrow's internet is really a difficult problem sometimes. And of course there's this whole thing about next generation networks. Trying to make networks sing and dance and do everything else all using proprietary technologies. And you've seen this whole issue with mobility and mobile services. And of course even this whole triple play quad play trying to pack IPTV down down your your wire. All of those things typically rely on a technology base that's not open. A technology base that is full of patents. Full also of occluded source code. Nothing is available. You're trying to reverse engineer things without truly knowing what was going on. So all of those challenge our concept of openness in networking. But you know, above all of those, let's just focus a little bit. What's the one thing that's really important? Um if you don't have open addressing, you don't have a network. If I don't know your address, I can't send you a packet. It's as simple as that. If you can't get addresses, you can't join the network, you can't be part of this community. We need open addresses and we've just run out of them. Which is a bit of a bummer. Um we're a victim of success. Every single internet graph, I make most of these technology graphs up and to the right, big curve, isn't that wonderful? Let's put some numbers on it just to sort of look at this. There are 4 billion addresses in the V4 addressing scheme, 4,000 million. Right? So, line in 2009, we gave out about 5% of those addresses in just 1 year. 189.6 million addresses. So, down there at the address giving out area, we were working really, really hard. Uh but last year we decided to work really, really harder and we gave out 248 million addresses. Now, if you equate an address with a customer, that's another quarter of a billion people on the network. In actual fact, because of NATs and everything else, normally that's 1 2 3 4 people behind each address. So, in other words, there are around oh, 300 million new things on the network. What were they? Phones. Bloody phones. In Australia alone, we gave out 9.6 million addresses last year. We only have 17 million people living on this island. You know, bloody phones. Um the revenge of the phone company. And in actual fact, this is a sort of a daily graph, the sort of how much each day, smoothed out a bit. And you can see in the northern hemisphere folk go on holidays a bit in the middle of the year and you know, they work on either side. But you notice 2009 sort of trundle along, 2010, can anyone spell panic? Cuz, you know, life is getting rough. Um this is a sort of a stock pile of where we have in the factory of addresses, we hand out addresses in lots of 17 million. It's a binary thing, 2 to the power 8, right? Or sorry, 2 to the power 24. And we count addresses using these blocks of what we call {slash} eights. In V4, there are 256 of these blocks. Uh the dear old standards folk have said 35 are ours, you can't touch them. Never, ever, ever. Fine. That leaves a few left. We have handed out 213 of these blocks. There are seven left. And we're getting through them at a rate of one block a month. So, you know, what are you going to do about this? Well, we had this wonderful plan cuz we knew this was going to happen in 1990. Cuz even before the internet was even a twinkle, you know, of commercial success, even before everyone started taking VIN seriously, it was pretty clear that once we'd entered into the revolution of a computer on everyone's desk, let alone in everyone's pocket, the address space wouldn't last. It would come to a crunch. And the predictions were sometime between 1998 and seemed like a good enough idea, we'd better do something. So, we had this really good idea that what we'd do is we'd create a new protocol. And we'd call this protocol IPv6 cuz someone had used five and blah, blah, blah. But, the idea was cuz we're all great at engineering, that we'd never actually run out of addresses. Because we're going to plan. So, as the internet grows, big green line going up to the top on the right, and as the available pool of V4 addresses, blue line, starts to plummet earthward, all you guys would be so well attuned to the environment, you know, and attuned to the situation, you'd start running V6. So, that by the time we got to, oh my god, we're getting a bit low, we'd all go, so what? We're all running V6. Now, I worked for Telstra for 10 years, much to my shame. Um it was a really enlightening experience though, cuz when I joined, I joined in about oh 1995, they told me about the year 2000 committee. Cuz we're all in the Y2K business. Oh, well, it's over. No, no, no, it's not a problem. We started our year 2000 committee in 1987. We're been planning. You know, these are the guys that look at disaster and you plan decades in ahead for it. So, we thought this was a really good idea. You know, this is an industry that thinks