Carterfone

Transcription

MALE SPEAKER: So it's my great pleasure to introduce Bob Frankston, who imagined a world with a spreadsheet before computers really had a useful thing to do. And so it was world changing, and I think what's really interesting is when I met with Bob a couple weeks ago and we start talking, we immediately got into a really strong agreement and a really good argument all within about four minutes. And so I thought, that's the kind of guy I want to talk here. So thanks for coming, Bob. I really appreciate it. BOB FRANKSTON: OK. As some people here know, my history goes back well before VisiCalc, which is actually Dan's idea. I will go through history, but let's start the talk. Hit the space bar. That's a good way to move forward. So I'm going to start at the end point. Both of the talk, where the high-level issues are, and the sense that I interested in the endpoints of connections, and not all the network crap in the middle. So the internet itself is just a technique. It's the way we can use whatever is available to communicate. The problem is that we have all these people who think the internet is important. That's thinking that there is a thing called the internet. And they sort of have to defend it and make it work. And for me, it's just that stuff in the middle. I just start with what you do with it. And you mentioned using VisiCalc, some people at the St Student Information Processing Board at MIT, which I co-founded in '69, trying to get people to use computers back then. I had a sense they were useful. Nobody believed me. And this is that actually the culture, the Project MAC, which is in the building that's inside one of the Novartis buildings now, which is the attitude of how do we make this technology useful? What can we do with it? So I'm skipping outside my talk, but for me the slides are mainly a guide. They try to keep me on track. There will be lots of points we can discuss later I'll just mention. One of the big insights is the least important thing we do is exchange content. Now that might sound heretical, but how many people have ever sent a copy of Wikipedia? You just say the word Wikipedia. You refer to it. You never send it. And that's part of the secret about what makes this work. Yet all our policies, all our measures of broadband and everything, running out of internet, are based on the idea we're shipping exabytes of data through this one router, which must be in the middle of Kansas City. But for me, the internet is really about opportunity. How do we create opportunity to do things at the-- well, there's no real edge, but outside of the net? And the other important point is that the internet is a discontinuity. When you read the classic histories of in the internet, they start with the ARPANET and have a packet radio. No, that's nonsense. Yes, the excellent technologies were there, but the interent really started from a very different beginning. It was the revolt against networking as a service. You didn't need a third party along the way to read your telegram like a scribe and repeat it down the line. You do it yourself. You just exchang these meaningless bits. And one of the really horrendous legacies we're still stuck with is the idea the carrier in the middle must make a profit for you to do anything. And I was reminded of this. I understand why companies need to make a profit, the question is whether that's the right model. For example, airlines. I don't have a better idea for airlines. But the consequence is I'm not going to be able to get to Los Vegas tomorrow because airlines no longer have enough extra equipment. So basically, they canceled my flight because a gear was stuck somewhere and there's absolutely no resilience. The whole system shuts down. And maybe airlines have to do it., but for us to use that technology for networking is absurd. So as I said, the internet is a technique for doing things ourselves. Now, I put in DIO as a broader point because doing it ourselves doesn't mean you do it alone. It could be doing it ourselves if you have communities that work together. But it's easy to just think of DIY as opposed to depending on a third party. Do it yourself. The other thing that we're going to get to is once meaning is outside the network-- oh, I just realized you're not seeing my slides. MALE SPEAKER: [INAUDIBLE] slides? BOB FRANKSTON: Yes, there are slides here. I'm more coherent than I seem, because the explanation is up on the slides. Speaking of depending on a third party and not having eyes in the back of my head. Ah, there it is. I won't go back to the other pictures, nor show my beautiful picture of the trolley in the snow. That explains why you were actually looking at me and seemed to be awake. I can go all the way back, but that's not important. So there's where you see DIY versus DIO. You wondered why I was-- OK. Do I seem more coherent now that you can see it? OK. So what does this mean? This is really about taking advantage of opportunities. And we're limited by discovering what we can do. For example, voice over IP was a discovery. Yes, we could get to work in some cases in the early studies, but how can you do voice over a-- when you lose your package, you got high jitter. Just like TV. People thought we need asynchronous TV. But some people were creative and figured out ways to take advantage of opportunities. Now it's not always obvious. The reason VoIP it works now has nothing to do with making VoIP work. It was the web that gave us the capacity. If we try to make VoIP work, we actually did. It's called SS7. Does anybody want to buy a used phone network? It's too expensive to maintain. But the key is to have an appropriate business model. And this is one of the main messages. So we don't need a third party the middle like the airlines do. Have to run at the edge, to only do what's profitable. It's sort of like depending on the railroad to get to a destination. Not many railroads are willing to tracks up Mount Everest. So we want to get away from this idea of the provider making profit for us to do anything. The other problem we have with the current business model is that providers' worst enemy, the people they fear the most, the existential threat, are the customers. There's a customer buying from Netflix will remove all the profit of selling content, as the cable companies do now. And there's a slide show I point to on my site from the cellular industry explaining that abundance is the greatest threat. That if they don't prevent abundance, people are not going to buy the services. So if you wondered why there's a problem with broadband crisis, it's obvious. Giving the customers enough capacity is a losing