>> Zhengyou Zhang: Thank you very much for coming over tonight and we love to have you here. This is the first lecture of a series on the state-of-the-art of technology. We are going to have the second Tuesday evening from 6 to 7 p.m. over here and this will be sponsored by a different society. Today we are going to have things like consumer electronics and computers and then we are going to have Computer Society, Signal Processing Society and each different aspect of what's going on in the world that are the most exciting things. We have some pizza here. If you would like some pizza you can go to the back and pick up the pizza and then come back to enjoy our lecture by Joe Decuir. Joe has been with us for many, many years. He is very active in the chair and in various other positions. He is pretty much a guru of Seattle section and he has become our new Fellow, 2015 and so congratulations to Joe. [applause]. And he gives distinguished lectures all over the world and does all kinds of wonderful things. He is on the board of governors for the consumer electronics society. He is a wonderful person. >> Joe Decuir: But the thing that we're most excited about is that we are going to be chairing the Global Humanitarian Technology Conference here in Seattle this October. >> Zhengyou Zhang: Yeah, and Joe is our chair. >> Joe Decuir: And she is the vice chair. She will be the chair the following year. >> Zhengyou Zhang: And so we were going to have a very big event and hope all of you will come to join us in the conference and help out, volunteers to get good humanitarian technologies and get things started and get very active in Seattle and connect with the world. So please go get your pizza and come back here and then we will hear what Joe has to tell us. >> Joe Decuir: Today. Is this mic lit? Good. I might as well introduce myself and keep going because we have 41 minutes. You heard the introduction. The one thing I would add to that is that I worked here, lucky me, starting in the fall of 1992 and I retired after eight years, time off for good behavior in 2000. I want to talk about some problems related to building the internet of things. In particular, I want to describe a particular solution that uses Bluetooth low energy. I want to talk about the problem to solve. I want to talk about the ways that we are considering connecting the low-energy world to the internet and then various ways to map it all through and then to talk about what to do with this and how it would work. Given the time I am going to be a little crisp, but I hope there is time for questions at the end because if one of you is brave enough to ask a question, it's likely that several of the others would like to know the answer and that means you're paying attention. The first problem to solve, I'm a connectivity guy currently and I'm interested in connecting the low-energy world to the internet. Today, this is industry hype. You can hardly open up another magazine article online hyping the internet of things. There are going to be probably trillions of dollars invested in this and my wife keeps asking me who cares. Why would I care about these things that I want to pervade my house with? How many of you have seen recent, and I mean in the last two days, articles on the web about people hacking into people's cars? Anybody? Yes. Several of you. As an aside, I work a lot on security and I believe that security for these things is a forethought, not an afterthought. I believe that the security, the customers will shun any devices that appear to be leaking. If anyone wants to kill this market quickly, they will produce a lot of stuff they gets hacked. To answer my wife's question, I say in this whole space somebody has to come up with an application that would get me to want to buy the device. For example, two years ago, 2 1/2 years ago I bought a Fitbit. That's nice. If I knew in advance that buying a Fitbit would allow me to lose 20 pounds, that's a sale. But that's just an attraction that doesn't mean it's what it's actually going to do. As you can imagine, willpower matters more than what this thing tells you. That's just an example of the attraction. Then I might use it if it actually connects to the world. My Fitbit will talk through low-energy to my phone, my iPad and tell Fitbit.com how I'm doing or am not doing. It's starting to accumulate information to tell me and it gives me little electronic rewards if I walk so many steps a day et cetera et cetera. A question to always ask is where is the intelligence. Is it in the device? Is it in something that you control? Is it up in the cloud somewhere? To pick an example, you all know, because you know what the Kinect is, that the Kinect is something where there is intelligence in the device staring at you, but there can also be intelligence in the complete Xbox and it can go up to the cloud and back. You know about Cortana and you know about Siri, which records your voice and they can do voice recognition and speech recognition based on what you tell your cell phone, and that's where the intelligence is in the cloud, not in your hand. Always ask yourself where is the intelligence, really. And if the intelligence is not where I can see it, someone can get to it. Like I say, you've got to be able to trust