ahead. So, as long as we sort of put this V6 out there, industry would go, oh yes, not a problem. We've got a thing with V4, we better start using up V6, you know, moving into V6. So, you know, you'd all act rationally. So, a few years ago we started worrying about this problem. So, we started looking very, very closely about what industry was actually doing. And we started looking really carefully about how addresses are being used. Um so, these are a few plots. You actually We've actually been plotting the data every hour for years. And you know, you do some maths and you do some projections and you start to look at this and I've been doing it for some time and I can tell you right now that IANA is going to run out of addresses in February and that's no surprise. And it's not going to be the end of February. And it's not even going to be the middle of February. IANA is going to run out of addresses in February and that's going to be pretty, pretty clear cuz we're right, right down at the bottom. And it's also pretty clear that there are five regional internet registries and one of them is here in the Asia Pacific. And it's pretty clear that the rate we're giving out addresses in the Asia Pacific, someone's going to come knocking on our door in about July and saying, I want a gazillion addresses and the answer will be, no, we've actually got none. None. Go figure. So, the real answer was and the answer we're going to tell you in July is, "Well, you should have been running V6." So, the next kind of question, which is a really cool question, is "How many folk run V6?" Put up your hand. You're weird. Because if you're like the rest of the world out there, only one person should have put their hand up. Because when Google do this measurement and look at the folk coming to Google, the number of folk that come to their wonderful dual stack Googley thing is 0.3% Not 1%, not 3%, 0.3% three in a thousand. So, a couple of people can put your hand up, you know, fine. That's where we are. And look at that graph. Is that an up and to the right graph? Hmm. No one's running it. So, you know, something is really, really not right in all this. That you know, this great plan we had is now looking pretty bloody weird. SO, DO YOU LIKE THE TRAIN? UM SO, SO THE real test is, you know, is this plan feasible that we can dodge the bullet? That somehow we can go from 0.3% to 100% in 7 months. So, let's just sort of examine exactly what I've said here. So, we're going to deploy IPv6 across currently 1.8 billion people. Multiple devices, multiple things. 1.8 billion people, okay? More than a billion in hosts. How many routers? Hundreds of millions. How many firewalls? How many pieces of middleware that don't work properly? And you guys have been so busy writing all those stupid filters and firewall lists, all that configuration code, all of that code you've been doing that's V4 only. You have to change all that. Every last piece of code. And we better audit it and all those support systems and all the paraphernalia that makes the internet work all works on V4 only. You've got 200 days. Why are you here? Go. I maintain my Yeah, right. Who actually really runs V6? Who's got the white screen of death on their browser about every day? Because sometimes when you run V6, really happens badly because you're doing fine and the other end has no clue. And the TCP connection just goes gack. You don't even fall back to V4, just goes gack, white screen. Sit there for hours, look at everything. Have I plugged it in? Have I turned it on and off again? Do it again. White screen. This is really frustrating. Who also you guys have been running it? How long does it take you to get to some websites? 35 seconds is pretty typical in some cases. Why? Try V6. Hmm, timeout. Hmm, sends another packet in V6. Hmm, the guy had a quad A record, it should be coming in V6. Hmm, hmm, what's the time? Should I make a cup of tea? Oh, bugger it, I'll try V4. Whoops, there's the page. Why do you bother turning it on? As a user experience, it sucks. So, you know, when you look at all this big plan, you're going to make all this work in 200 days to make the network suck. Something is really weird. So, you know, when you look at the real transition plan we're about to head to, it's even more broken than that. We have no clue. And we have no clue in about 7 months time. We just have no idea. Because over in the factory that makes iPhones, they're busy saying that Apple made $6 of profit last year and they're going to make $68 trillion of profit next year. They're not cutting down on their production. They're busy pumping out chips like you wouldn't believe because as far as they're concerned, yep, 250 million new addresses next year, not a problem. I'm just pumping this stuff out there. So, how are we going to do it? V6 is not the answer. Not in that time frame. You're going to need a lot more time