idea. So this is not sustainable. We need to have a sustainable business model, and this is, again, Google, since you use connectivity, to encourage capacity. As I said, we exchange references, not contact. Capacity is not the same as bandwidth. It turns out we have so much capacity, we're using, at the first approximation, 0% of our capacity. And that's what's a conflict of interest. So there is an alternative model, which is the way we fund sidewalks. The sidewalk owner does not view the user as a competitive threat. But people talk about privatizing the Mass Pike. Then Route 30 becomes a competitive threat. We already saw this in Detroit. There's a private bridge in Detroit. They try to keep the city from building a public bridge because it's competition. So you've really got to be careful about what incentives we create. Now the resource I describe as ambient connectivity, and I use this term for the idea of wherever you are, you could just assume connectivity. Now this is sort of like having open Wi-Fi everywhere, except I don't want to tie it to any particular technology. Current Wi-Fi is much better than it might be, but it's still a problematic technology. But the key is to be able to assume we're connected wherever we are. I'll get more into this. I add the word permissionless because people somehow seem to assume you've got to show your credentials to prove you can pay for it. But how many of you have to show that you're authorized to use a sidewalk? That's the exception, not the rule. And the thing to realize is this whole idea of networking, we so implicitly have a provider-centric view. We talk about the network as what's important, and how you maintain the network. And as we'll see, our language, our framework is infused with a provider-centric view of protocols and everything. What you have to do is forget about the network. I want to be able to think just about the relationships between the two endpoints and what I'm trying to accomplish. It should just work everywhere. Or you shouldn't have to log into this or connect to this. Yes? AUDIENCE: One of the big struggles we've had is that users want to be able to define what they get from an ISP, and the ISP is usually unwilling to commit. BOB FRANKSTON: We shouldn't have the concept of an ISP because the interent is not a service that's being pumped out of the salt mines in Louisiana. It's what we do with the available resources. The internet is a technique. That's what I'm saying about flipping the model. People assume they have to ask the ISP. No, you're important. The ISP should come to you and say, please, I want a job. Can I do anything for you? I'm very interested in how language works, so I have to use language people are familiar with. So I say open access, even though words like that are still provider-centric. One of the big words we use, terrible words, is broadband. What is broadband? AUDIENCE: Effectively, a bigger pipe than the other. BOB FRANKSTON: Well actually, it's a technical term that comes from radio technology. It turns out the companies that extended the TV antennas up the top of the mountain took the broadband over the air signals and put it on a coaxial cable, versus baseband. And that became-- we're going to give examples later-- associated with a business model because that's the way language works. And then it became literally a contronym. Literally itself has become a contronym. A contronym is a word that's its own antonym. And the way language works, literally fills in the same slot as figuratively, so people tend to confuse the two. So broadband we use to mean the internet, when the internet is really what you tunnel through despite the broadband business model. This is what makes it so hard to talk about this. And I have to emphasize, I'm talking about an architecture, not any particular implementation. So when I talk about having open Wi-Fi, that's just an example. It's one way to do it. What amazes me is how often it's, oh, we'll give you wireless mesh. First, why do I say wireless? A wire is just a wave guide for signals. There's nothing wrong with wires. And mesh is just a particular technology. And if I've got a fiber that takes me across the country, why do I want a bucket brigade? It's an economic problem to solve, and technology's just a means. What we do have to do is preserve the magic of it just works. And that's the real architectural goal. That's the power of architecture. Network just disappears. And apps can use any available means, including the QR codes. So if we're just exchanging a reference, I just give you a QR code. So one way to explain this is our view of the history of what I call digital. Now when I say digital, I mean discrete symbols. It doesn't mean 10 fingers. But the fact that we can have these sharp distinctions and you can regenerate the meaning. That's the real value. The zeroes and ones are going to be regenerated. But it's also alphabets. If I write letters, even-- . pictographics. Hieroglyphics were very distinct symbols. Little off, you can guess what it is again. And that allowed the message to be preserved over distance. You can carry the message. If it got a little distorted, you can correct it. Smudged. But also over time. And we actually had some forward error correcting oral traditions, but that's for when you want a discussion of details of language. But one of the legacies of the early days of inarticulate digital is the telegraph. We had to use the wire to keep the single contained, to maintain the relationships between the bits, the endpoints. So it flooded our skies with all these wires. Every call needed its own wire. And you think that's bad? Even when we had wireless, no wires, we created virtual wires. This cockamamie idea of single frequency signaling creates scarcity. It's as if we said no two people in the room could wear a blue shirt because it'll be confused. Interference is not a property of physics. Waves pass through each other. Interference is another word for confusion. So whenever anybody talks about interference, they're just saying we're creating confusion. If you use all the rich data, and in a sense, if you look at meaning come from context, big data's fundamental. Information in isolation is meaningless. Yet our telecommunications system is trying to carefully preserve-- this is the inarticulate digital. We have to carefully preserve all this stuff because-- it's a great hack, by the way. The tuning fork was a wonderful discovery back then. The fact it's been used to compromise the Constitution and prevent free speech, well, collateral damage. The other reminder is that business models-- again, like broadband, the business model came from