In the 90s, devices would be connected to the internet through a gateway and that was called the internet of things. It’s actually still like that, except that the UX has improved a little and moved to mobile – which makes the IoT concept more sensible. There’s also a cloud that is actually relevant. Also the device is no longer connected via serial, but via ZigBee, Bluetooth, …. The reason for IoT has also changed: it’s no longer a way of controlling things, but rather a way of gathering data.
Real IoT problems:
- Security issues
- Current compute power may be enough for the gateway, but the sensor has some way to go. But there are no big changes to be expected anymore.
IoT has much lower-power hardware than what Linux is used to (enterprise, smartphone, embedded).
Does it make sense to make Linux small, since the smallest DRAM you can buy is 32MB? It still makes sense for embedded RAM, but that is still not large enough to run Linux anyway. And the embedded RAM is not going to grow that much anymore. Still, there’s a chicken and egg problem: if Linux is not going to run on it, there’s not much push for making bigger RAM, but vice verse if the chips anyway won’t work then there’s no push for making Linux smaller.
Security problems haven’t been exposed sufficiently for people to make it a real priority. Also, in embedded systems, the features needed to make security real are often missing. You also need to be able to do secure updates over the air – possibly with live updates because downtime is expensive. It will be impossible to keep an entire IoT system secure, so in addition you have to be resilient to security failures in part of the system and be able to regain control. Here there is platform work that could be an enabler. Containers were also quoted as a way to add security, but personally I fail to see how that helps.