than 7 months to make the next V6 only device that only talks V6 a useful member of the network. It won't happen in 7 months time from now. So, somehow you've got to do something else. You've got to persist in using network address translators. You know, those things that somehow you bought out there, your ISP gives you only one address and you share it amongst all your machines at home. Yeah, NATs? Does it work? Yeah, cool. Cool. Let's test this theory. Because we're going to go into the when doesn't it work mode. Uh because the frightening fact is it won't. So, you know, right now NATs are really, really cool. The thing is about NATs is most of the time from an internet service provider's perspective, you bought them, you paid for them, I don't care. From the ISP's perspective, there is no such thing as address shortfall or address scarcity right now. You guys have a problem, I don't. So, they honestly don't give a stuff about that. And you know, most of the time most of the stuff works. And even then when you go, "Oh, I don't work properly. I need a real address." Nice people say, "Oh, sure. Pay me 10 bucks a month and I'll give you a real one." So, out there in consumer land, oddly enough, you know, addresses already have a price. But, when you add more NATs, all of a sudden the ISP has to start putting NATs in. Thank you. Uh it is a bit evil. In fact, it's really evil. Um it's evil for a number of things. One, when they put in a NAT, it's not going to be a little NAT, and it can't afford to fail. It's going to have to be a rather big NAT, and it cannot afford to fail. Which means it's going to have to be an expensive NAT. Who's going to pay for it? Look to yourself. Yeah, you're going to pay for this stuff. Yeah? How well is it going to work? Hmm, interesting. Um at the moment, a lot of you guys write applications. Yes? And you sit there and go, "You know, is there a NAT in the path? If I send out one of these packets and get back one of those packets and send one of these packets and do this thing, I can work out what kind of NAT I'm behind and get around the problem." Yeah? You do. Whole bunch of traffic like that. What happens when the carrier puts a NAT in? What about two? Or maybe three. Just to make life interesting. And because NATs don't have a standard, because the IETF got so snotty that standardizing NATs was beneath their dignity, everybody who writes a NAT writes it creatively. Every single NAT behaves differently, and even NATs from the same vendor change across models. As an application designer, you're stuffed. And when carriers start deploying it, things get really, really, really ugly. Because the more intensive use of NATs is going to start breaking things like crazy. The aperture through which your application looks at the internet is slowly shrinking. But at the same time, applications are now discovering parallelism. How many web pages only have one element on them these days? Oh, How many have 500? All of them. How does your browser cope? Well, I just launched 500 TCP sessions at once. I've got all the ports I need. Today. But when you have a carrier grade NAT, that's precisely what it's rationing out and sharing, port space. We're now eating into those next 16 bits of UDP and TCP port space. And they're finite. And they're really small. And all of a sudden that aperture through which you look through won't work. How many ports do you need per customer? The answer isn't one. The answer's a lot closer to a few thousand would be good. But if we want to put in a quarter of a billion more new users next year, and the year after, and the year after, and the year after that, one looks like a frighteningly soon prospect for NATs. We can't make that work. Cuz all of a sudden, everything you've assumed about NAT behavior won't. So, if you think that well, I don't need to do V6, I can just do NATs. The network will still work. NATs are just fine. Which is what I thought I heard 3 minutes ago. Yes? You're wrong. They can't scale that far. We can't make a network at say 5 to 10 times the current size of the network using NAT. IPv6 is almost your only choice. Isn't that cool? I believe it was Russian. I never know if it worked or not, but it's so cool. Um When I you know, when you look at this stuff, the first thing I read was the engineering plan for V6. Because being good engineers, we wrote down our requirements. And there's this long piece of crap you can read it somewhere. It's in an RFC. But you got to grab You got to do a search for two key critical words in this order. Backwards compatibility. Cuz it was in the spec. V6 was meant to be backwards compatible. I speak V6, you speak V4, we can have a fine old conversation. Nah. Nah. Nah. Nah. Doesn't work like that. I speak Swahili, you don't, we're stuffed. I speak V6, you speak V4, we cannot talk. So if I'm speaking V6, you have to speak V6, too. How many