the accidental properties of history. The single frequency single hop signal gave us the business model of radio. And if I make it to CS this year, I ask about radio, I've found Pandora is radio. Tubes and transistors are no longer radio. Beware the language. The talking telegraph really created havoc. We had this great technology. Digital can cover any distance. And then this guy across the river there disobeyed his father-in-law, stole an idea from Elisha Gray, who retired to West [INAUDIBLE], and created the talking telegraph using analog technology, and made a mess of things. Distance now became difficult. Distance used to be easy. Telegraph would go over any distance. Distance became hard. And with that explicit understanding, we could tune things, we could tune knobs and all this stuff. But it was very hard to preserve a signal across the street, let alone across the country. So this guy Claude Shannon did a lot of research, created this whole technology for how to preserve a signal within wire. Within the channel. He didn't use the term wire. He used a channel. It wasn't very abstract term. He called it information for the same reason physicists, we'll point out, like to use words like work. Well we have giant brains, all this notation, anthropomorphic notation. We use the word communicate as if these wires carried our meaning. And we're still living in that legacy. Matter of fact, here it's more than a century after that picture I showed you of the wires covering all the streets, yet it's even worse. All we've done in a century and a half is learn how to take all those wires and bundle them up very neatly into little tiny packages. And worse, we have multiple bundles and wires from competing bundlers, or bunglers, whatever you want to call them. And it's even worse. Inside each of those we have virtual circuits to maintain each conversation. Now how you can maintain a business model of carrying bits where there's no longer any way to compete. In other words, they're just bits. They're generic bits. How you differentiate. But the reason we have the legacy is we're still stuck in stage two, the idea that you need pipes, pipes, pipes. And it comes out in many ways. One is carriers don't want to become dumb pipes. Sorry. That's not the worst of it, is the fact we don't need pipes anymore. Bits are all self-identified, meaning it's assembled outside the network. We don't need pipes. Yet our business models are still pipe-based. Now the digital technology that gave us computers did set the stage for the future. But we first have to-- again, this is why I brought up the issue of language. How many people think physicists should run the NLRB, National Labor Relations Board, because they understand what work is? Why would a double E be allowed to run a school of communications and the PR department? Those people understand communications. Apparently E's don't because they confuse information in the Shannon sense. How much information is there in not answering a question? Or in a pause? There are people who spend a lot of time figuring it out, and you have to slap them and say, no. That's a stupid question. Now this sets the stage for what I'm calling stage three. Now we're very interested in [INAUDIBLE], but I took a class in spring of 1973 from Mike Hammer. Does anybody remember Mike Hammer, reengineering? It was studying networks. So one of things we studied was ALOHAnet. How many people here know about ALOHAnet? OK. It was a radio packet network in Hawaii. Do It yourself packet network. There had been various experiments integrating packet networks, but I remember ALOHA was the one we studied in class. And basically it was what any hacker would do. It looks at a packet and resends it. No big deal. Maybe it's the latest temperature. You don't resend it. You just give them temperature. I mean, we called it a net. It was ALOHAnet, but it was a network in the sense Facebook is a network, or, sorry, Google+ is a network. It's a social network. That's the problem with language again. This is why I'm so fascinated by language. The word network is so general as to be totally useless for communicating. Worse than useless. It sort of like diplomacy. You can both use the same word, but mean entirely different things. But you think you're talking to each other. And at the end of the class, some other kid sitting next to me-- this is May 1973, did his baseband project. He took ALOHA and put it on a coax. Now I wanted to do it over the campus network, it would have been broadband. But I didn't follow through, so it doesn't count. But it was obvious this is a great hack. 2.994 megabits, as he reminds me. That's Bob Metcalfe, you know. It was wonderful. But to get his Doctorate he had to prove to the EE people that it would perform well. Turns out he was wrong. It performed much better than his thesis said. So how many of you are familiar with the paradigms framings? When you change a paradigm, everything changes. So when Copernicus did his little thing of moving a reference point 93 million miles, it was a nice mathematical hack, and it did make navigation easier. But it had two side effects, which he wouldn't have anticipated. One, it destroyed people's idea that they were the middle of the universe. And that made it very hard to accept. And I'm telling you meaning is not fundamental, if you want a scary concept. The other thing it did was give Newton his physics. Because suddenly you could see a uniformity out of all these disparate things, and you just get a deeper understanding. And in that sense, what the packet radios forced us to do was learn how to preserve meaning outside of the network. Discover what we can do with the opportunities. And that, I argue, is the real origin of the internet. Learning how to solve problems ourselves, learning the techniques for using any available resource as a way to communicate in the human sense, not just EE sense. And that's why I say a fundamental discontinuity, yet we're still thinking stage two because, as I said, language, information, communications. It just sounds just like the railroads of the 19th century. And in fact, we have a thing called the Federal Communications Commission. What do they mean by communications? What's worse is they're explicitly spun off from the ICC, Interstate Commerce Commission, which managed railroads. Now that's gone. It's been replaced by the Department of Transportation, which manages the means we use to travel. It does not manage the actual freight itself. In the meantime, the FCC is there telling us what words we're allowed to use, how we're allowed to run elections. Really weird. You'd think these people didn't understand how