 Good morning, everyone. Good morning. I'm John Curran. I'm the president and CEO of the American Registry for Internet Numbers. What you know is Aaron. He probably interface with us mostly through things like who is looking up IP numbers and where they are. I'm going to give a presentation about the depletion of the IP version 4 address space and the rollout of IPv6. This is something that I give this talk to a lot of different audiences because it's very timely now. I would say 30 percent of the content, 40 percent of the content goes right over the heads of the audience, so I'm actually very happy to be presenting here because I should be able to get some good questions and some good discussion going. We're going to go 50 minutes. When we get near the end, if it turns out we still have more, we'll go across the way to 114 and do a Q&A session over there. So let me get right in. Internet protocol, you guys know this. What we call IP is IP version 4. It was developed with the original Internet in the spring of 78. The ARPANET was actually ARPANET and MillNet were the first two networks that used them. They were built by a company called BoltBoranek and Newman BBN. I served as a CTO of BBN for 1990-98 and it was a company that was involved in doing some of the original protocols. IPv4 took off in a very fast pace. People don't understand. People think about the Internet today being a wild success. The Internet was a wild success in 1992. We already were experiencing 400 and 500 percent growth rates for the first few years of the Internet. The Internet IPv4 address space is 32 bits, which is just over four billion addresses. This is a major problem. In 1992, the IETF, the Internet Engineering Task Force, did some work. They did it under a working group called the AL working group, ALE, address limitation extensions working group. In that working group, we did forecasts of when the IP address space, IPv4, would run out. We came up with somewhere between 2010 and 2017 as the consensus estimate. In 1992, based on the rate of Internet growth over the preceding five years, we already knew the Internet was going to run out of IP addresses in 2010, because it's a 32-bit field. This was something pretty amazing. We knew all the way back then that we had a ticking time bomb in IPv4, and we were going to run out of addresses. Having said that, it's not an easy problem to solve. The Internet was commercializing in 1992 and 1993. The NSF National Science Foundation Network had been replaced with five of the initial five commercial networks. Those were Sprint, MCI, UUNet, PSI, and BBN. When I started working with the Internet, we had a host file, host.txt, still in place, so there was an enumeration of every computer in its IP address. We had just really begun to deploy DNS. Yeah, I'm old, that's right. The fact of the matter is that we were already running into scaling problems. That's what DNS was about. We couldn't keep a central host.txt file updated. In 1993, we did some changes to how we did IP address assignment. What we did is, the first one we did is we adopted variable blocks. Prior to 1993, if you got an address block, you either got what we call now slash 8, slash 16, or slash 24. Back then, it was a class A, class B, or class C. Basically, too big, too big, and too small. We're really the three block sizes. A lot of people needed 500 addresses for their business, and we couldn't give it to them, so we ended up giving them instead 65,000, because it seemed like a good number to give them. So we knew that was a mistake, and we rolled out classless inter-domain routing, CIDR. We introduced the notation that you all love called slash notation, where a network block has its significant bits designated by a slash and the number of bits, and so that allowed us, instead of giving someone a 24, we could give someone a 23, or a 22, which had two class C's or four class C's, and give them the right size block. Even with that change, we realized we were still going to be running out. CIDR made us do a better job of address management, but didn't solve the problem. What did we do? We formed a working group, Task Force, actually, called the IP Next Generation Task Force, IP and G. 13 member panel, I happen to be the official operator member of the panel. We had academic researchers, government, host companies, operating systems, a little bit of everyone, and we looked at six different successor protocols for IP version four, and we had the typical, you know, interwarrior fighting between the groups, and some groups merged, and some groups stole each other's ideas, and we ended up with a short list of two, and then we selected, in 1994, we selected one of them, and it was designated IP version six. We skipped IPv5, guys. IPv5 was a BBN experiment five years earlier. We took the internet protocol number, we couldn't use it. That's why the internet protocol goes from four to six, because version five had already been assigned. IP version six, a fixed 128-bit address field. Okay, so 128 bits, two to 128, 340 trillion, trillion, trillion, also 340 undecillion, if you happen to be into big words that represent a lot of zeros. If you took and you gave IP addresses a certain size, a certain density, and you could pack all of IPv4 into a golf ball, a similar sized sphere packed at the same density of IPv6 addresses would be the size of the sun. Okay, so