Let’s look at other software on the desktop, to understand why there isn’t (as a broad, popular platform) Linux on the desktop, then how there could be.
Over on De Programmatica Ipsum I discussed the difference between the platform business model, and the technology platform. In the platform model, the business acts as a matchmaking agent, connecting customers to vendors. An agricultural market is a platform, where stud farmers can meet dairy farmers to sell cattle, for example.
Meanwhile, when a technology platform is created by a business, it enables a two-sided business model. The (technology) platform vendor sells their APIs to developers as a way of making applications. They sell their technology to consumers with the fringe benefit that these third-party applications are available to augment the experience. The part of the business that is truly a platform model is the App Store, but those came late as an effort to capture a share of the (existing) developer-consumer sales revenue, and don’t really make the vendors platform businesses.
In fact, I’m going to drop the word platform now, as it has these two different meanings. I’ll say “store” or “App Store” when I’m talking about a platform business in software, and “stack” or “software stack” when I’m talking about a platform technology model.
Stack vendors have previously been very protective of their stack, trying to fend off alternative technologies that allow consumers to take their business elsewhere. Microsoft famously “poisoned” Java, an early and capable cross-platform application API, by bundling their own runtime that made Java applications deliberately run poorly. Apple famously added a clause to their store rules that forbade any applications made using off-stack technology.
Both of these situations are now in the past: Microsoft have even embraced some cross-platform technology options, making heavy use of Electron in their own applications and even integrating the Chromium rendering engine into their own browser to increase compatibility with cross-platform technology and reduce the cost of supporting those websites and applications made with Javascript. Apple have abandoned that “only” clause in their rules, replacing it with a collection of “but also” rules: yes you can make your applications out of whatever you want, but they have to support sign-in and payment mechanisms unique to their stack. So a cross-stack app is de jure better integrated in Apple’s sandpit.
These actions show us how these stack vendors expect people to switch stacks: they find a compelling application, they use it, they discover that this application works better or is better integrated on another stack, and so they change to it. If you’re worried about that, then you block those applications so that your customers can’t discover them. If you’re not worried about that, then you allow the technologies, and rely on the fact that applications are commodities and nobody is going to find a “killer app” that makes them switch.
Allowing third-party software on your own platform (cross-stack or otherwise) comes with a risk, that people are only buying your technology as an incidental choice to run something else, and that if it disappears from your stack, those customers might go away to somewhere that it is available. Microsoft have pulled that threat out of their briefcase before, settling a legal suit with Apple after suggesting that they would remove Word and Excel from the Mac stack.
That model of switching explains why companies that are otherwise competitors seem willing to support one another by releasing their own applications on each others’ stacks. When Apple and Microsoft are in competition, we’ve already seen that Microsoft’s applications give them leverage over Apple: they also allow Apple customers to be fringe players in the Microsoft sandpit, which may lead them to switch (for example when they see how much easier it is for their Windows-using colleagues to use all of the Microsoft collaboration tools their employers use). But Apple’s applications also give them leverage over Microsoft: the famed “halo effect” of Mac sales being driven by the iPod fits this model: you buy an iPod because it’s cool, and you use iTunes for Windows. Then you see how much better iTunes for Mac works, and your next computer is a Mac. The application is a gateway to the stack.
What has all of this got to do with desktop Linux? Absolutely nothing, and that’s my point. There’s never been a “halo effect” for the Free Software world because there’s never been a nucleus around which that halo can form. The bazaar model does a lot to ensure that. Let’s take a specific example: for many people, Thunderbird is the best email client you can possibly get. It also exists on multiple stacks, so it has the potential to be a “gateway” to desktop Linux.
But it won’t be. The particular bazaar hawkers working on Thunderbird don’t have any particular commitment to the rest of the desktop Linux stack: they’re not necessarily against it, but they’re not necessarily for it either. If there’s an opportunity to make Thunderbird better on Windows, anybody can contribute to exploit that opportunity. At best, Thunderbird on desktop Linux will be as good as Thunderbird anywhere else. Similarly, the people in the Nautilus file manager area of the bazaar have no particular commitment to tighter integration with Thunderbird, because their users might be using GNUMail or Evolution.
At one extreme, the licences of software in the bazaar dissuade switching, too. Let’s say that CUPS, the common UNIX printing subsystem, is the best way to do printing on any platform. Does that mean that, say, Mac users with paper-centric workflows or lifestyles will be motivated to switch to desktop Linux, to get access to CUPS? No, it means Apple will take advantage of the CUPS licence to integrate it into their stack, giving them access to the technology.
The only thing the three big stack vendors seem to agree on when it comes to free software licensing is that the GPL version 3 family of licences is incompatible with their risk appetites, particularly their weaponised patent portfolios. So that points to a way to avoid the second of these problems blocking a desktop Linux “halo effect”. Were there a GPL3 killer app, the stack vendors probably wouldn’t pick it up and integrate it. Of course, with no software patent protection, they’d be able to reimplement it without problem.
But even with that dissuasion, we still find that the app likely wouldn’t be a better experience on a desktop Linux stack than on Mac, or on Windows. There would be no halo, and there would be no switchers. Well, not no switchers, but probably no more switchers.
Am I minimising the efforts of consistency and integration made by the big free software desktop projects, KDE and GNOME? I don’t think so. I’ve used both over the years, and I’ve used other desktop environments for UNIX-like systems (please may we all remember CDE so that we never repeat it). They are good, they are tightly integrated, and thanks to the collaboration on specifications in the Free Desktop Project they’re also largely interoperable. What they aren’t is breakout. Where Thunderbird is a nucleus without a halo, Evolution is a halo without a nucleus: it works well with the other GNOME tools, but it isn’t a lighthouse attracting users from, say, Windows, to ditch the rest of their stack for GNOME on Linux.
Desktop Linux is a really good desktop stack. So is, say, the Mac. You could get on well with either, but unless you’ve got a particular interest in free software, or a particular frustration with Apple, there’s no reason to switch. Many people do not have that interest or frustration.