to communicate. And we're still compromising the First Amendment by creating scarcity in order to keep the business alive. So the internet happened when we're interconnected the LANs. The LAN, you were , anywhere, plugged into the coax, you were just there. But the consequence of that was you could not see beyond the individual packet. So therefore, we had to have a system-- and this is the real insight-- to separate TCP from IP. And it wasn't just Vince or Bob Kahn. There were a number of others. David Reed locally, and Dave Clark, and all these other people. How do you preserve an application relationship independent, when you could just exchange nothing above packets? And this where the end-to-end argument. You had to have meaning outside the application. The earlier slide said TCP, mentioned explicitly. What TCP is, it says transmission control protocol. Very interesting. Because if all you want to do was make sure all the bits got through, you keep retrying til it gets through. But TCP was also protocol for cooperatively sharing the resources in the middle by detecting congestion, by backing off. And what's amazing is how well that worked. That it's really a social protocol for sharing resources. It's not just a reliable stream protocol. And basically, yes, you can violate it, but if more than a few people violate a collapse, you get immediate feedback, people used standard libraries. So it's a whole interesting discussion about why that works. And sociologists should study this some more. And I won't go into problems like buffer blow. There were lots of other issues. So where I come into this, and this much prouder than doing VisiCalc, was home networks. The original triple play plan, or n-tuple play. [INAUDIBLE] is a good example. They would haul a pipe into your house. They would put an a ATM card cage in the basement. ATM was asynchronous transfer mode. The base for DSL. I'll try to avoid too much acronym soup. By the way, ADSL is interesting. You know why DSL exists? For interactive TV. It was not created for the internet. That was a later repurposing. There's a lot of repurposing discovery. Hindsight history makes things look so purposeful. People have a reason for everything. But if you remember at the beginning when you had to run a separate wire for reach relationship, well those wires, virtual wires were inside a pipe or any other triple plays. Each phone line was delivered to the card cage. In the early days, remember, the phone companies owned your phone. The service was the conversation. It was not the wires coming into your house. And therefore in the 1950s, AT&T had to be sued to allow you put a box around a phone. it was a kind of device called the Hush-A-Phone compromised the quality of the service and dispersed their reputation. Later the Carterfone allowed your equipment on, but the Hush-A-Phone, I think was a more interesting case. Where the boundary of the service is. And you can see the example of how that works in practice. Carriers would want to charge for each IP address as if it were a phone number. You're going to pay for each IP address. Since they looked inside your house, they can do things like trying to ban webcams because those were an abuse of their service. It really [INAUDIBLE] home. Repurposing a video distribution service, rethink connectivity. And today, the modern version of that is called Google Fiber. And Google Fiber's attitude is you can share. They're telling you what you do with the bits. Now what company would be so evil AUDIENCE: Are you referring to the no servers allowed? BOB FRANKSTON: No, that the least of it. AUDIENCE: That is [INAUDIBLE]. BOB FRANKSTON: No I understand. Verizon had the same attitude. And the nice thing about terms of service-- I'll be nice to the FCC, so I won't say fuck them-- but back when I was doing home networking-- and I'll get into that-- I basically ignored the terms of service. Because the phone companies said you're not allowed to add more than one PC to your home network. To the router. They're going to tell you what you can do with the device inside your own home. AUDIENCE: So for the Google Fiber no sharing, you mean you can't [INAUDIBLE] community over to watch TV? BOB FRANKSTON: No, it's not just watch TV. If I have a Google Fiber programs, I'm not allowed to share that connection with my neighbor. Now I can understand the business model point of view, which is why you need a business model that doesn't create such terrible behavior. But are there people who try to bring, like housing communities. You want to bring broadband to the poor. You got a gigabit. That'll serve complete apartment houses. Steve [INAUDIBLE] here uses just a few megabits to serve the [INAUDIBLE] project. So that it's as if you-- yes? AUDIENCE: Yes. [INAUDIBLE] one gigabit. BOB FRANKSTON: No it's not. You play the odds. Sure. But let people discover what it does. AUDIENCE: Providers tell you up front, yes, you may have one gigabit of bandwidth, but you only have 5%. BOB FRANKSTON: Yeah. Let the people decide and discover that. Sure. But it's sort of like a sidewalk. We're going to limit how many people can live in the house. This is where you have the idea of discovering what you can do from the edge. I don't want guarantees. The reason Skype can do video is because Skype does not promise to do video. If they promised video, they could not do video. There are a lot of counterintuitive aspects here, which is why you have to start thinking from the resource and from outside. And you also have some interesting things, like if you're above a threshold, you've got 10 people with 10 connections, you get 100 times of capacity. But basically, if you have a lot of DSL wires, you share them, you get disproportional improvement. It's like a one-lane highway will get more than three times the traffic jams of a three-lane highway. Because you start to use the resource effectively. You fill in the holes, you accept delays. And it all depends on the application. We're just extending the word Wikipedia. Who cares? A gigabit is wonderful. Even if you've got a 10 megabit. Let me share. Let me decide. But it affects the business model. That's why I keep coming back to the funding model. I'm not saying it should be given away, we need funding models that work with societal needs. Yes. AUDIENCE: My one big sticky thing here still is legal issues, including intellectual property enforcement. BOB FRANKSTON: No. But that's not in the wire. The bits are bits. They have no inherent meaning. This is a homeopathic theory of communications, that bits have a memory of their original intent. AUDIENCE: We share a network connection that [INAUDIBLE] multi-family home. And