that's a relative comparison here of two to the 32nd versus two to 128th. We hope to never go through this change again. I have faith, even with the best healthcare technology, I will never go through this change again. So fact of the matter is that we adopted IPv6, and it's similar to v4. All the protocols you know, so HTTP, HTTPS, SSL, it all runs. Everything SMTP, everything runs because it's an IP change. TCP and UDP above that still sit there, still do the same old thing. We just have bigger addresses to deal with. So we adopted this new protocol in 1994 called IP version 6. I will tell you that I'm the guy on record as saying we made a mistake when we chose IPv6 because IPv6 materially has no new features. It is literally very similar to IPv4. We did some cute things. We built in security, which ended up becoming IPsec for IPv4. And we built in auto configuration, which didn't exist in 1994. You all call it DHCP today. So everything we tried to put in 6 to make it unique and valuable, people necessarily said if it's that unique and valuable, let's put it in IP version 4. So we have no carrots to cause people to roll to IPv6. There are some differences. Some people can claim this minor improvements, but materially, it's very, very similar. Having no carrots makes transitioning v6 very difficult. Your only reason to ever move there is because v4 won't work. And the only reason v4 won't work is it's running out of addresses. This means that the demand curve goes from zero to 100% for everyone at the same time. Not a good deployment strategy. And I went on record saying this, and now I'm up here being the one championing it. So you sort of see the irony of that situation. Okay. So summary, 1981 for IPv4, 1999 is when the IPv6 standards were finalized. We actually did do this and get IPv6 through ITF processes, draft proposed, final standard. Look at it. 11 years ago, we got this done. Why? Because we knew we needed to have IPv6 deployed in every host operating system we cared about in order to make a transition even viable. That meant getting this protocol ready and out to host operating system vendors and router vendors probably five years before any customer ever asked about it. 128 bits, a dress format. We all know dotted decimal notation. It's what a v4 address is. Okay. So you've got 128 bits. If you're trying to use dotted decimal notation, well, you'll be here all day. We ended up going with a hexadecimal notation, four character groupings, A through F, and you've got eight four character groupings. So your IPv6 address, 2001 colon 0db8023 colon, it's long. Get used to it. We're going to have long addresses. This actually colon's pretty good and the format's pretty good. Turns out we have one little challenge for those people who are BNF posher guys like I am. It turns out to be challenging to write this in a way that also handles port numbers on the end with a colon. And so people, ITF went one way and the W3C went another and there are ways of doing it. Your local libc, some libraries that will help you. The DNS resolution libraries actually do a pretty good job of this. Okay. So examples of prefix notation work the same. Okay. So we've got V6. It's out there. It's 1999. We want everyone to deploy it, but no one has any reason to deploy it. Okay. Now, fast forward to today. We have allocated 80.8%, 80.08%, as of June 2nd of this year, of the IPv4 address space. Allocated means the IANA, the Internet Assigned Numbers Authority, which is a logical function that holds the free pool, has delegated address blocks to the five regional registries. ARIN is one of those five. APNIC and Asia Pacific, RIPE NCC in Europe. LACNIC, which was originally part of ARIN, we spun that out for Latin America so they could have someone, a group that was more local, had language and cultural issues that would be more responsive. AFRINIC was originally part of ARIN. We spun that out as well. So there's five regional registries. The five RIRs all draw from the same pool. So one bucket, and we all have five straws. You've seen this. If you've ever had a Mai Tai bowl at a Chinese restaurant, whoever draws fastest gets more. It's that simple. And so we actually have been drawing down at similar rates among the regions. Fairly comparable. I have a slide that shows that. But it's very consistent. We have been going through this address, these addresses, at about, well, I hate to say it, but about five to six percent per year. Well, if we've allocated 80 percent and unallocated is unallocatable, is really what that should say, 13.67 percent. Those are addresses we can't allocate to ISPs and hosting companies, can't allocate to end users, cannot use for general internet connectivity. Why? Well, that's Network 10. You know, RFC 1918 addresses live in this space. For those people who know multicast and anycast addresses, they live in this space. This is the address space we can never get back. It's permanently allocated to other purposes. We're down to 6.25 percent of internet address space. 