it, otherwise, people are going to run away. We may, in fact, sometime have people looking back on now as the good old days before someone could hack my car. The Bluetooth SIG where I work for a living has identified three different ways to connect low energy devices to the internet depending on the use case. The first one is let's take an existing device which doesn't know where the rest of the intelligence is. It's just what they call a generic attribute protocol server. And the RESTful API allows the device going through a gateway from anywhere to attach to and control this device. We don't need to do anything different outside of the gateway. The second way is if you have a really smart little client app and you teach it about HTTP, so you can tell it to go through a gateway, through a proxy server to talk to the internet. Make sense? You all know what proxy servers are by now, maybe. And the third is to put IPv6 intelligence directly in the device and that is the subject of today's conversation. I just walked through these and this is sort of a picture. Existing stuff can go through a gateway that's RESTful. Does anyone here know what I mean by talking about restful? Some. All of you who have been on the web for the last almost a generation now and know what HTTP that appears in the URL for the place you want to go, right? HTTP stands for hypertext transfer protocol and it is a restful design. We're going to do more of this going forward. Stepping back, I was describing that if you teach this thing to go back out to a proxy to the web, so for example, a pair of these things together, they are speaking native Bluetooth low energy and they can go out to the server first, punch a hole in the firewall, which needs to be there and explained to something out here where they are. Third point is if I put IPv6 or some other native IP directly in the device, in which case it needs to talk to a border router from clients and servers out on the web. Here's what it might look like in your house. There are two common ways to find internet-access. One of them is 3-D and 4-G through your phone in your pocket or if you've got some kind of broadband modem, cable, DSL whatnot from an ISP and again you need the firewall so you don't get hacked, and then you can talk to a gateway in your house which is doing the 6L border router to talk to all these smart nodes in between. This is a good time to check for questions. Is this mysterious or does this make sense? I've got 35 minutes. Let's suppose I want to connect these things. Who here does not know what the internet protocol is by now? It was such a wonderful thing that we built it and reconnected the world. Unfortunately, the people who designed internet protocol were not aware of how popular this was going to become. When they designed it, they had 32-bit addresses which is 4 billion unique addresses and that is no longer enough unique addresses to address every living human on the planet. If you're going to have more than one gadget per person, for sure, IPv4 is not sufficient. We're imagining, maybe in a nightmare, tens of devices per person and that means we need a lot more bits. Today, to function we have network address translation. We were working on network address translation at Microsoft 20 years ago. That's when a set of local devices are mapped into a single external device and port addresses are used to map particular functions inside the protected network to the outside. IPv6 now has a huge address space so to put it in perspective, 32 bits is not enough to address every living human on the planet. 64 bits is enough to address every living insect on the planet, if you can imagine how large that is. 128 bits is that much larger. If we have now defined a system that can basically address anything, then we've got a foundation to start connecting to all of those devices we might conceivably want to have. IPv6 has a larger header and if you want to adapt this to things that are light, you have to do some adjustments. Otherwise, you spend so much time on the air just shipping those 128 bit addresses back-and-forth that you don't have time to do anything useful. The IETF, Internet Engineering Task Force, has been building a way of mapping IPv6 to smaller devices. Their first target was the 802.15.4 radio. To translate that into English, how many of you have ever heard of Zigby? That's the radio it uses. Z wave also uses that radio. There is a cable modem keyboard control function that also uses 802.15.4. Their original plan was to start porting this to a new and interesting radio that was Bluetooth smart, low energy, but the people from the Zigby space said that we own Ypan. They've just decided to name it 6LoBTLE instead and this is a decision that happened like three weeks ago when the document, it was in voting. This is sort of hot off the press. Stepping back, there's a technique for adapting IPv6 to small devices. It's been done to the first radio. It's now been done in ballot to the second radio and it's going to be applying also to 802.15 radios and so on. Yes? >>: I'm sort of in principle like natting so what's the big advantage to having native IPv6 addresses? You know, if you are doing some kind of a subset and mapping it. >> Joe Decuir: What they do basically in the border