one person who was visiting downloaded-- BOB FRANKSTON: Yes. And this is why lawyers are one of the greatest threats to society. Basically, you should move the trial to Salem, which is the appropriate venue for such theories. AUDIENCE: My point is, we do need, as a society, to-- BOB FRANKSTON: No. We have to come to the 20th century, let alone the 21st. AUDIENCE: But isn't the fundamental problem referring to the fact that you're trying to share the resource that has a huge level of [INAUDIBLE]? BOB FRANKSTON: OK. Let me get through this and I'll get to this other. The short answer is we're using 0% of the available capacity. So before you start telling me about a shortage, start telling me where that shortage is. When I've got three broadbands coming through my backyard and 1% of any one of them is my full internet connection, I've already got 100,000 times the capacity. So what do you mean? Why do you tell me shortage? Where's the shortage? When the police will run a fiber to their gun detection sound thing, which only a few kilobits you need for gunshot, but deny anybody the ability to share that. People now run fibers to fire alarms and things. That are terabits of capacity going unused. So before you run about scarcity, show me there's a problem. And the other example I won't get into is meter reading. Now meter bits are special. They can't go over the real internet. They need their own network so the phone companies can go back to triple-play and collect money for each meter read. And if you read the IEEE numbers about how many gigabits of capacity and everything they need to read the meter, they need a whole fiber infrastructure, gigabits capacity. Though if you run the numbers, you'll find out it's less than a bit per second. But you need fiber. We're really beholden to all these myths. So we need to start from the edge. And the other important point is local connectivity is more important than distant connectivity. We became fixated on long distance. Anybody old enough to remember when long distance would put fear in your hearts you'd go bankrupt on a long distance call? That's because of the stupid thing with analog telephony that Al Bell did. With digital it's no longer an issue. And even more important, the people we know and stuff are nearby. I was reminded of this. We're so hung up on social media, the network. My son last night was caught-- foolishly decided to take a long drive to Phoenix. And he was going to go from Indianapolis, or he was going from Indianapolis to St. Louis. It turns out that outside Terra Haute, traffic was stopped for 12 hours in the middle of the snow. Online social media is no information. If he had CB radio with him, and they almost took one, he could've found out from the truckers and everybody what was happening locally. And so you can get across the country easily to get to a website, but you can't send a message from your light switch to the light fixture, because that's still using technology Ben Franklin understood. We have TV controllers. There are so many problems with you have to be on the right LAN, you've got to point the IR at the right device. We do all this work at a distance, yet we don't realize it starts locally. And then interconnect these local islands is how the internet happened. And that's important lesson I built on. Actually, we're going to finish. So what I did with home networking is I made it opaque. I was at Microsoft at the time. Basically I prototyped it at my house. I did it for myself. And I try to do things that are generalized. When Dan Bricklin did a spreadsheet so he could do his homework, we just did a general version of it. So the NAT is a nice hack. So people couldn't peek inside your house. They couldn't enforce the terms of service. You could share all you want. On occasion they'll try to tell you, and the problem is yeah, some people look at the IP address and all sorts of how bad ideas. Accuse you of being a witch because your bits don't look right. But the key is to make it just so you can give people the tools to use the resources available. How many people worry about how many gigabits of video they're showing within their home? There's no cost. It suddenly becomes expensive once you get outside the home. So the benefits to the ownership is you get as many IP devices as you want. There's not scarcity of IP addresses, no meter running. And right now we ca run gigabits in the home. Remember it used to be ten megabits was very expensive. Modems used to be $300 for kilobits. But we're not quite there. We still have this perimeter security, these complex rules. Instead of abstracting security, making the application. We still think it's a property of the network. And these all long discussions in their own right. And we still don't understand generic bits. We have USB, we have SATA we have FireWire, we have HDMI. All these special packets, as if they're are all so special and different. I did this display port here because that's really an application protocol which we should be pushing, but instead we're stuck with HDMI, which is equivalent to the old digital telephone. Sort of faux digital. So with home control we can learn by doing. But people still think it's special because it's within a home. It's behind a NAT. When it's really just the example of a do it yourself environment. We have discovered, for example, want to solve a problem connecting devices? One thing from the edge is, v6 is a dead issue at this point. Instead we use v4 reflected to the cloud. Which I think is a terrible idea, but it works. It's an example of you go with what works. But I won't dwell on that either. Another topic. But we still use wires to define relationships. So how do we move beyond the home? The answer's very simple. We do what we did with the original internet. We basically interconnect this islands. We've already tunneled between the islands. When you get a quote, "broadband connection", you're really tunneling out of your home through telecom. They own the other end of the tunnel, but that's a temporary historical accident. We won't go into why it happened. But we can take the same ownership perimeter and expand it out from your home. So if you and your neighbors share, which some people are already doing, an apartment house, you can get many benefits. To get two major benefits. One is cost savings. You share the capacity. But the other thing is you can get permissionless ambient connectivity. You can now do all those IP-based devices. You can buy some now, but the problem with perimeter security makes them very brittle and fragile. AUDIENCE: It depends on what the perimeter is. In our building we run our own services. BOB