6.25 is a very small number. Okay? Let me give you an idea of how small it is. If we look at the address space in terms of available IPv4 slash 8, so these are class A equivalents, slash 8 means one 256th of the address pool. Okay? And we look over the last five years, we had 62 slash 8s available to be assigned to hosting companies, ISPs, end users. And March of 2006. You can sort of see the steady drop throughout the year. In March 2010, that last bar, you see 22. Well, I'll tell you, this year in the next quarter, we allocated even more. So we're actually, this number's now 16. Okay? We're going through this year's rate at about twice the previous rate. You can see this trend line, folks. If you follow this trend line out, I will tell you, it's 340 days from now that this hits zero. Okay? 340 days, the central pool runs out of address space. That's pretty close, guys. So, at that point, each RIR has three to six months of space. We'll continue to meet requests from ISPs, hosting companies, end users. The way this works, by the way, is you don't buy IP addresses. You don't rely to your regional registry. You show that you have a need. If you're an ISP, you show you've used 80% of the last block we gave you. And when you've used 80%, you can apply for the next block. We have every ISP and hosting company in the country. Actually, Aaron handles right now, Canada, US, and parts of the Caribbean. They all come back to us every six months to a year and say, we've allocated all these new customers. Can you give us more address space for the next six months to a year worth of customers? Every ISP who has been relying on this for years for their growth is going to get told no starting in about a year and a half. 340 days plus about three to six months of inventory. So every ISP has this essential business demand to have more addresses to connect more customers to their network, and they're all going to start getting told no at the same time. This is wicked cool. So IPv4 demand, if you're curious how it's been allocated among the regions, this gives you RIPE NCC, Latin America, APNIC for Asia Pacific, and it has been fairly evenly divided. We actually have an agreement among the RIRs, the regional internet registries, that none of us will take down more than two slash eight blocks in any three month period. This is necessary because you could have a government show up to an RIR and say we have this huge demand, and the RIR would have to fulfill it, and then would have to turn around and potentially get address space, and we've sort of said anyone could create an unofficial demand that would be hard to audit. We don't want to suddenly see Africa say, oh, we have a need for the address space and take it all at once. So we've already instituted sort of throttling mechanisms so that the run out will be fairly equitable among the regions. Okay, so we've been allocating between eight and 12 slash eights each year nationwide. There are 16 slash eights remaining in the pool. We have allocated so far this year 10 slash eights. We will probably allocate 20 before the year is out. Huh, how's that work? Well, we've allocated 10 so far, so if we allocate another 10, the 16 will drop down to about six, and then that six will last us the first half of next year, so we're going to run out. I don't care who you are or how you do the math, this is pretty simple math. Our demand rate is such that even at eight to 12 slash eights per year, we have two years. If I use the math of our present rate, we have one year. It's going to happen. So we're going to run out. IPv6 is the approved answer. It is already sitting there in your laptop. If you are running a Linux derivative, you have IPv6. If you're running OSX, you have IPv6. If you're running most Windows variants, you have IPv6 and what are you doing here? Sorry. So it's out there. It's actually also supported by the layers above, so Apache knows what an IPv6 address is, IIS knows what an IPv6 address is. This is not inconceivable to turn this on. One little problem. IPv6 is not backwards compatible with IPv4. IPv6 is not backwards compatible with IPv4. Now, you would think the bright engineers at the ITF would have figured out a way to make this backwards compatible. The problem really isn't that we couldn't have done it with v6. We could have, but we would have had to change v4 once in order to make v4 have the backwards compatibility so that we could then roll out v6. It's a lot like if every cell phone in the world could dial seven digits and only seven, not six, not eight. The number shall be seven and seven shall be the number. If you could dial only seven digits and we ran out of numbers and we suddenly introduced 14 digits. Well, the problem really is that, yeah, you could, I could call you because I dial seven zeros and you're seven digits with my new funky 14 digit phone, but you couldn't call me because my 14 digits and someone else's 14 digits may have the same lower seven digits and we overlap. The problem was the existing base of IPv4 machines, it's code guys as tables and buffers and they're all set up for 32 bits and there's no way to generically change that to make it accommodate 32 bits and 128 bits unless we first upgraded v4. And the decision was trying to upgrade v4 is as hard as trying to roll out v6. We'll just roll out v6. This is a interesting design decision that was made in 1994. It is, however, the reality of life. Suck it up. It's what we have to live with. We really can't keep v4 running. You can do some cool things. We all know what NAT is, Network Address Translation. You can use NAT to do some of your growth. There are ISPs today who use NAT internally and use Network 10 and 182, 168, and all the private numbers for connecting