router you have to compress the addresses to the nodes and I didn't want to get deep into exactly how 6LoPAN works but that's what's going on. It's not the same as natting; it's more a compression. >>: And then it's technically different but why are you going through all this trouble, you know to get native IPv6? What's wrong with NAT? Why would they… >>: It is IPv6. Exactly, it's an IP overfull definition, just like IP over ethernet or IP over Wi-Fi, IP over 15.4s. It's just like that, so that's a high order bit. There is no natting. >> Joe Decuir: You know where to find me, so if I don't answer you enough today, we'll talk. I'll give you a card later. As he was saying, IPv6 over foo. Problems to solve, first of all, connecting it to IPv6 and all those giant addresses and all the other functionality. The topology, the limited packet size because if you want these radios to be small and cheap, you don't want to put, has anyone looked at the specifications for the newest Raspberry Pi? Nice machine. You could run Windows 8, or Windows 10 is going to run on. That's way too much functionality to put in a small cheap thing. We're talking about things that probably have 16 or 32 bit processors. And they don't have memory management. They will have much less than a quarter megabit of memory, because if they get larger they're first of all going to consume a lot of power because they're gonna leak and then you start putting them into lights because you have to put it in something that is powered all-time. You have to do service discovery and you have to do security. That's just the top-level list. The characteristics are you wanted to have a small packet size. If they're large it's not low power anymore. You want to shorten the Mac address. You want to have low bandwidth and low duty cycle. You can support Star and you can support mesh. Does anybody know what I mean when I talk about a mesh? Good. You are trying to keep it low power meaning you want it to run on a battery if not a coin cell rather than being lines powered. Obviously, you want a cheap. You want to be able to support a lot of devices and have a technique that supports a lot of different things that are distinguishable and groupable in some way. You want it to be ad hoc deployment because our customers are not going to be your corporate IT department. The people who deploy this are going to be basically anyone and we have to design systems that can stand it to be installed by anyone. You want the devices to sleep a lot to save power and assuming that you have at best a 32-bit processor but probably a 16. What the border router does is it looks at the first packets that are sent in and it recognizes things and it'll compress them so that on subsequent uses they are much shorter. It's actually pretty simple. Then you compress the subsequent packet headers as well because you already can assume things about what was done before. You still have to do address resolution. And you still have to handle different kinds of radios and some applications will have different parameters to optimize. Without a lot of time I am not going to get deep into that. We sat there in the Bluetooth SIG and 2 1/2 years ago the IETF came to us and said can you give us the foundation so we can put IPv6 in these things. And we said okay. And being the Bluetooth SIG we take twice as long as we ought to be but we eventually finish. This was adopted in December. We want to build new use cases. The market for 6LoWPAN is expanding. Who has heard of thread? There are more than 100 companies and consortia who think they are going to tell all of us how the internet of things is going to be wired up. Thread is one of them. You'll hear about more of it later. People have been putting 6LoWPAN into systems for managing power today. Step back for a moment and say what is a smart house? Any ideas? Oh come on. >>: Smart house? >> Joe Decuir: Yeah, and how do you measure the IQ of a smart house? Some people think a smart house saves them money on power, close enough. And a lot of the original applications of 6LoWPAN are in that space. Some people like to have their entertainment follow them around the house. That's what DNLA is for. Some people want better control of their security and who is in the house and can they turn lights on and off and et cetera. A lot of work is being done in that space now. Of course, some people like my dear wife think that they want a house that takes care of itself, that not only will it vacuum itself, maybe it will clean the toilets. I had to explain to her that that is probably a ways off and it's more of a physics problem than a communications problem. There is an organization called Internet Protocol for Smart Objects, IPSSO. It is closely related but not the same people as the IETF, Internet Engineering Task Force and this presentation is about how to hook together a system with a Bluetooth radio using a particular profile talking to 6LoBTLE talking to this. That's the stack. Let's start with the smart radio. In this design the ordinary Bluetooth tools are used for finding and connecting to a device from something in the center but they are not used thereafter. There's a feature that is in the second version of Bluetooth Smart called connection oriented channel and this