FRANKSTON: Yeah, But do you have a WEP or other authentication? AUDIENCE: It's been years since I was using any of those services. BOB FRANKSTON: No, but the answers is-- AUDIENCE: A server you can [INAUDIBLE]. BOB FRANKSTON: Right. You have that, but what you want to do is at least be able to have devices just connect. I'm glad to see Google does not have an agree screen. But the problem is you get things like this at Jet Blue. Now can somebody explain why first I have to call my lawyer before I can call my lawyer? There's something wrong there. I've already the airline terminal. I've already agreed to much worse things they could do me, yet they have me type agree. This is something everybody should do what they can to get rid of those. And again, if you ask people fear, they'll say child pornography. They don't have any better idea, it's just they've learned this mantra. But if we paid for infrastructure like we do roads and sidewalks, we wouldn't have a need to shut things down just for revenue. AUDIENCE: Will you accept, especially in neighborhoods like where my house is, that there are thieves, drug dealers, all sorts of criminals using those public rights of way to get by with no record that they passed because-- BOB FRANKSTON: You mean you let people walk by your house without signing an agreement? And without paying you? AUDIENCE: Yeah. Well the ping is a totally different issue. BOB FRANKSTON: Well ask Rand Paul. He'd demand-- AUDIENCE: You have a very different expectation of how freely usable for potentially illegal activities a sidewalk is, compared to-- BOB FRANKSTON: And this is what's amazing. We're lucky by accident in the Constitution days when people were really pissed off at the king, so they wrote the free speech thing in. But every opportunity since, the fear is so-- it's spelled NSA. And yes, it's amazing how people fear what they don't understand yet. Which means you're not going to get any discontinuities, and you're not going to get any surprises. Yes. But an important part of the whole lecture, why the First Amendment works. Why things work despite the fact that crossing a street is too dangerous to do. So if we fund the infrastructure-- again, as I keep saying-- as you do sidewalks, these problems go. The reason for putting up a wall, putting fear aside, financial reasons go away. Because it's so cheap, you pay for it once like a home network. And then you create opportunity to discover [INAUDIBLE]. We haven't even started to explore what we can do. Because we can do high value applications. Like at home control. I don't know where my slide is, but home control is a wonderful example, because we just start to discover what we can do with connectivity. But one of the lessons I learned to total control is that we could do high value applications. We call them trophy homes. There are lighting scenarios, all these complicated high value stories. TVs coming down, the shades, all this stuff. We can't do light switches. OK? That's the problem here. Without permissionless ambient connectivity, simple things become hard. It's not just permissionless. The DNS, the current-- you must ask ICANN for permission. There are lots of other problems, but if we solve the financial problems, we then have an incentive to solve problems. We introduce resilience to the edge. You connect through an access point. You're walking down the street, you've got Comcast and all these others connected to, and then Google Maps stops working. Because one access point got stuck. Authentication chain, somebody broke it. The protocols of the edge aren't resilient. So there's a lot of work to do better protocols and have resilience of the whole network. Not meshy, necessarily. That's just technology. But resilience. So the old model, we assume value is created inside the network. Thus we could fund the network out of the revenue. And that's where the cable TV companies are. They own the pipes, they sell resources, just like railroads sold rides. If you go to Netflix that collapses, which creates the conflict. Fortunately there's a new business model, which is the one I keep saying. Just pay for the wires. They're so cheap. You put them in once. It's amazing how inexpensive. And if we didn't put the fiber in, we would have been doing gigabits over copper. DSL was running megabits 25 years ago. Now it's copper, right, so not pairs. You don't follow DSL rules. You be very innovative, you do the bundles as a whole. And at this point, one way to use copper is to sell it. Less copper and buy fiber. I don't care. I'm just saying the economics is what drives Moore's law. And then we won't have the need to exclude people for no reason other than forcing people to pay. Imagine sidewalks, the paywall around them. To show you the value of this, let's look at a case study. Philips Lifeline is one of the local companies there now in Framingham. They started locally. And initially they just had the radio from the device going to a phone and it would dial up. It worked well enough for the purpose. Hopefully the phone line was not in use, and all the other things. But they put it before, and you can get it to work enough times to be useful. Now people are going to the next stage, which they carefully set up connections to cell phone. If you set up the Bluetooth right and use the right phone, and you've paid your bill, and the matter of-- and you [MUMBLING]. It can work. But the problem without permissionless ambient connectivity is the default is failure. Shouldn't the default be to just work? And that's really what I'm getting to. The idea that we want just works connectivity. Now there are no guarantees, and this is why the VoIP story is important. Because we can get voice to work. It turns out Skype works much better over lines that the traditional phone companies can't use because their protocols are all brittly dependent upon the network doing everything right, preserving a path. Turns out cleverness can make up for a lot. So this is why so important are the incentives and create opportunities. And then there's a whole discussion, which I won't get into, is then we can start rethinking the current internet. For example, the Post Office, I understand, is routing far better than any of the internet people. Because it try to send mail to you. It sends mail to a place. That's its job. You tell your friends where you are. So I can go through, if anybody wants to, the other big topic we could do is where are the smarts. Smarts are valuable. So people want to own the smarts. But if you go into a restaurant, you don't want your wait person knowing what you're going to eat. You'd like to have some say in the matter. So I won't go into