customers. Those customers sometimes run into interesting address conflicts when they use the same numbers for their home and the router attempts to figure out why is 192, 168, 10.7 both on this interface and on this interface. So the reality is that you really can't scale indefinitely with IPv4 NAT. We've got to deploy v6 for the public Internet. IPv6 for the public Internet is going to be necessary if ISPs are going to be keeping growing. We either have to put a banner out that says the Internet is full, go away. Which is tempting, by the way. Or we have to deploy IPv6 on every public server. Every public-facing server needs to get IPv6 addresses put on its current interfaces. Now you guys know this is not new wires, it's not new fiber, it's not new cop. We're literally talking about putting a second address on the same interface. It's laying an address family on top. It's not that hard. I do get people in some presentations who are like, but John, I can't afford two new routers and I can't afford a second set of web servers. No, we're not talking about that. We're talking about adding addresses to your existing devices. We've been allocating IPv6 addresses since 1999. So when this was officially adopted, we started doing the allocations and people can get them. There's policies for service providers, community networks, end-user organizations. You can get your IPv6 address from Aaron, you can get it from your ISP. If you get it from your ISP, it's part of your service. That's pretty cool. So you don't have to pay for it. The bad news is you change ISPs, they'll probably ask you to renumber because you're no longer paying them. If you get it from Aaron, we do charge a fee for administration. Someone has to pay for the who is to keep it running, the DNS and adder service that's with that, all the infrastructure, someone has to pay to get me here to speak, you know, it's the whole thing. So there is a fee to keep Aaron running. We have 3,700 members in the U.S., basically every ISP hosting company, a lot of large enterprises, but you can get your addresses directly from your ISP. If you call up your internet company, and in the case of V6, you call them about six times and you bounce around about ten times, you eventually find the guy and you can say, yeah, I can give you V6, I can turn it on on your router, I can get you an address. It's not inconceivable. This is something that's a priority if you want your web server on the internet to have an end-to-end connection. Now, we all know what an end-to-end connection is, we know how valuable it is. It means your server is talking to the end host and there's no proxy in between or funky gateway operated by some carrier. Okay? That is the default plan, guys. If you don't put a V6 address on your web servers, then the customers connected with V6 will go through a carrier V6 to V4 gateway. We'll dynamically get an IPv4 address out of that carrier gateway and we'll connect to you. Okay? So this is pretty bad. If you care about latency and jitter, you actually want to run Skype, you actually want to do video streaming, the chance of you succeeding with that through a carrier gateway, well, I mean, it might work, but it might work one day and not work the next and then what do you have to do? Well, you have to get that user and Verizon and you on a phone call to talk about Verizon's overloaded carrier gateway. Oh boy. That's a good call. Have fun. Knock yourself out. Okay. So what does this mean? I'm going to handle this from the hardest organization to the easiest. Broadband access providers. Who's a broadband access provider? A lot of people, folks. This is, sure, Telcos who sell internet service to consumers at home, broadband, DSL, cable, like fiber, like that. But it also includes Wi-Fi companies. If you're providing a wireless ISP, you're a broadband access provider. Congratulations. You're going to run into the same problem. If you are a large enterprise, you work at a company that actually has lots of divisions and you're the internet backbone for them, you might be effectively a broadband access provider. So pay attention to this one. Okay. What do we know? I want to talk about Granny. Granny, grandma, grandpa and grandma next door, they access the internet. They pay someone to get to the internet. They pay the cable company or the phone company or they pay someone. And they think they have access to the entire internet. This is true. I guarantee you, you go to your neighbor and you ask, is your internet service all the internet or just 88% of the internet? They'll say, oh, I bought the whole internet thing. I didn't buy a subset of the internet. Okay. This is very important. So ISPs come to us and they say, we want more V4 addresses. Aaron says, that's great. We're now out. It's two years from now. But you can have V6. They go, great. They start connecting customers up with V6. Well, right now about 2% of the internet sites have IPv6 addresses. That means 98% of the internet sites don't have V6 addresses. Something has to connect those customers who have IPv6 addresses to the public internet. Now, I want to be really clear. Granny has an IPv4 address in her living room. It's her broadband access modem that has a V6 address because, well, they're generating 10,000 of them a year and they can't assign V4 to their living room. So there's already one level of mapping in that, sitting