flatsiedatsie
commented
3 years ago Wouldn't a full desktop OS create a much bigger attack surface? And as you said, wouldn't having two images create management overhead?
In my experience installing a minimal window manager and privacy-hardened Chromium browser on the headless image worked very well.
https://www.candlesmarthome.com/adding-a-touch-screen-to-the-candle-controller
I created an install script that sets all this up. It can be found here: www.candlesmarthome.com/tools/kiosk.txt
It has cool features, like showing an image / looping logo video during boot.
It did have a downside: what I created did not handle multiple browser tabs gracefully. There was no UI for switching tabs, and this was by design. I tried to just not have the gateway every open other tabs in order to make kiosk-users not have to window-manage. For example, I used browser flags to forbid opening external links. But this still allowed users to open a new tab for the internal logs page, which opens in a new tab.
Perhaps a pragmatic solution would be an addon that manages this window manager and browser installation. But that would require some serious sudo-ing during that installation process :-)
benfrancis
dizzwold
In the last WebThings monthly public meeting multiple people (myself included) expressed an interest in adding touch screen controller software to WebThings. There have also been many requests to add built-in support for voice input via microphones directly connected to the gateway (e.g. using the ReSpeaker hat on Raspberry Pi).
I have created a new controller repository (which replaces the things-controller repository) intended for work on a controller application which provides a web runtime to render the gateway's web interface on a screen and accept user input via touch and voice. This application is intended to be used either on a combined gateway/controller device, or on standalone controller devices which connect to a central gateway.
The default Raspberry Pi image of WebThings Gateway (built on Raspberry Pi OS Lite), is designed for headless (i.e. no display, touch panel, microphone or speaker) smart home hubs. Rather than try to augment the existing OS image into a distribution for smart speakers and smart displays (adding lots of dependencies that headless hubs don't need), my suggestion is to create a separate distribution specifically for those devices (e.g. based on the full desktop version of Raspberry Pi OS). It's quite common for Linux distributions to have separate desktop and headless versions (e.g. Raspbian & Raspbian Lite, Fedora Workstation & Fedora Server, Ubuntu Desktop & Ubuntu Server etc.) so this makes perfect sense to me.
The obvious challenge we have is who would maintain this distribution, as we're already lacking the resources needed to continue to maintain the existing cut down distribution for Raspberry Pi. Something which could make this easier is to pick a containerised OS like Ubuntu Core or Balena OS where another organisation maintains the core/host OS (via automatic updates) and we can focus on our maintaining WebThings Gateway & WebThings Controller applications inside their own container(s). We already have a docker image and (WIP) snap package of the gateway application.