the challenges, but obviously there's a telecom industry which exists and to rightfully sees this as a deadly threat. We have an FCC, which is trying to transition to IP, as if they're desperate attempt to keep us in stage two. By simply saying IP is the new phone wire, but everything else is the same. You don't transition. You let the whole thing fade into the sunset. At some point nobody's going to care about it. You resell the facilities. And want does one do with lead acid batteries? How do you decommission them? There might be a lot of money. Somebody find a good use for lead acid batteries. You'd make a lot of money. AUDIENCE: They recycle them, I thought. BOB FRANKSTON: As what? Maybe they do, but I'm just saying. The whole internet thing is just in the infancy. Everything I read about it is very special hacks. There's no general connectivity. But then again, these long discussions-- I just wanted to introduce these as ideas at this point. I can also, if you want, go to the other slide, which I'm proud of, is how do we run out of the internet when it's just a technique? After hacking at night here, went to Union Square for French toast. And I was told we're out of a French toast. We asked, said are you out of bread, milk too? He said the guy didn't know how to make it. And that's really those wires are, is running out of internet. So any questions? Anybody confused? Anybody awake? AUDIENCE: So what happens to your do it yourself home small community network when it breaks? BOB FRANKSTON: What happens if your water pipe breaks? AUDIENCE: Well, I get it serviced. BOB FRANKSTON: Yeah. So we'll have the new plumbers. And in fact, I'm actually trying to find people who want to start. In other words, the point I'm making is I'm actually in favor of business, capitalists, and all this. It's just you've got to have rational businesses, and the model changes. If you want to read a book "Railroaded", by Richard White. It's a wonderful story because it tells how the railroads in the West-- this is opposed to the ones in the East, which have different densities-- were never financially viable. That's why you've got the FCC. Or subways in New York were never viable. Each infrastructure, you have to look at how these centers work, what's the business model? Instead of just continuing forward. So yes, if we're viewing a time and materials model as a business, we need plumbers. Of course they're a time and material business. Your plumber does not charge you for the cost of water at the price of whisky. When you buy flour you don't pay for the cake. So you have to understand where the values is added. Question back there? The first step, one thing to do, and people are already doing it, the first step is education and understanding. But if you're in an apartment house, get your friends and neighbors to basically make that a connected zone. Because petitioning the city is hard, because there's too much education, the language, everything makes it very hard. How can you even talk about the FCC when they were created to keep a business model alive? They're not a technology They were designed as part of a deal with AT&T to make AT&T viable as the stewards of our ability to communicate. Things have sort of gotten out of whack. And they also got a right to do it based on spectrum allocation. But you understand how hard it is to get people to think differently, but if you change economics, if your building-- and there are a million buildings-- so if you start to get your building to be a connected zone, and work out from there, repeating the history of the internet, you start to get to an alternative which is viable. And I think that's the starting. AUDIENCE: When you start to get, having been involved in small building connectivity zone efforts over a couple of decades, problems you have is still, as far as I know, too much expertise. BOB FRANKSTON: Yeah. If you want a problem, it's really a problem in California, the Bay Area. Too many people who understand too much. That's my parallel to put in fiber, ripped out fiber. So yes. It's a matter of venue shopping and set example. AUDIENCE: But more than this, plumbing isn't a simple thing. But plumbing was simplified so that it could become a trade. And by having it established as a trade, regular people can become people that maintain plumbing. You need that for networking. BOB FRANKSTON: Well actually, remember, I did that already for home networks. When I first started doing home networking in '95, think of the reaction I got when I said people are going to maintain their own home networks. So I did a couple things. First thing, remember I had this position having been in a class where we were doing this stuff. So I knew there was nothing hard there. Nothing secret. I just had to remove all the things that made it hard like configuring IP addresses and things like that. I've learned a lot since. There's a lot of ways they can make it even simpler. So yes. Part of it is this is why I talk about the connectivity of net routing I compare with crossing a river by jumping on icebergs. Turns out we have bridges. Just use the bridges. It's like the Post Office delivers to addresses. So insights like that make things much simpler. All hubs and routers should be self-diagnosing. But until we have a market incentive to do this, it's going to be hard to make that happen. Unfortunately, I was at Microsoft. I was able to keep them, and I purposely made sure they had no way to monetize it. To control it. But they did get value out of the connectivity. Just like Google has a value out of connectivity. The more people can connect-- even if 99% of the traffic is local, you can still start to offer all these new services and these new innovations. But I am actively looking for people to work with who do want to do business in both educating people on how to do it themselves and providing the appropriate expertise. So yes. I don't expect most people to want to do their own networking or build their own networks, but there's no reason that it has to be hard. We're just used to it. The same as DNS. We have this whole legacy because so much of today's internet goes back to the mainframe days, port numbers, all these issues, which are legacies. We've learned a lot since we first did the current internet. Yes. AUDIENCE: So, to [INAUDIBLE] idea up, you just have to make it easier. But you mentioned a problem being firewalls and such. BOB FRANKSTON: Yes. AUDIENCE: One of the [INAUDIBLE] benefits of that is I think at one time if you put a Windows computer on the public [INAUDIBLE], it would compromise in 10 minutes. Definitely less [INAUDIBLE]. BOB FRANKSTON: Yes. And when I first did