Embedded Linux development: the old way

Delivering software for embedded Linux devices has traditionally been problematic.

The proliferation of a format per distribution in standard Linux packaging, whether a Slackware package, a DEB or an RPM, means developers have to learn them all to reach a wider audience. Discovering new software for embedded devices often means hunting down PPAs, grabbing packages via command-line tools like rpm, apt and pktools, or building from source instructions. Furthermore, handcrafted packages often require complex code to manage installation and updates, are incompatible with each other, have unmet dependencies or may write all over the filesystem.

To solve the challenge of installed libraries having dependencies on specific software package versions, users often obtain newer or older versions of the dependent packages. This may break compatibility further, pushing the problem to another set of packages.

An equally painful story holds for updates. Because those packages are not transactional, software update strategies are mostly a manual, non-atomic package-based operation.

Whereas an apt-get update works fine for servers within a secure environment, intermittent power and network outages for devices in the field mean an update will likely be interrupted. Delivering reliable software updates to low-powered, inaccessible, and often remotely administered embedded devices requires a well-thought-out solution. With traditional mechanisms, faulty updates can cause IoT devices to become unstable, and fixing them requires costly manual intervention through on-site engineer visits or device recalls.