design takes the connection channel, sets it up with a minimum MTU of 1280 octets which is big enough for your average IP packet. It lets it run underneath of the upper stack allocated a fixed channel. This is what it looks like conceptually. Here is the node. It's got the regular radio. It's got the link layer. How many here have not heard of the open systems interconnect model by now? Good. This is the multiplexer layer. These pieces are used to make the connection but afterwards it runs that stack. In fact, I need to change that because they just renamed it and I didn't catch it in all of the places. Here is a border router which connects to IPv6 sources out here and then connects them to all of the child devices. In regular Bluetooth low energy org smart today it's low energy because it's designed to be very low duty cycle. It, you can do a transaction of advertising for attention, forming a connection, getting and receiving a command and turning off again in 3 milliseconds. How often do I need to turn off the lights in this room? Maybe, maybe twice a day. On and then off. If I'm doing that the amount of actual communications time is tiny and for a lot of other things that you can imagine in your house the actual time that they are going to be on, or the time when any changes occur, those are tiny compared to how long they are on. And that's how you can build things that run on coin cells. Ordinary packet sizes are 23 octets. Even a long executed 512 octets, so transferring a long message cost a lot of battery time compared to 3 milliseconds. Common internet protocol packets are 1280, or they can be that large. Here's what the stack would look like inside a single Bluetooth device. It has access profile for finding things, profile and attribute protocol for controlling things once you've found them and the multiplexer, the link layer and the physical layer. Applications profiles and services are what does this thing do once I've decided, what causes me to want to buy this thing? I didn't buy this thing because it has LE in it. I bought it because it measures something that I consider important. The connection oriented channels is we are now bypassing the simple mechanisms for controlling things in a device and going to something that's less constrained in terms of how much data it can deliver to deliver it to something that's beyond the scope of the Bluetooth SIG. Let's suppose, for example, I have something that I can turn on and I can have a characteristic that is simple automation I/O and I write a 1 to it or I write a 0 to it. An ordinary at and get, that thing will take, a single transaction will probably take well under a millisecond. If I have to deliver longer pieces of things that boil down to turn the light on and off, I'm choosing to spend energy in other ways just to have a more general method. Does that make sense at all? We added long packets. We made room for this stuff and we let devices that live above the BTLE layer decide whether to switch or accomplish any other action that we want. Why would I want one of these things? Like I say, there are probably 150 companies were consortia that one to tell us how the IOT is going to work and they're not all correct because most of them have few followers. You can classify them by category. IETF, Internet Engine and Task Force we've already talked about. AllSeen is an organization that Microsoft is a member of developing plumbing for connecting devices. MQTT and don't ask me to recite all of the names. There are automotive specific things. As you seen, people are wiring up cars and then connecting to the internet, God help us. There's healthcare. Has anyone ever heard of IEEE11073? I expect most of you have not. 11073 is a set of IEEE standards for talking to health and sports and fitness devices. It defines what the data looks like. It was designed originally for the convenience of organizations that write the applications that use these semantics. In the Bluetooth world what we've done is wrapped all that stuff up so that lots of these things now speak that language or speak to a translator that speaks that language. Home automation, big complicated space today, but we're trying to drive some convergence. Industrial, utilities and smart grid, so home automation, the organization IPSO has been building a set of specifications of semantics for how you talk to all of these different parts of a house. The point, of course, is that if I want to write you bunch of applications, how many of you remember the history of computer operating systems? Some of you should. The applications writers are the actual proxy for the end users. The end users are the ones who actually spend the money on the product. The applications writers want to have as much commonality as possible of the things that they're going to write to so they don't have to write many different versions. Way back in the mid-to-late '70s we first started building personal computers and the people writing word processors had to write a different printer driver for every different printer and that was before Microsoft came and other operating system companies came along and abstracted printing and abstracted communications and all those other things. The applications writers would like to be less confused about