it, what I was saying is OK, v4 closed the ports by default. But IPv6 or IPsec, we wouldn't be stupid. We designed it to be aware. Now it turns out we're still stupid, but that's another issue. But we can't depend upon permanent security. So now anybody know the Maginot Line in France? And how how well it worked for them. Now it turns out Andre Maginot actually was not stupid. He knew you still needed an army. But he made a mistake. He died. And the politicians then said, army? We don't need no army. We've got a wall. So people are building in these hyper things. But look at Google's problem. Chromecast is on the wrong network. The device doesn't work. So the answer is yes. We have to first have a better security model. Because when you go to Starbucks that same machine gets connected. That's based at the application level, the software level doesn't depend on the firewall. In my model, even if you have permissionless security, you still can have legacy stuff protected. So I'm not saying we get rid of all that. I'm looking new opportunities. AUDIENCE: One of the big incentives to have an apartment network, basically. A small community network when I was having a college-ish lifestyle, was that we could have our server serving the copies of our CDs to us [INAUDIBLE]. Now that isn't done anymore. Maybe it's done centrally. Do you see obvious ways that Google could enable that sort of community hub when we're otherwise serving. BOB FRANKSTON: It's interesting. One of the things that Western Digital is selling is a personal cloud, which is basically a disk in your house as a server, Pogoplug. A lot of people doing it, so I can easily see products. Because if you want to watch multi-screen 3D ultra high definition video, it would be nice to cache on a server right in the building. It's interesting. This is sort of a Maraschino cherry problem. Maraschino is you take a cherry, take all the cherry-ness out of it, then add it back. And in a sense, we've done such a good job of conquering distance, we have to put locality back in. But that's pragmatic. I mean, there are two reasons to [INAUDIBLE]. One is performance. And fine, put a cache server there. You can put that in a box. Chrome PCs should be able to work with local cache as well as remote. By the way, a little footnote. Be careful about the word cache. Sometimes you'll explicitly put it there. You want to make sure this information is there in case you get disconnected versus just for performance. You've got multiple issues. So you can easily have protocols that allow peer sharing locally. This is content-centeric networking. So lots of reasons for doing it. But you also want your thermostat, your other devices, to work within your house when your connection further out goes down. So we have to understand that the relationships are separate geographically. So there are reasons for having geographic proximity. Sometimes I want a light switch to say turn on the light in this room. And you might use techniques like IR not for the message, not for sending a message, just so you know what room you're in. Again, that's why I'm saying we have to rethink a lot of what the internet is as opportunity. And the distance stuff, yeah it's nice, but we forgot there's a physical world there. CB radios, we need to do virtual CB radios. So next my son is stuck on I-70 outside of Terra Haute, the truckers and stuff have a local social network, not just CB radios. Though it's probably nice that nobody else has CB radios anymore, knowing them. AUDIENCE: You made a comment at the beginning about something like most [INAUDIBLE] involved in sharing. BOB FRANKSTON: Well, we don't exchange content. We exchange references. Is that what you're saying? AUDIENCE: I don't know. I was trying to-- BOB FRANKSTON: When I say Wikipedia instead of giving you a copy of Wikipedia? AUDIENCE: Right. BOB FRANKSTON: Yes. You want me to expand on that? That's a very interesting topic. Because I used to give my talks and ask people how much do we compress the Encyclopedia Britannica? And I say to an ISBN number. And if you're a communication theorist, you want to fit it into the old model, you say, well the ISBN is a dictionary just like any of the coding we use is a dictionary. But remember, in communication theory, you have out-of-band signalling. You have an out-of-band clock, which enables a three kilohertz line to carry more than three kilohertz because of precise timing. But then I decided that's nonsense. it's too hard. It's really that we're just making references to things. And really that's what language is. You send URLs. Now URLs have URL problems. And then, so your effective capacity is meaningless. This is why clocking the number of bits of value is bad enough. It's already inversed. The longer the movie, the less valuable each bit is. But you can't even measure. Same movie can be a terabit or it can be ten bits depending on whether you just need a reference or you need a copy. My mother wanted to send a toy to her grandson in Seattle from New York. I said, fine. I'll have it there in an hour. I returned it to Radio Shack, gave my brother the number, and he bought it for him. AUDIENCE: The bit usage, bit used driver has been people watching movies and such. BOB FRANKSTON: It actually turns our that's not that big. That's what Akamai is about. They're cached locally. There are two ways it's not a big thing. Number one, it's cached locally. Number two, Netflix does the compression that the originally DSL people only envy. So at two megabits, they can give you a high definition experience. You don't really care [INAUDIBLE] picture. You want to enjoy the movie. A cute idea would be to put a camera on, and if you're enough feet away from the TV, they'll de-res it because you're not going to see the extras reses. When you come closer-- the point is there are no rules here. AUDIENCE: But my point about that is that is a lot of what people consume from an ISP. BOB FRANKSTON: Well, consume is an interesting word, but we'll put that aside. Yes. AUDIENCE: A lot of how they're sending bits or receiving bits. What we've found with the service I work on is that if you call something caching rather than keeping a copy, then the content owners are much more willing to allow you to have a copy on the client side. Therefore, community networks that merely cache the video in ways that the [INAUDIBLE]. BOB FRANKSTON: Yes. Our language, our legal system. Well, our legal system is still in the 13th century. Instead of science, which tests ideas, our legal system proves things by example. Not that I have anything against lawyers, but I have something against lawyers. OK. MALE SPEAKER: Thank you very much.