in every living room. We all know that. 90% of the people who have broadband have not turned on in their broadband living room. You just get used to it. But the problem now is that when she goes out to the internet, instead of getting a public IPv4 address, because the carrier has assigned one or assigned a pool of IPv4 addresses, now the problem is all those devices are connected with V6. They're connected to the entire internet that gets a dynamic V4 address. And when they go to access your website, you're going to go through that gateway. Now, there's no choice. If carriers don't put this in, they can't use V6 addresses. If they can't use V6 addresses, they're not going to grow for long. You can't layer NAT and keep it running very long if you use V4 to V4 NAT. So what happens? You're going to have a lot of management and say, I don't have full internet connectivity. And the telcos, well, they actually are regulated in some things, and Common Carrier is one of them. They'll all get dragged in and get told very loudly by the state public utility commissions fix Granny's internet access. So we know this is going to happen. We know that the job of providing connectivity for you're providing wireless service to a conference, and someone connects up with IPv4, you have to realize the fact that if you don't have enough addresses, you may have a challenge, you may have to assign them a V6 address, and you may have to figure out how to gateway them as well. Now, there is technology for this. Carrier-grade NAT, 6RD, there's a whole bunch of them. It turns out we don't have one transition plan. In their own unique and individual ways. The fact of the matter is that the ITF ignored the transition problem until about five years ago and then realized the Carriers needed this and began working on a lot of technology. So if you go to the ITF web page and you look for IPv6 trans, the working group, you'll see all the drafts. There's a whole bunch of ways to do it. But the point to understand is that if you don't put a V6 access by a gateway in about two or three years as customers come on with V6 public addresses, they'll get dynamically mapped back to V4 and you'll have to suffer the consequences of that latency and delay. So broadband access providers are paying a lot of attention to this. ISPs, let's say you're a backbone. All you do is provide transit. You've got a big backbone, a bunch of routers, you run IPv4. Well, you're going to have customers that have IPv6 service because if you're a broadband access company you need to provide V4 and V6. If you're a website, you want to get a V6 address so that you can also serve to those customers end to end. So ISPs are already getting demands for IPv6 service. There's a bunch of ISPs who are really good at this. So NTT, big program, they have IPv6 globally, global crossing as IPv6 globally, Hurricane Electric, little RISP that made its mark on IPv6 service, actually doing very good job. You can go get commercial IPv6 service. And so if you're an ISP, if that's your business, be prepared for business customers calling you up saying, I want not just IPv4, but I want IPv4 and IPv6 on the circuit that I have today. Not hard to do, turn on dual stack on all your routers, run a second routing table. I'm not saying it's trivial, but you know how to do this. It's just a question of picking up your input. Internet content providers, this is the fun category. So if you make a business providing content to the Internet, this change from V4 to V6 is actually very exciting because it's going to put every broadband access company, every carrier in a situation where they're gatewaying customers to you because they're connecting customers with V6 and your websites V4. Well, some of the brighter and larger, brighter folks who provide content to the Internet have realized this is a problem. Why is it a problem? Well, you want to control the user experience and you have no control over what the carrier is going to do. If you can run V6, though, you can get packets all the way to the broadband gateway, you can get them all the way to the customer premise and that might be a better experience. So this is why Google has turned on V6 for the vast majority of V6 on www.google.com. They've put it on v6.google.com. Let me describe that a little bit. If you have an IPv4 and an IPv6 address, you have an A record and a quad A for your DNS for a given name, okay? And you have a customer who thinks he has V4 and V6 connectivity. And the operative word here is thinks. He has V4 and V6. So he has an interface that for some reason thinks it has an IPv4 and IPv6 address. It will use the IPv6 stack. Why? We want to move to V6. It's the right thing to do. Okay. So if a customer has misconfigured IPv6 on his local LAN, because some box is announcing itself as a router and he's going out to the internet right now using IPv4, just like we all are, everything's fine. When someone puts a quad A record for www.foo.com and they put a quad A that says I can be reached via V6, the poor user who thinks he has V6 connectivity now starts syncing packets into a black hole router that can't get them anywhere. So you had connectivity, someone puts a quad A IPv6 record in for a public web server, suddenly that web server is unreachable because the user has broken V6 connectivity. How many broken users have IPv6? How many public users have local broken IPv6 connectivity? 