what's underneath. The convergence I'm interested in driving is convergence on the models from IPSO because if I can build systems that then translate to using things like AllJoin [phonetic] then I can attract those people and build markets. A couple of the devices from the dedicated application services, right, I talked about writing to a printer. I don't want to have to deal with a bunch of different abstractions to turn a light on and off. It's always an advantage in software development to not have to reinvent wheels. Do it once and test it. I want operability over machine to machine protocol. There are several examples that I will quote later and before that can be translated into the AllJoin connectivity system. I want the developers of the developers and of the and gadgets to concentrate on what they're really good at and not having to reinvent all of the connectivity stuff in between. These people have been writing applications that run on top of 6LoWPAN in homes and with the development of 6LoBTLE we can take the same applications and run them on a different radio. They use something called CoAP which is constrained application. I think I've got a slide on CoAP. Where is it? I'll jump back and forth. I want to explain what CoAP is. You have all heard of HTTP. CoAP is designed for light weight stacks and uses UDP rather than TCP, so it doesn't count on the system underneath it to keep track of reliable delivery. It just sends a message and moves the timers up to a different level. It's very similar to HTTP. In fact, it's kind of a subset of it and there are now web browsers extensions that you can get that speak to CoAP, so imagine it may be the next version of Internet Explorer might have an extension that knows how to talk to these lightweight home control things. Let me go back up. These people decided to take the most commonly accessed items, 18 of them, and build what they call a starter kit. Here's an example of a smart object. They build models for how to talk to a fridge, how to talk to HVAC. That means hitting, vacuuming and air-conditioning control in your house, how to talk to random digital inputs to random actuators. I'm not going to explain the French stuff. Specifically, it talks about general I/O, generic sensors, how bright is it? Is somebody nearby? How warm is it? How humid? Measure power in some device. Sometimes you want to measure what is the load behind the power source. You want to control lights, course. You want to turn power on and off. You want to measure acceleration in some things. You want to measure magnetic fields sometimes, measure barometric pressure. These are things that are in the starter kit. The people developing them have actually been working on probably 100 different devices and again, if they are doing such a good job of doing 100 devices, I don't want to have to do it a second time. To take just one example, drill into digital input, it wants to know the current state. It could have a counter so it could see how long it has been in the current state. It could say what polarity it is. I might want to specify a debounce period. How many of you have designed embedded systems where you have to deal with what is called debounce? I see a few hands. Good. I used to work in video games, 35 years ago. That's where I learned about debounce. That's why I'm a fellow, and yes, you have to be very careful about how long people are pushing the buttons. And whether you believe it or not, you can provide for edge protection. You can reset the counter that is keeping how long it's been toggled and there can be application types like people are putting air pressure sensors and cars. In my car I was driving around and the sensor went on. It turns out that it was a sensor failure, but it started telling me that I had an air pressure problem in my car. We should get used to that. Any questions about that before I proceed, because I don't want to run over. I think I have enough time. Here is more detail on the things that I showed you before. Here's the radio part. Underneath this is connection oriented channel. Then we started acting that to the IPv6 and then above that we put Co-AP which generates UDP traffic and above there I put an application that is trying to drive a smart object of any kind. >>: Hey Joe. On that slide where would the lightweight M2M transport fit. >> Joe Decuir: As a matter fact, lightweight M2M maps directly into IP smart objects. I've got a copy of the document and this implements this thing in lightweight M2M so it's an attempt to bring everybody together. Good question. I think I went back to this already. I should move it to a different place in the next slide deck. Like I say, there is a lot in trying to tell us all how it's all going to work. At CES show a month and a half ago the chief executive of Samsung, which I'm sure all of you have heard of by now, he says that in two or three years 80 percent of all Samsung's products will be internet enabled for the internet of things. He says all the work on standards has not succeeded so far. We need help. We can't do this alone. If I knew who he was I would write him a fan letter because that's exactly the right attitude right now. He needs the help of all of the rest of us. There will be Darwinian evolution. When