people spoke V6 again? Put your hands up. Well, what about the you know, there's a problem going on here, isn't there? Not all of you speak V6. So if I only spoke V6, I can only talk to you few people. Rest of you leave. I'm sorry. You can't understand me. Well, if that doesn't work, I've got to speak V4 as well. So all of a sudden, because it's not backward compatible, V6 isn't a replacement for V4. It's an and. So during this transition that we're about to go through, I still need V4 addresses to fuel an internet that grows at at least 250 million new people a year. And what did we say about NAT? They don't work. So do you start to see the beauty of the problem? You know, it's brilliant. So, we had this wonderful plan. You know, when you put it in PowerPoint, it all looks so plausible. Um you have these bits of V6 at the edge. That's true, you know. I can even do words and sing and dance. You know, you know, it's just wonderful. You know, you do these V6 islands. You all run What was it from Microsoft? 6 to 4? Anyone run 6 to 4? Don't. Do you know what the failure rate is for 6 to 4 on connects that I see at the server end? 15%. 15% of all folk running 6 to 4 can't complete the connection and fall back to V4. So, if you're running 6 to 4, what's the only thing you get? A really, really slow internet. Turn it off. Okay. Um so, this stuff got to The idea was that it was meant to sort of then promulgate into the ISP land that all of a sudden these isolated islands would start doing bridges and we'd run this sort of strange hybridization world and, you know, in the end goodness and light would prevail and the world would be wonderfully V6 and V4 would be a legacy. So, the only problem with that plan is that economically it's rubbish. So, no one ever did it. And now it's too late, so it really won't work. So, the combination is really, really bad. Um the idea was that that only works if you have enough V4 addresses to last through the process. The other model is the transition would only take a small number of years. I said that those 1.8 billion users and those millions of lines of code that were done by contractors years ago wouldn't be done in 200 days. That the world would not be ready for V6 in 200 days. Fine. How long is it going to take you guys? Geez, you're optimistic. Honestly, there is so much even the last 20 years, we have been so bloody busy building and just putting it out there and then sealing it up and moving on. There is so much infrastructure, so many weird rules. I don't know about you, but where I was at one particular location, we turned on V6 and I found this was bizarre. All of a sudden, there was no more firewall, no more filters. Every single port on my machine worked. It was Nirvana. And the real answer from the local sysadmin was, "Well, you know, no one's ever used it before. You know, it's just open." When you turn on V6 as an ISP and you go to a customer and the customer was running a NAT in V4, what are they running in V6? Nothing. When you turn on V6, how open is the network? Completely. You know, all of these little issues that will start to fly up. So, all of those assumptions about V4 really get down to the issue that this entire transition, which will take Some economists have said 70 years, which I think is but, you know, five strikes me as equally insane. So, somewhere around 10 years to undo the last 20 years. We need V4 addresses and that ain't going to happen. So, dual stack isn't, "Well, I'm going to run V6. Let's turn off V4 tomorrow." Dual stack is actually double the fun, double the cost, and double the pain and no one knows for how long. Right? You've got to actually have both. So, we're going to have to stretch out IPV4 forever. Now, this is tough. This is tough. Um because we don't have any idea how to do this. Australia wanted 9 million addresses in this last year for all these new iPhones. What if Apple go and invent something in between the iPad and the iPhone or something, and all of a sudden we need three devices, then we need four, we need one for every car. Where are you going to get the addresses from? If they want to talk to the rest of the network, how's that all going to work? Because quite frankly, when the pool runs out, you're going to have to start looking at each other. So, it's pretty clear we're about to do something we've never done before. Actually, we did do it once. We did do it once. Anyone remember the spectrum auctions? How many companies bid so much they sent themselves bankrupt? Oh, the Dutch company KPN. How good are these carriers at market economics? Oddly enough, really crap. They're great marketing firms, but in terms of actually understanding how markets work and their supply lines, they're remarkably weirdly bad. So, what are we about to do to the V4 address space? Well, if we're not going to start giving them out, cuz we've given them out