0.1%. It turns out there's a lot of places where you might be getting a V6 address assigned to you that goes to nowhere. A lot of corporate networks have a mistake. They actually have someone turned on and routing V6 with no connectivity. Sometimes vendors help us with this too. There are home gateways and OS is that when acting as a gateway announce themselves as a V6 router even if they can't get anywhere. So the reason Google doesn't put in www.google.com and add an IPv6 quad A record is because 0.1% of the people who tried to access it would run into a problem. Now, does that really matter? Well, you know, you make your money on advertising and click throughs and you have these pages come up and 0.1% means 0.1% of your revenue is gone. Multiply billions by 0.1%. Yeah, it turns out to be a pretty big problem as it turns out. Now, it's getting better. That used to be 0.5%. It used to be a whole percent five years ago. About 0.5 last year. It's about 0.1% this year. Now, if you look at the IPv6 record for your public web server, recognize some people can't get to you anymore. Now, those people who can't get to you anymore are broken, okay? Stupid people, but that's okay. They can't get to you anymore. If you look at Aaron's website, you'll find we have a quad AN. You look at a lot of websites, we've put quad AN. This means that some people can't get to us. I'm not saying every business can do this. Google can't. Facebook has v6.facebook.com, but no, they have not put a quad A record on www.facebook.com. Again, 0.1% of a business is pretty big for some people. We're trying to fix broken v6 islands that are out there. If you happen to see that you have a v6 address on your local land and you know you don't have any go fix your router, it's sending router advertisements out and it shouldn't because it's not connected anywhere. It shouldn't be saying it can get anywhere if it can't go anywhere. So this is a pretty hard challenge. You need to put a v6 address out there. You need to be ready for IPv6. Depending on your business, I may tell you you need to be ready but use a different domain name for now. If you're a business that lives and dies on page views, okay, and number of clicks or advertisements served, okay, you might not want to put it in. Or if you put it in, you have to realize you're going to get customer calls. But for the vast majority of people, put it in. If you want to be ready for a very large number of users who are accessing you via v6, you should start preparing now. Enterprise customers. Mail web application servers must be reachable via v6 in addition to v4. Open a dialogue with your ISP that is hard, but we can't tell you to deploy this. This is North America. Canada, US, Caribbean. In other parts of the world, they solve this problem much easier. They tell someone, if you want a business license to be an ISP, you'll turn on v6. If you want a business license in order to be a public content site, you'll turn on v6. There are parts of the world, like South Korea, Singapore, you also have very high v6 deployment. That's not how we run North America. Just really not the way it's being done. So I can't tell you, you have to turn this on. What I will tell you is, you probably need to take your public website and know what happens when you put a v6 address. I have a private website. I actually went out and turned on v6 one day, put the interface on, configured Apache, everything was great. 6am the next morning, my website disappeared. I couldn't find it anymore. Turns out, I had some scripts that I wrote that process log files. A lot of people process log files as it turns out. You want to know where you're being accessed from. You want to know what addresses they are. Turns out I wrote a script that never thought about an IPv6 address being processed, and Pearl is Pearl, and we all know what it does. It does what it wants to do. That's what's going to break guys. It's not Apache, what's going to break is all your scripts that use IP addresses in them. Any configuration file, you've got people out there who have little databases that they've set up where they type in 16 characters. That's going to break. For larger organizations, you've got a help desk system. Remedy, Clarify, RT, Request Tracker. You've set up fields. Oh, a field IP address. Try typing in 40 characters into that field. Someone can't call in and open a problem with IPv6 until you've tested your tools with IPv6. So I highly recommend getting the lab set up two machines, turn on IPv6, run your applications, run your public-facing web server, process your logs, open a ticket, see what it does before you have to do this because someone runs down the hall and screams at you, we need to have IPv6 connectivity. And they will. If you don't have IPv6 turned on, then what's going to happen is that at some point you're going to be streaming video or streaming audio or you're going to be doing Skype or you're going to go to pieces because of the latency and jitter. And then someone's going to tell you to upgrade and you have a choice. You can either do this in advance and be ready or you can be told, this weekend you need to upgrade. Sorry, I don't care. We have a new business partner and they said we're on the old network and I don't know why, but go fix this. Unless you like crash weekend projects upgrading your public infrastructure