Far down the vault a man was screaming. His fists were tightly clenched and he was screaming out imprecations against the humming computer. There was a hopeless rage in his eyes -- rage and bitter, savage defiance. Even as he screamed he began to slouch forward, with the terrible weariness of a man trapped beyond all hope of rescue.
The Internet of Things! Existential menace or meaningless buzzword? Automating away drudgery or just eroding privacy? Cornucopia or Panopticon?
The Internet of Shit
But why such skepticism in a world of rampant technophilia? Why are we not, in fact, welcoming our robot overlords?
Very Valid Criticisms
There are a lot of criticisms out there, but for me the top ones are:
- If you're not the customer, you're the product. The backend services have to get paid for somehow. and if you're not paying directly, you're paying by accepting advertising and/or by having your privacy sold.
- Your hi-tech device may suddenly turn into a container of hummus if the service provider no longer feels like supporting it.
- Uselsss if the Internet isn't available. Even the most reliable networks sometimes go down. Consumer routers fail.
- Crypto support on IoT devices is often very weak. Partly because of lack of entropy, lack of CPU resources, etc. It's tricky to use a protocol like SSL on a tiny CPU.
- It's hard enough to get people to change their smoke alarm batteries or update Internet Explorer. No-one ever is going to reflash their thermostat.
- The software for things like lightbulbs is often awful, and open to all kinds of exploitation. This shouldn't be too surprising since even the manufacturers of electronic locks have trouble getting this right.
Not Shit Things
Turning the criticisms around, what requirements do we get:
- Devices need to be retargetable. That Nest Thermostat is really just a rotary encoder and a display, so why shouldn't it do other things, or more things, if you no longer want it to be just a thermostat. Why should all your lightbulbs be locked into a single vendor?
- Communications need to be local. Sure, it's nifty that you can switch your lights on from Low Earth Orbit, but it is ridiculous to bounce every light-switch-flip off some Cloud somewhere.
- The network needs to be protected. Mostly from the Internet, but also from the local network.
Any candidate solution has to address these requirements.
I'd like to do some more work on defining a simple candidate protocol which would meet these requirements and hopefully get it published as an RFC. In the meantime, I'll be talking about The Internet of Toys at Buzzconf Nights and continuing to develop these ideas.
- I'll be talking about this stuff at PyCon AU's inaugural Internet of Things mini-conf which should be fun! More news as it comes to light.