you have a lot of choices customers will be attracted to something that might be useful to them. They'll keep it if it connects to things that they care about and then they will not throw it away if they don't discover that it can't be trusted. We'll support all these use cases that I talked about above because Microsoft may decide to take a technique and bring it to one group and CSR may take a technique to another group. Nokia, for example, has been one of the people trying to drive IPSP. They'll take that into IETF. They'll spread it however they're are going to spread it. The Bluetooth SIG itself is not going to be telling us how this works. I think I left out my favorite cartoon. What I did instead is give you three slides worth of references. I'm providing this so you can go find all of this stuff. IPv4 and IPv6, how they work, network address translation, where you can find the engineering task for different kinds of radios. That takes too much time to explain. I was talking about health standards. This is the 11073. AllSeen Alliance is a group that Microsoft is a member of. Machine to Machine Connectivity, Open Mobile Alliance. Someone was talking about lightweight M2M. Smart grid is an IEEE standard for how to control power usage in homes. And there is that starter pack. I invite any of you that are curious about this stuff to download that and see what it has to say. Here's how to find me. If this was my own computer at which are you a couple of cartoons. One of them, a few years back Dilbert had a cartoon where Dilbert's boss gives him something that is basically a Fitbit. Gilbert is so excited that his boss cares about his health. In the last frame Dilbert and the cat are saying this guy is not breathing hard enough. We should give him some more work. [laughter]. Get used to the fact that all kinds of people can get access to your information and they may not be friendly at all. The other cartoon that I like is about 15 years old now about networking houses and refrigerators and all this. The man comes into his kitchen and his refrigerator has an eye on it kind of like Hal 9000, and he says Dave, give me a beer. And the refrigerator says the scale in the bathroom is reporting disturbing flab abnormalities. I can't do that. And the guy is saying, I don't want my equipment to be smarter than I am. But imagine someday our houses will care about our health so much that they will be intervening constructively. You know where to find me. Yes, sir? >>: Just describe what the current version is of Bluetooth and also where beacons fit in. >> Joe Decuir: Okay. Apple and some other companies are making use of beacons. Beacons are about where it is something. A Bluetooth smart server will generate advertisements. There are three channels that are chosen to avoid all of the Wi-Fi channels in the 2.4 GHz band and so a device will be saying I'm here. I'm here. An older version of that technology, who has heard of DoCoMo. DoCoMo is to Japan what AT&T is to the United States. DoCoMo makes a whole lot of money off of beacons and commercializing where things are. They had this problem a few years back because they wanted to do indoor location, but GPS doesn't penetrate the ground. Shinjuku station in Tokyo is the largest train station and the largest shopping center in the world and DoCoMo wanted to put beacons that would radiate the message here's where you are. Here's where you are. Here's where you are and here's how low you are or how high you are so they could tell which stories you were near and they would have a separate record of what the specials were at the store at this time. Apple has been popularizing iBeacon, but they didn't invent it. It was invented in Japan. All they do is they take these things, stick them on the wall. They are powered by light, because the power is low enough that they'll run basically forever on the ambient light in the place. And they're programmed with their location and all they do afterwards is say I'm here. I'm here. I'm here, and they run forever. They don't have to be maintained. Just don't cover them with paint or they will stop working. >>: The notion of doing different device types over IP over Bluetooth seems to be in some sense competing with the notion of Bluetooth specific profiles that don't have IP in the middle of it. So how do you see Bluetooth SIG going forward? Do you see that is being, that they will continue doing that competing? Do you see that as a dead-end, or what you see looking into the future? >> Joe Decuir: I don't know how it's going to resolve, but I believe there will be more than one choice. In fact, I am preparing to make a proposal within the Bluetooth SIG that what the Bluetooth SIG should do is take the IPSO starter pack and create exact GATT equivalents for each of those 18 items because that means that if I'm an applications writer and I want to write to IPSO and a translates to a gateway which is near a bunch of low-energy devices, if they don't have IP in them directly they can convert directly into native AT in GATT, [phonetic] and drive the devices and the devices will be lower power because they'll be lighter, but the application vendor still sees the same thing. I think that might be a winning strategy, but I have to convince some other people to