to v6, this is probably one you want to lay in a little earlier. Software vendors, software vendors. This is actually pretty cool. Here's a demand curve that's going to go from 0 to 100% a year and a half from now and no one knows about it. I was talking to Vaughn, voice on the net, big conference packet networks, et cetera, et cetera. And the folks at Vaughn were like, this is pretty cool. One guy got up and he said, John, you're telling me that some of my customers are going to want to have PBX support for IPv6, SIP over IPv6. And most people don't know about this and they're going to want it because well, they may have new sites connected with v6 only. And if they try to use carrier gateways and SIP, it's not going to work. This is great. I want to tell my marketing team to build this, my product team, build this. He says I want one favor. I'm like, okay, sure, what do you want? He says stop talking about this. I don't want my competitors to know that they need to have IPv6 turned on. If you build hardware or software that speaks to the public internet, you need to pay attention to the fact that you're going to have people asking you for IPv6 about a year from now. Probably not right now, but about a year from now you're going to start seeing the demand. And it is happening. I know the database vendors the other day were beginning to get requests from it. I had one call me up, literally, like what's this v6 thing? And I'm like, oh boy, here's a conversation. So this is going to go from zero to 100% very quickly, and then you're going to have to build a lot of hardware. Guys, if you build, you know, programmable logic arrays and ASICs and like that, you need a year. You should be working on this now. Government, I kind of talked about this. We don't, at Aaron, we don't advocate any public position for this, meaning it's not our job to work with the government and tell them that they should tell everyone to deploy v6. No. But we do work with the government both of them have IPv6 road maps for their own systems. You'll be hearing more about that later on this year. This is actually more important than people think. The reason your routers and firewalls support IPv6 is because in 2005 we did a memorandum with the U.S. government that mandated every U.S. government network had to be IPv6 capable. And in fact in 2008, by June 30th, a number of them turned on IPv6, pinged each other with IPv6 packets and then very quickly turned it back off really quickly. Because a lot of them had routers that worked with v6 but no firewalls. So the fact of the matter is that people were like, well, why bother? Why bother having the federal government do that? U.S. federal government buys $61 billion worth of IT systems and services each year. And in 2005, mandated IPv6 capable networks, a number of major router vendors stood up and said, oh, that's us, right? Okay. Yeah, we'll go do that. And put it into their main code base. The U.S. government is now working on standards for IPv6 for public websites, which should help the load balancer acceleration crowd to get products in place so that if we need these two years from now they're available for commercial people. So we try to prime the pump with the government wherever possible. That's about it. So what if you want to do v6? Well, first you can do it in your lab. You can grab private address space, work with it. You want to do it on the public network, get address space from an ISP or from us. IPv6 connectivity. Native is ideal from your ISP. You can get tunnels. There was a talk earlier this week, very nice, talked about setting up tunnels. You can set up tunnel v6 over v4 so you can get connectivity, make sure your site works, see the implications. Check your operating system, software, network management tool. You're pretty good on the OSs. That stabilized a lot. Ubuntu, Debian had some real issues two years ago that I was busy screaming with people about code trees. We're okay there. Your network management tools are really where you need to pay attention. Rotor firewall and other hardware upgrades. It is possible. Your network management tool. That supports v6. That's in a different license category. All you need to do is send us money. I'm not saying that's not the case. Think about the fact that over the next three years, you're probably going to upgrade it anyway. Just make sure when you replace the silly thing that you get v6 in the replacement. When you're buying a new one, they'll throw it in for free. I guarantee it. Just make sure you get it. You can get v6 on in two years and find out you didn't get it. Resources. www.iron.net slash knowledge slash v4 v6. Again, if you have broken v6 connectivity, you can't get to this, but that's okay. Social media. You can get to it over IPv4. You can get to it over IPv6. But don't have broken v6 connectivity because that will slow you down. Tmarin.net is where we have all the presentations we've done. You can download this presentation. You can give this presentation to someone else. I highly encourage that. www.getipv6.info. We have implementation notes there. It's a wiki. Log on, see what's out there. When someone learns something, they put slide decks or notes up there. If you happen to have some learning, put it up there. We all know how this works. The IP addresses are set by you, i.e., the public community. You don't have to be a member