Comment Dear Santa. For Windows-10-end-of-support-day in October, please may we have a dead simple bulletproof all-free OS that gets old PCs online without a Google account, and does nothing else?
There are a lot of desktop Linux distros out there. It’s the number one thing non-Linux users complain about: too much choice. One of them only comes bundled with hardware, and yet, it has more users than all the others. It’s ChromeOS, and it could be facing the chop soon. How come there is no all-Free Software tool that even tries to do what ChromeOS does without needing an account with The Borg?
Even if it may soon be replaced by a desktop flavor of Android, ChromeOS does just one thing and it does it well enough that vendors sold billions of the things last year.
The idea of ChromeOS is simple: it’s just enough Linux to get you online. It turns a PC into something akin to a tablet, with a full-screen icon-based app launcher. The desktop is very simple and vaguely Windows-like: there’s a taskbar at the bottom, a file manager, drivers enough common hardware that most things just work out of the box, including a bunch of common GPUs, networking including Wi-Fi. In terms of apps, there’s a built-in Google Drive client, and of course the Chrome web browser.
And that is about it. Anything else that you want to do, you do in a browser window. Productivity apps? Use Google Apps. Messaging or video calling? Log in to your chat system of choice in a browser window. The file manager can show images, but not much more. Your bookmarks, passwords, and what few settings there are are stored in your Google account.
All the components are there, including a potential revenue model. Strip out absolutely all the complexity that can possibly be removed, and leave something which can run on any old PC from the last 15 years and gets the user online – and nothing else. How hard can it be?
There are no optional extra native apps, and no way to add any. If you have branded ChromeBook hardware, there’s the Google Play store and it can run some Android apps in a built-in compatibility environment. Around the Irish Sea division of Vulture Towers, we mostly just use ChromeOS Flex on whatever spare laptop is lying around, which works very nicely indeed. There’s no app store, but you can open a Debian VM and install Debian apps in there. We installed VLC so that it could play movies or audio files, and it automatically integrated with the desktop – suddenly, clicking on a movie in the file manager just worked.
This is the sort of simplicity that we feel a desktop Linux for the masses should aspire to. There are no local apps and no questions: no choice of desktop, no dual boot, no fancy add-on cross-distro package managers like Flatpak or Snap. The desktop is just Windows-like enough to be instantly familiar, unlike GNOME; on the flipside, it’s not cluttered with a hundred options to tweak, plus two Help/About menu entries, three text editors, menu bars in some apps and hamburger menus in others, as in KDE Plasma.
Where Google’s team put innovative effort into ChromeOS was in making it robust enough to be sold to the masses in the hundreds of millions of units, with no tech support. It’s immutable, with image-based updates. It has two root partitions, one of which updates the other, so there’s always a known good one to fall back to if an update should fail.
This is a more fault-tolerant design than SUSE’s MicroOS-based systems, which use the rather fragile Btrfs. It’s also much simpler than the Fedora Atomic immutable systems, including offshoots such as Universal Blue, which use the Git-like — for which, read “fearsomely complex” — OSTree. For added entertainment, Fedora also defaults to Btrfs, with compression enabled. If you don’t believe us about the problems of damaged Btrfs volumes, refer to the Btrfs documentation. We recommend taking the orange-highlighted Warning section very seriously indeed.
It seems to us like there’s nothing to rival this in the non-Google world. There have been ready-rolled web-kiosk OSes such as Webconverger. It was open source, based on Debian, with updated over Git. It was end-of-lifed a couple of years ago.
Like Chrome itself, the upstream project behind ChromeOS, ChromiumOS, is open source. There are a couple of projects forked from the basis of ChromiumOS, such as FydeOS, which adds a second authentication system so it can be used behind the Great Firewall of China. We plan to return to FydeOS and take a deeper look soon. NayuOS is comparable: it’s broadly ChromiumOS, with some added tools for developers. It’s mainly intended for use in guest-login mode, but it can also talk to Nexedi’s SlapOS back-end servers.
If Mountain View does decide to stop developing ChromeOS, expect both of these to stop being maintained very soon afterwards.
At the UI and functionality level, there is nothing in ChromeOS that would be difficult to replicate for any distro vendor. All the bits are there: an embarrassment of desktops, networked authentication via LDAP or something, network file sharing over Seafile or the like, umpteen email servers and webmail systems, such as OwnCloud and NextCloud.
There are even all-in-one servers such as Zentyal, which we looked at in 2010. These could serve as the back-end and offer a revenue stream: paid subscriptions for extra storage, for instance – or selling private servers for customers to run their own fleets of clients.
Nothing in this is really difficult. All the pieces are there. For our money, Btrfs is just too fragile for remote unsupported client devices, but if it’s locked read-only in normal use that might make it resilient enough. If for some reason updates and rollback must be performed using a single boot volume, we’d rather see it use OpenZFS, but redundant partitions and fail-over, as used in Valve’s SteamOS, seems a simpler answer.
The goal is banishing any and all questions of what to download, what desktop, what apps, packaging formats, software stores, user accounts, backing up, all of it: run the browser locally, sync the data – and only the data – to the cloud.
And all the components are there, including a potential revenue model. Strip out absolutely all the complexity that can possibly be removed, and leave something which can run on any old PC from the last 15 years and gets the user online – and nothing else. How hard can it be?
The ideal time would have been long enough before the End of 10 to work the bugs out beforehand, but it’s still not too late. Distro vendors can still offer their rich local clients for those geeky enough to want them. This is alongside existing offerings, not instead of them.
Simplicity is key. Long before there were such things as personal computers, Antoine de Saint-Exupéry said Perfection is achieved, not when there is nothing more to add, but when there is nothing left to take away. But there’s an even pithier version than that. Mies van der Rohe put it best: Less is more. ®