already, where are you going to get your next address from? Look to your left and look to your right. How much is it going to cost? Won't be cheap. Scarcity has its own price. So, all of a sudden, IPv4 addresses are going to cost a lot. And that's going to be weird. Um So, I've seen a whole bunch of folk with a whole bunch of ideas, and I find it strange cuz some of them are sort of mutually contradictory. Um there's a whole bunch of folk saying, "Do more NATs. Deploy more NATs." and a whole bunch of other folk saying, "You know, this is really daft. NATs only just work now. If we start ganging them up and layering NAT upon NAT upon NAT, most applications that currently try and slide their way through NATs won't. They just won't work. Or we could use a larger private address pool, start privatizing more of the public space. I always find that a bit weird, but some carriers, oddly enough, are really keen on that. Wonder why? Hmm. Um there's ideas of rationing. Uh there's idea that we're all be very public-spirited and we'll give back our last before addresses rather than sell them. Um yeah, right. Um there's this idea that somehow we'll magically figure out how to regulate address transfers or another group that wants to deregulate them. Some folk want markets, some folk don't want markets. But the real question or the issue is in all this, we have no idea what future we're building into. Which for an industry that has completely given up all other forms of networking culture, TCP/IP is now a monoculture. That's all we have. There's nothing else. And within a year, we have no idea where we're going. So this entire world's production and communications is now sitting staring in an abyss where we have no particular idea where we're going. Because when we try and make V4 stretch across this transition, we really don't know for how long we need to do that. We've got no idea where the sort of the cumulative end point is, how long we have to make it work, for how long do we long do we need to hold our breath, who gets to win, who gets to lose. And then there's this other minor problem. This is a deregulated industry. There is nobody in control. So the real question is, as we explore some very, very hinky areas of technology, will we ever get to V6? Because if we have to create a network that really is not V4 or V6 in this process, will we ever come back to openness again? Will we ever come back to an open network? Or will we end up during this transition privatizing a huge amount of the open space and never resuming its openness again. Some kind of curiously struck with that question because that question seems at the heart of what openness was all about. That if you have to make some temporary measures that sort of fix over the problem because you've got this good place V4 and you've got this good place V6, but you have no idea how to get between the two. And then you sort of start asking, "How do we wedge ourselves on this? How have we come so close to this precipice? That having created a monoculture of TCP/IP, we're now about to drive a sort of dramatic stake through it and go, 'Geez, that's a problem.'" Um I suspect there are a number of really fundamental reasons that are quite frighteningly bad. Um The one group of folk that lost over the last 10 years and their share price just shows it. In Australia, it was Telstra. It lost. Telstra was part of a monopoly cartel across the globe that controlled communications. They owned the handset. They owned the wire. They owned the vendors. They owned the technology. They owned you. Completely. They had complete control vertically over the entire market. Over the last 20 years, they've lost everything. And it wasn't because we deregulated telephony. It was because of computers and the internet. It was because you guys started moving packets around. And they lost everything. Because now we're moving voice through IP as you well know. And all of a sudden, the rich rich rewards of being a monopoly just disappeared. And now, who are you asking to spend money to build V6? Those same people. Are they interested in spending more of their money to create an even gloomier future for themselves? Well, the evidence is no. How many products does your local carrier have commercially that offer V6 today? And that's not an observation about Australia. Almost every country in the world. LT We'll get into that. It's not what you think. But at the moment commercially there's just nothing out there because these guys aren't motivated in creating an open network because for the last 10 years the only thing that openness has done is ripped apart their control of all parts of this industry. Openness has never been in their interest as a carrier. So they see all the cost and none of the benefit. They don't want to spend more. Geez, what about Google and Amazon and eBay? Surely they'd be interested, wouldn't they? No. Why