of iron. Anyone can participate in the policy discussion. As we continue to slice and dice the remaining IPv4 address space and fairly allocated out, there's a lot of interest in what policies we use. If you want to know what policies are being used and what changes to policies are being used, www.iron.net slash participate. You can see the policy process. You can participate in iron entirely online. You can see the billing lists. The meetings are all remote cast. You can participate, ask questions online. The full thing. You don't need to go to our exotic meeting locations. That's about all I have. Fellowship program. If you know someone who wants to participate in iron and can't afford it, we actually give out fellowships. I think the fellowship program is open for another three days or four days for the next meeting, but it opens every six months and the program will pay your tuition and it will pay your flight, hotel and meeting expenses to come out. You can get involved in iron even if you can't afford to get involved in iron. And no booth. Don't worry about that. I'll take this down. That's it. Thank you. And questions. Questions here. If you want to follow up with questions across the hall on Capri 114. Yes, question. This is a larger than Y2K. The government's been a little busy with other things recently. The US government. This will end up becoming Y2K, you could test your own systems in the lab and know how you do. In this case, your systems have to speak IPv6 in addition to v4 for customers who are coming on to ISPs you don't know. So we knew when Y2K was. It was probably the only thing we did know when it was going to occur. This is a moving date of uncertain impact that affects everybody globally. So, yeah, it's going to be bigger than Y2K. You'll be hearing about it. There's going to be a lot of business in being an IPv6 consultant. If you want career stability, I will tell you there's no better field to be in. Other questions before I move across. Yes. How long are we going to have to maintain v4 and v6 on the public internet? 5 to 10 years. We won't have people actually connect servers servers with a v6 only address if we can get so much as one IPv4 address to put on that server. But even that's going to run out eventually. So, 5 to 10 years and then what will happen is we'll end up inverting the model. We'll have companies that say, I don't want to do a dual stack. I'm sick of maintaining this. I've enabled v6 to my desktops because that's the next step, obviously, is you push v6 down to your desktops. At some point, 5 to 10 years from now, you're going to say, I'm done with v4. I'm turning it off. I'm putting a gateway at the edge that will translate my v6 users to v4. And if I get a business that's having a trouble, I'm going to tell them, look, get off the friggin, you know, telex, you know, get to the modern age, turn on v6. But that, you're not going to have that until 2020, okay? Sure. Oh, so there is something that's going to happen that's very cool with IPv4 addresses. So Aaron's going to start telling people, we don't have any. Yep, we know you need it. You'll get approved. We can see your demand, but we don't have any. It will be possible to transfer addresses. This is an approved policy. From someone who has addresses to someone who needs addresses, where we've approved their need, i.e. there's going to be addresses on eBay at some point. But the only party who can get those addresses is someone who's applied for address space and then turned down, not because they have the need, they've been approved for their need, but we didn't have the addresses to give them. That gives you a chip. I'm qualified to get a slash 21. I'm going on eBay to find it. So I do expect the addresses that are coming off that people don't need will very quickly be sucked up and they still want to use v4. Other questions? Yes, mobile devices. Wow. So as it turns out, this is one of my... I've been dealing with the three and fourth generation network projects. A lot of the mobile devices actually have IPv6 in them. I actually posted on Facebook the other day a picture that looks capable. If I happen to get on a wireless network with... amazingly, they call it iOS. iOS 4. If you happen to get on Apple's iOS 4 with your phone and you're on a Wi-Fi network that gives out v6 addresses, you'll get a wireless v6 address. It works. There's a lot of devices that support IPv6 Now, about two years ago I met with most of the carriers and most of the standards organizations and I said, the odds of you getting v4 allocations past 2010 is pretty much zero. They know this is coming. I know that Verizon, for example, their LTE, their next generation project, they announced the handsets for their new project. All of them must support v6, must assume that a v6 address is assigned. They optionally can support v4. Optionally, okay? That's going to switch over the next two to three years. Most people will have no idea that's happening, just like you don't know when you're sitting behind a nat right now other than the fact that you see a private address. A lot of people using their cell phone have no idea what IP address they've gotten. They're going to get a v6 one and for now go through a v6 to v4 gateway. Okay, I think we're overdone and have to evacuate the room. I'll go over to Capri 114 for anyone else who has questions. Thank you very much.