Fireside chat with Wienke Giezeman, The Things Industries and Ray Ozzie, Blues

I'm Ray Ozzie. I've been in the software business for quite a while. I was introduced to IoT and the nascent Things Network at a time when there was a tragedy happening in Japan. I had left Microsoft, where I was Chief Software Architect, in January of 2011, and in March there was a tsunami that took out the nuclear plants of Fukushima. A number of us technologists did what we always do and tried to say, how can we help? A couple of weeks after the meltdown, a group of us converged in Tokyo and decided that the best thing to do was to use our skills to measure radiation on the ground because the government and the power company had their hands full relocating people but weren't giving people any data to ascertain whether their children were safe. So we created a project around citizen science and open data, using Arduino, Adafruit, and SparkFun modules to put together a network of mobile and fixed radiation measurement devices. The mobile ones worked extremely well, people put on Tyvek suits and drove around measuring radiation. We also got radiation detectors near the plant and needed a way to backhaul, but there were no active cell towers, so we constructed a mini LoRaWAN network out of spare parts. From that experience, I learned two things. First, it was eye-opening in terms of the impact mankind can have on the environment, regardless of whether you're pro-nuke or anti-nuke. There were no baseline radiation measurements, and it gave me a greater awareness of how IoT technology can give us a viewpoint of the state of our environment, because giving people awareness puts us in a much better position than treating all resources as infinite. Second, as a technologist, it showed me how challenged people who want to build rapid solutions are, and how things we think are easy—like connecting to a LoRaWAN or cellular network—are actually very difficult, with hurdles like power management, security, and dealing with AT commands and state tables. So I created Blues Wireless, early on focused on cellular, to help customers make it much easier to build cellular-enabled products. But it pertains to every aspect of wireless and building devices.

Well, I've been fortunate to have been in the industry a while and seen different transitions, from client-server computing to web computing and so on. Every time there's a major technology transition, right off the bat, technologically astute people build little platforms for their own thing. When the web first started, you saw dozens of startups building mail servers, web servers, and little programming platforms around them. It becomes very chaotic, but through natural selection, it reduces, and the community of experts tries to make things simpler for customers to reduce choice. People finally decide, let's not have so many ways, let's just have two or three. That's how the transition happens to a growth phase. The web wouldn't have grown up had it stayed fragmented in terms of technologies. The IoT industry we're all part of has taken much longer than anyone expected. People made growth projections early on because the demand is clearly there, but I put the blame on us—we have not done the hard work of picking the winners yet. It's still a very fragmented ecosystem. We get excited about new radio access technologies, new edge inference technologies, but things are not going to become simpler for customers until we collectively decide to put all our weight behind one RTOS or a tool chain that everyone can rally around. For example, Microsoft Visual Studio Code—you can like it or hate it, but the community of developers has decided it's going to win, so everyone has rallied around it. Even if someone has an alternative, it's not helping customers to create more choice. So let's rally around that. There are many layers of the stack in building hardware where we could help customers by just rallying around certain things.

There has been a viewpoint where people use the word 'convergence,' specifically pertaining to wireless technologies. I'm not big on convergence because it makes it sound like there's one point everything comes to, when in fact there's a reason LoRaWAN exists side by side with cellular, and Bluetooth is not going away. It's important for us to find some harmony between them so we decide on common semantics or APIs to transport things over these. Let's agree on the higher layers of how applications deal with them and how they can live side by side. I may disagree with some people at this conference, but I view LoRa as more of a LAN than a WAN because I believe the cellular network is really the only WAN we have at this moment. Perhaps satellite will be another viable WAN, but unless something has the ubiquity of cellular, it's hard to call it a WAN. LoRa has amazing potential even from a LAN perspective because of its unique penetration characteristics and long distance in agriculture. We can help customers by not saying all technologies are good for everything. If you've got a cornfield, it's probably not good to put a cellular module on every stalk of corn. There are sweet spots for certain technologies, but they can be in harmony.

Yeah, we just have to—it's not going to be overnight. But I would encourage people, and I'm mostly repeating myself, instead of becoming enamored with the next thing down the line, the more we can get together in common interest groups like this conference and decide what we're going to believe in. How many of you remember LAMP as a paradigm that characterized the stack? Or when HashiCorp came out with Terraform, suddenly DevOps people had a common viewpoint: I need something to describe a model of my infrastructure, I need a logging package. They taxonomized it like that. It's only going to be our industry that does that. We can keep becoming enamored with things like putting WASM containers on IoT devices, but that really does not help customers at this moment in time. I personally believe the biggest challenge and opportunity for us is to reduce the time from concept to value. That's only going to happen if this group says, 'No, you need FreeRTOS plus this kind of microcontroller plus this edge inference thing, and this is how you get it to the cloud, and this is the right security you should use.' Forget about all the other choices—they all have merits—but here's a toolkit of limited size, and we're all talking the same way and integrating the same way, like with webhooks.

It's really interesting you say that. From a technology perspective, I'm very excited about what's going on in the NTN, the satellite space. I think there's untapped potential there. People really haven't conceptualized that there's another potential wide area network. The rates for sending messages are going to drop, and I think that's a big opportunity. People familiar with it have stayed away because of the rate structure, just like they stayed away from cellular due to rate structure and other impediments. So that's another big opportunity. I'd love for us to work together to develop—maybe like the 12-Factor App methodology—some practices and layers we all agree on. I'd love to work with anyone who wants to. Do we need to have a traditional kind of IoT architecture book? Everyone knows there's no one type of IoT device, but at least for the layers of complexity, if we could say this part is application-specific but the rest snaps into the normal way of doing it, that would be great. I'd like to see us make progress there. If today it takes three years to go from prototype or concept to deployment, I'd like it to be six months.

I just want to make sure people understand: in consolidation and simplification, not everyone wins. There will be things that go by the wayside. But the opportunity for all of us is growth. We're not going to have the hockey stick curve everyone's talked about until this industry decides we're just going to pick some winners. I look forward to that a lot.

Thank you, thank you.