not? Because they're incumbents. And being incumbents, oddly enough, spending money to make it easier for more competition is not a strongly competitively good move. So oddly enough, pandering to the existing base and making sure if you will that you actually don't create a field that's more open to future competition is actually not that bad competitively. So right now the folk who have benefited from complete openness in the past are actually a little bit ambivalent about the future. Because in some ways closed networks where you have to pay to enter are well within the financial reach of incumbents. They can afford it. And if what that does is increase the barriers to entry to new competitors, that's a good thing for them. Not for you. So now, those incumbents are part of, if you will, a field that sort of says, "I'm not that interested in openness per se above all else." So, what about consumers? So, okay. Here's the proposition to you. You're all customers of me today, aren't you, lucky? Now, you can have a dual stack network as of tomorrow, and it'll all cost you 20 bucks a month more. It's the same old internet. Exactly the same. Everything else is the same. All the apps are the same. I'm going to give you V6, but you're all going to pay 20 bucks more, or you can go and join another ISP. What are you going to do? You're all going to leave me. I'm sorry. You know, it's been a great idea, and you've been wonderful customers, but you're going to vote with your wallet. And short-term gain always trumps long-term common interest. So, even consumers aren't motivated to spend the money to allow the ISP to do this. So, this whole cost of transition becomes a real problem. Cost and benefit don't align. And so, that leads me to a rather depressing view. That the longer we toy with disaster, and we are toying with disaster, the real loser is openness. Because if anyone remembers the 1980s, it's really easy to build a network as an amalgam of closed walled gardens. It's just you can't get to there, you can get to my content. You can't get to their content unless you're a customer of them. It's really easy to build a very, very cheap new network starting using V4 with address one all over again. It's extraordinarily cheap. It is really expensive to build a network in six. So, in some ways it's not really obvious to me that we're going to continue with a truly open network infrastructure because all of the economic pointers tend to point to a far more gloomy, far more depressing world. I don't like it. And I hope that you don't like it either. Cuz I think that's the worst possible outcome we could possibly envisage. It's just terrible. But the problem and the challenge I'm going to leave you with cuz I'm not an optimist is that we have to fix this and we have to fix this by altering our environment. Folk like Telstra, folk like Apple, folk like Google, folk like eBay need to see motivations as to why they should continue with open networking. Why V6 is in everybody's interest including theirs. The fact they haven't so far is a failing of mine. But I'm very sorry to say this, give you this problem to you because somehow you really need to prevent a multi-trillion dollar industry falling flat on its ass. Thank you. Thank you, Jeff. We have time for a few questions. I'll have a couple of people I guess start over here. For the final IANA allocation, you said early February. From things various people have said, there is actually a day. Can you tell us the day? I'd love to be able to tell you that. No, I unfortunately I can't. Um, I I would, however, advise that if you start looking at the schedule of meetings over the next few weeks and try and find where meetings are on this planet, you might actually have a decent clue as to when, but you know, literally, we have reached that point and and things will be happening in the next few weeks, obviously. Um it's it's all over. It was nice while it lasted. Can you tell us a bit more about these unadvertised addresses? Um we've always said that when we handed out addresses, you didn't actually have to use it on the internet. And well, you didn't. You just If you needed addresses, private or public, you know, you could do so. So, there were allocations made into whole bunches of places, including, say, universities, where at some point they used it and numbered all their machinery and then found themselves with NATs at the edge. So, the addresses that they used internally are no longer being advertised. Now, there are some awfully big allocations that are very, very unadvertised and quiet. And if you look very hard, you might find some into the US defense establishment, for example, where there's a massive amount of address space is allocated, but you've never seen it openly. But equally, if you look in corporate networks and similar, you'll start to see other, you know, networks that are