Government open source needs an engine, not pushers

Something is a bit off between government and open source land. Every conversation about open source in government is a conversation about policy and funding. Both are hygiene. Neither is the engine. Everyone is busy pushing the car, and nobody asks how it is supposed to keep running once they stop.

This week I spoke with a policy officer who walked me through how the latest open source initiative is being kicked off. A funding programme of a few million, a coalition of parties around the table, a plan that names the early users, and the familiar policy cycle of funding, monitoring, recalibration. It could just as easily have been about Common Ground, or about any of the ERP refresh projects from the past decade. The story sounds the same every time. Entirely correct. Entirely necessary. That's how you get the car out of the ditch.

But as he was talking I sat there thinking: everyone at this table knows how to push. Nobody is asking who's behind the wheel once we're back on tarmac. You don't win races by pushing. You don't even drive by pushing. Sooner or later you're spent, and if there's no engine running by then, the car just rolls back into the ditch.

Government is a starter, not an ecosystem

Let's be honest. Government has no innovation capacity. That may sound harsh. It's observation. Almost every piece of government-led open source from the past decade is either an implementation of something that was conceived elsewhere (Collabora, Nextcloud, OpenProject, openDesk building blocks), or hyper-specific to one use case (vacation rental rules, Signalen, the electronic voting machines). Inside those constraints the work is fine. That's exactly where government's competence sits: setting the brief, taking it into operations, evolving it where needed. What it can't do, and what it shouldn't be doing, is experimenting on risk as the commissioning party for its own citizens.

For open source that means: government is a great starter and a hopelessly bad ecosystem in its own right. Starting an engine is not the same as keeping it running. A funding programme is a crank handle. After that, an engine has to run somewhere else.

What happens when the engine never starts

What happens when government never hands the engine over to someone else, I've seen more often than I'd like. One kind of customer with one kind of requirement set is a monoculture. And monocultures die of one disease: one austerity round, one coalition agreement, one state secretary who says "let's just go with Microsoft", one department that sighs and rips it out. Every feature ossifies into something that only works in a government context. The bill for maintenance, security, and ongoing development is paid 100% by the taxpayer for something that hangs 100% off a funding programme. Developers and vendors stay away, because there's no bread to be made, only contracting until the budget runs out. And when someone eventually says "we should have done open source after all", everyone points at the government that should have carried it.

Need examples? The Dutch government ICT graveyard is well-stocked. KEI, the digitalisation of Dutch justice, burned €200 million in six years before the plug was pulled. oBRP, the online citizens register, went through €100 million before the same happened. The Tax Authority's ETPM hit €203 million. Those are the headline numbers; estimates put total Dutch government ICT waste at €1 to €5 billion a year, year in, year out. The car park behind the policy department is mostly wrecks at this point. Open source projects do not get a special pass to avoid it; the discharge problem is exactly the same.

The question is how to keep the next one off that lot. And it is a sharper question than "fund it longer", because if you crank for too long without picking up the direction, somebody else picks it up for you. One day you look up and a commercial outfit has rolled in with a tow truck and made off with the wheel. You traded your dependence on Microsoft for dependence on whichever vendor happened to be parked nearby.

And if you ask: but isn't that exactly what sovereignty is? No. If government is the only one who runs it, builds it, and maintains it, you've just swapped your dependence on Microsoft for dependence on government. That's nationalisation with a nicer word.

Sovereignty without a market underneath isn't sovereignty. It's a hostage situation with extra steps.

Why neither model transfers to the Netherlands

"Yeah but," says my policy officer, "look at La Suite, look at openDesk. It's working over there, isn't it?" Correct. La Suite numérique runs on roughly 600,000 Tchap agents and over 500,000 daily users across Visio and Docs. By every measure that is a major sovereignty win for France, and it works because DINUM is one central button in a centrally governed country. Germany with openDesk is taking a similar path: solid adoption inside government (Robert Koch Institute, the International Criminal Court), and a partner programme for the private sector kicking off in 2026. Both are real successes in their own context. There is nothing wrong with either model.

The catch for the Netherlands is that neither model transfers. We have 342 municipalities, twelve provinces, twenty ministries, hundreds of independent agencies. We are never getting one central button, so the French procurement trick does not work here. And we do not have a single federal IT body the way Germany does, so the openDesk pattern of one government rollout dragging the partners along has no obvious Dutch equivalent either. For us the market has to carry it, or no one will.

The engine is already running. Just not in The Hague.

The good news: that market is already running, and not only here. Nextcloud ships the upstream out of Germany. T-Systems and IONOS are lined up as openDesk delivery partners. The big French and EU integrators are positioning around La Suite. The commercial ecosystem has decided that sovereign open source is real product, and they all want to be in the engine room.

The same pattern is alive in the Netherlands. KPN, together with Nextcloud, has put a sovereign Nextcloud-based offering on the market as a commercial product, and SURF rolled out the same stack to 100,000 users in higher education and research in 2025. One layer deeper into the education system, SIVON, the Dutch secondary-education IT cooperative, just started an OSPO pilot on the same Nextcloud stack. Stacked up, most of Dutch education is moving onto a shared open source platform. Students from secondary school through university will spend their school years getting used to Collabora and EU-built office tools instead of Microsoft. Quietly, that is one of the biggest open source wins in the Netherlands of the past decade.

Centric sits on the same line, with partners like Lucidon and Apura. These are the parties building the engine. They are putting capital in to ship open source to small business, to vital infrastructure, and to all those care institutions and municipalities that will eventually have to leave Microsoft. Their math says it can be done, and their math counts heavier than a funding programme, because they have to make a living from it.

This is the conversation Dutch open source policy is not having. The white papers and coalition agreements stop at policy and funding. Adoption is what matters once those run out, and almost nobody is talking about that one.

It gets worse than just not having the conversation. Government is actively blocking adoption outside its own ecosystem. The rules on developer.overheid.nl and code.overheid.nl define ownership strictly as government-owned. A non-government party cannot host a repository there, let alone share ownership of one. That single rule kicks the most-adopted Dutch open source projects straight out of the car park. Valtimo and the FrankFramework, two of the most-used open source platforms in Dutch public-sector IT, are co-owned by parties outside government (insurers, mostly). So they are ineligible for the very forge that government open source policy claims to encourage. We are handing out parking violations to the cars that actually run.

What government has to do is make sure its work can land there. Not crank the engine once and then start over. Keep going until someone is behind the wheel and the pushers can go home.

Step by step

Policy and funding are hygiene. They get the car out of the ditch. What is needed after that is adoption, and specifically adoption outside government. Inside government there is always another funding round to keep a project breathing. Outside government there is not. The engine has to run on its own there, and that is the only honest test an open source project ever gets.

An open source solution that lives only inside government won't get a second life. The same goes for the next ERP refresh. And the same for every other funding programme from the past decade we now find ourselves wondering whatever happened to.

KPN, Centric, and the ring around them are ready. SURF showed with 100,000 users that it works. The engine is already running, just not where the government has it standing on a crank.

The alternative is the next Little Engine That Could: well-funded and well-meant, charming on the policy slides, never quite making it over the hill.

Step aside. Unleash the engine and let the race begin.