agree with me. >>: So your point is they have to use the same schema and then transport can be less. >> Joe Decuir: The same semantics. Schema is the right word for it. There was another hand? >>: I think there was another one before me. One of the things I've been finding is a lot of people expect hardware to be achromatized and you don't make a lot of money on it so they tie it into their service and they figure the money is going to be in the service so you are working for open standards where anybody's service and anybody's device and anybody's app can now work together, I want that. The problem is I am finding that Wink which is available at Home Depot, yeah, we're kind of open in a way if you can talk to us in a way, but everyone wants to own that cloud service because they think that is where they are going to monetize. >> Joe Decuir: That's right, because they want to hold on to something that they control, something that's unique. I used to work at Microsoft. We wanted open standards to apply to everyone except us. >>: They want to apply the implementation of the open standard. >> Joe Decuir: Twenty-two years ago I got away with telling a joke to Bill Gates. So Bill, how many Microsoft engineers does it take to screw in a light bulb? He says, I don't know. You tell me. I said none Bill, because you would declare darkness a standard. [laughter]. He gave me good stock options later, so I didn't piss him off too much. >>: What I is all of these vertical stacks that frustrate me because I'm trying to build out a smart house myself to do these things and my video system is different. It's got its own stack. My alarm system has got its own stack. My Wink has got its own stack, so I'm building, I'll talk to anybody that I can get to use Node.js. >> Joe Decuir: This is going to be complicated for a while, but to step back, Samsung has joined the thread group and the thread group today uses 6LoWPAN and things that look a lot like IPSO over them. I was talking to Samsung at CES a month and a half ago and they said this one, the only tested parts are 6LoWPAN today, but we've already got Bluetooth low energy in it and we're going to try to sort it out and make it open 6LoBTLE also, so that's the direction they're headed in. Samsung will still add some kind of proprietary value on top of what they do and that's okay because people want to have a profit motive. I mean that's what pays for this building amongst other things. The boundaries of what is proprietary and what's public will get pushed back. I would suggest that in this space some of the winners will be the ones that are more open than the ones that are less open because the ones that are more open will attract more support. Microsoft learned that big time decades ago. Bill Joy who was a college buddy of mine said you should design your company on the assumption that not all the smart people in the world work for you. So they built, not an open source, but they built an open API Windows system and working with Intel they built the PC industry which still has life in it. When I look around the world the people who have at least open APIs if not open source attract more support and you all know the success of Linux and you might know that AllJoin is built on remote procedure calls in Linux so it's going to attract people to add value on top of it. >>: You kind of mentioned it in one of your former slides that Wi-Fi has 82.11, I remember. Is that true? >> Joe Decuir: It's in development. It's going to be a 6Lo system over 82.11 as well. >>: How do you see the future of coexistence between them? >> Joe Decuir: If the applications are disconnected from what the radio is, it's better for the application, and so there are some applications for which I would say Wi-Fi is a better use than 802.15.4 or BTLE. Anything that needs a lot of data or that needs the hundred plus meter range, then yes, you want to use that. If the same 6Lo IPv6 lives overall these radios, that's attractive to me as an implementer and it's attractive to me as someone trying to build a system that meets actual market needs. Does that make sense? Speaking of which, another friend of mine who is also a distinguished lecturer in the Consumer Electronics Society wants to come out here and talk about his ideas about home networking and whatnot and I would like to know who to connect Bob Frankston with here. Any volunteers? Anybody who works on this stuff here at Microsoft? >>: You can send him to me. Actually, the financial industry went through a very similar thing trying to standardize the communications between banks and so on, and they finally went to EDIFACT and that's pretty much the international standard now. There were a lot of basic principles that are completely missing from what I here, the whole concept that as the protocol advances you should send a code which tells you which version you are using and it should be backward compatible so it's possible to upgrade parts of the system without upgrading the whole system. >> Joe Decuir: Does EDIFACT to this also? >>: Yes. >> Joe Decuir: That's standard computer design. >>: If IPv6 had compressed addresses, much of the complexity of this stuff would disappear, wouldn't it? The complexity appears briefly when things are found and connected. The 6Lo