private. Some of that stuff might reappear in a market. Because once addresses start getting a price, some folk might be motivated that the cost of renumbering or the cost of whatever they need to do is low enough that, you know, they might sell some addresses. This may happen, and you know, some addresses may become available, but we chewed through an awful lot of address space last year, an awful lot, 250 million, 5% of the V4 address space. The unadvertised pool is only 20%. So, even if you flushed all of that out, you won't buy yourself much time. And my suspicion is that the price of some of those addresses will be eye-watering. I I remember sitting in a talk that you gave. I think it was at Aug 1998 in Sydney. I was an optimist then. And the words that you used, I believe, I'm not quoting exactly, my memory is a bit hazy, but it was IPv6 will never happen. Um and and and true at the time, um we were getting through four {slash} eights a year. And the pool of a V4 address space was going to last us until about 2040. And it just didn't seem reasonable to sort of jump into one technology and then have to do another. So, it didn't look like that would happen. But in 1998, no one ever dreamt of an iPhone. You know, the technology and the exploitation models continuously change. So, yeah, um no one, I think, has perfect vision in all this. And yes, we've now reached an entirely different future. We've run out of V4. So, would would you say that it's the uptake of mobile that's the main thing that's changed? And I guess, not to heckle, but isn't it sort of your fault partly? I'd like to think it was, but it's not. It's a much bigger industry than that. Two things happened. Um one is that the middle class evolution in large Asian markets, particularly China, has happened far faster than anyone ever thought. The idea that China would have, you know, a car, a mobile, you know, a a broadband service within a very short number of years was just amazingly un- unlikely at the time. So, that pool of expansion was never going to happen. The other thing was that, you know, when you look at the form factor of these things, it kind of got a Who would ever use that for the internet? Because our model of the internet at the time wasn't a model of apps and visuals, it was a model of email and browsing and keyboards. See, the idea was the internet was some special thing that you needed clunky stuff for. So, yes, a lack of vision, a lack of understanding, you know, just how bright marketing is all about, yes, we unleashed a creative industry that, you know, for them volume is everything. And we unleashed it too early. And now we're stuck in a transition problem where we honestly don't have answers. Good day. Um I've got two questions and one you probably can't answer, but one you hopefully can. Um the first is the big thing that I see is big companies and big government departments just not implementing IP version 6. And I'm assuming that people like yourself are getting to actually talk to the people involved. I was wondering about the issues that you're seeing that they're pushing back, why they're saying, "Oh, no, we can't do IP version 6." Uh the other one, and I suppose it's almost more to the room, uh you know, some of us are sys- sysadmins, some of us get to actually control this process of implementing IP version 6, but all of us are customers. Um I is there, you know, other than writing an email to my bank saying, "Please use IP version 6." Um what else can we be doing to help those Yeah, to push the the um put the the uh thumbscrews on them to to get get around to it. Right. Uh the first question, businesses, enterprises, and governments. The Australian government's actually been really quite out there in terms of putting stakes in the ground for all Commonwealth government agencies to have IPv6. I think it's within 18 months. Now, that's a qualified have IPv6 because there's the stuff inside the agency and there's the public facing services. There's the view of government from you and I. The commitment is the public facing service. From the inside you kind of think, "Well, hang on a second. Why should they spend the money? Everything works internally. They don't need V6 to make their business function today or even tomorrow. As long as the supply lines of their equipment, you know, their vendors and so on can continue to support them, they actually don't need to change. So, they don't. And it's very hard for anyone, even like me, to come and say, "Well, you really should run V6 internally." Why? Well, cuz you should. And well, show me the business case. I'm going, "Well, I can't. I can't." But from the public side, the answer is yes, they must. Now, what can you do about this? You can make better software because V6 sucks. Every single time I turn on dual stack, my performance goes through the drain. And the only answer that I've seen