border router recognizes what happens when things are first discovered and hooked up and thereafter, things are compressed. Does that make sense? It occurs infrequently. If it had to happen in every single transaction, it would be a much bigger cost and it wouldn't save you anything at all. Does that make sense? >>: You have to deal with the whole IP address when you are configuring it and then after that [indiscernible] >> Joe Decuir: That's correct, so it's a lossless compression if you understand the difference between that and lossy compression. >>: [indiscernible] IT computers 1620s and 1401s and so on it had a variable length number and they never had the line as low as it could be because there was no limit. If it went to a coded system like that, you wouldn't have these issues except if you had to big of an address in memory or something. >> Joe Decuir: A lot of the routing that's done in the internet is done by really optimized fast hardware and that optimized fast hardware knows about 32 bit and 128 bit addresses and not a lot of variability in between. I can see building all of that out of plane transistors. >>: You've still got the [indiscernible] 64, right? >> Joe Decuir: Another question? >>: I have two questions if I may, Joe. One will Thread support IP smart objects and two is the OMA device management a smart object? >> Joe Decuir: Yes. In reverse order, there are explicit references in IPSO for OMI lightweight M2M. That's why all of those are listed in the references. And the fourth is I have found the people who are working on the Thread stuff for Samsung and I'm trying to find out from them whether they're planning to use IPSO or not, and the answer is I have no idea. I'm trying to wait for them to get caught up on their e-mail. >>: You obviously, on Sunday night there was 60 Minutes and Leslie Stall was in the car and they hacked and so she couldn't stop her brakes. That can be a big liability. Is there some way we can deal with this to have security so that nobody can hack in? Or your houses, so they can open it up? >> Joe Decuir: Doors are being open before locks are being installed. >>: [indiscernible] before your things get abstracted. >> Joe Decuir: Take a car and connected to the internet. That's a door. The question then is where are the locks. One of the comments that was made because I saw the same 60 Minutes is that the manufacturers differ in how they handle that. I would say at this joint if I was going to go buy a new car, I would ask them how they control this now and I wouldn't buy the car if they didn't have controls built in. >>: It is inherently insecure. >> Joe Decuir: Duh! Yes. I've been talking about nobody wants their body hacked and nobody wants their house hacked and this is an example of nobody wants their car hacked either. We have been connected for a while. Many of you are familiar with OnStar by now, so you have had at least 2T level connection to General Motors for a long time if you've had one of those cars. And advantage is if you get locked out, you leave your keys in there, you call OnStar and then lock them for you. They can tell when your car is bidding stolen and they can cause it to slow down and come to a stop remotely if it's been stolen. That's a nice feature. That's something that attracts people, but the doors open as soon as the internet connected it to anything. I'm guessing that the OnStar system is GM proprietary and not an open system, so that there is no way directly for hackers to get into your car, but people are now starting to connect 4G to your car and Wi-Fi within your car which means it's now got the door open. And I don't know exactly how they're going to solve it, but as I say, I would ask them this question in the showroom floor. >>: [indiscernible] going backwards in a way. I mean IPv, the reason why [indiscernible] other radio technologies have adopted IPv6 presumably is because a lot of people know that it's the common language. Upstairs the common language is HTTP. There's a new version of it. You probably talked to a lot of people who are very interested in that layer. Have you heard of people saying rather than Co-AP we would rather use HTTP 2? >> Joe Decuir: Or 5. When I was mentioning the ways that the Bluetooth SIG has, the first way is HTTP and a RESTful API that takes, if you build a browser with particular capabilities and explain to it basically how to drive a Bluetooth low energy device directly, it converts those things into HTTP gets and puts, sends them through the gateway which then converts them into native stuff. When I was talking earlier about how I would deal with the conflict between IPSO versus this native GATT, what we would wind up doing is building basically the GATT RESTful API version of those made of things. From the far endpoint they look the same, but they're lower power lighter at the far end. I think I got distracted, but yeah, we're working around HTTP now as well as Co-AP now. What I didn't put up there is I have written several articles on the subject and there's another one that's going to be published in April and basically it's this presentation and it's got a lot more references. Is that a good enough answer? All right. Thank you for hosting me here. I have cards if anybody wants to find me. [applause].