so far is that vendors who used to put out software that did V6 first and then fell back to V4 have now reversed it. So now, Mac OS, Amazon, all the Microsoft products, even if they have V6, don't even use it. So, the amount of V6 out there, oddly enough, is declining. Why? Because V6 implementations suck. They're poorly written. They're written, I think, many years ago and no one's truly looked at this and said, "How do I make V6 robust and fast?" If you did that, you could reverse the preference in everything and it would just work. But for as long as the folk who turn it on keep on seeing frozen white screens and 30, 60, 90-second delays before the webpage comes up. And in a 500-part webpage, add it up. Yeah, that sucks. Folk won't use it. So, yes, the challenge is actually back at you. It's not a case of saying run V6 and make the user experience worse. That will never work. We've got to make the user experience better. There has been some recent things like I think it's called happy eyeballs. Rather than thinking serially, well, I've got to connect to this webpage and I can either use V4 or V6. Tell you what, I'll use V6 and see if it breaks and then I'll back off to V4 a year later, you know. Start both things at once. Look up the DNS in parallel. Start your TCP sends in parallel. Figure out who gets the answer first and continue with that connection. Make the experience fast. If you do that, everyone says this works. So, I throw the challenge back to you. It's actually our software at the moment in six, which is the real turnoff for folk saying, I could run dual stack, but it makes the experience worse. We've got to say, I could run dual stack and actually things will get better. Yeah, what do you say to the There's an economic argument that suggests that as we run out of IPs and the price of an IP goes up, that that will cause companies with, you know, a vested interest with a lot of IPs uh that they own to push towards IPv6 because then they can sell their IPv4s. Well, there are a whole bunch of economic plays out there in a market. There are a whole bunch of theories as to when you would sell if you had spare addresses and so on and so forth. Should I sell now? Should I sell later? Will the price go up? Will the price go down? In any new market, the first couple of years is just rampant speculation and weirdness. V4 addresses will be no different. It's pretty true to say that one sector of the growth is these things. These things have the highest revenue per user of any device we've ever dreamt about. If anyone can afford to pay an enormous amount for an address, oddly enough, it's the folk who actually back end these because this stuff is not cheap. So, I suspect the market in the addresses for mobiles will appear and oddly enough, the operators of that stuff will pay. And for some amount of time, they will live there, sort of persisting in V4 cuz it's cheaper. As to when the economics change, interestingly, it's not a unilateral decision. I just can't do well, I'm going to go V6 if I shut my back on the rest of the world. Yeah? But think about this again from a different angle. If I have the 500 million richest consumers on the planet who spend the most and I go V6, what are the rest of you suckers going to do? Every market has tipping points where the critical mass of the change becomes irresistible to the rest. And I suspect there's this strange game going on about V6 as to who gets to adopt first, loser. Who gets to adopt second, possible winner. Who gets to adopt third, loser. And it's trying to position yourself at the cusp of tipping where you've got a product just as everyone else realizes they've got to join you. That's the critical piece of timing. It's sad to think that we've got to this kind of game theory in one of these sort of vital pieces of our society. We shouldn't be playing these games with coms, but we are. What can I say? I I wish it was something else. I wish that we'd actually managed to do this earlier. We haven't. This is the last question. Everybody else will have to talk to the speaker afterwards. Is hello. More places have done uh more to adopt the uh IPv6 like Japan. What lessons can we learn from that? They have not. Um literally it was all just hot air. Um the only thing that's actually happened in Japan is that NTT did a IPTV multicast on its service using V6. But, you know, the major providers out there still offer V4 the same as everyone else. The thing is to activate a customer in V4 uses conventional technology that's been around for oh years. It's really, really cheap. To do that in V6 today, how do I do a V6 only customer as a carrier? I have no idea. And that translates to an enormously amount of cost. So, across the world, no one does this, not even the Japanese. So, no, unfortunately, there are no lessons out there from that respect. Thanks, Geoff. Okay. Thank you, indeed.