Born classified

A paper from Space Week Nordeste 2025 came across my desk this week. Three Brazilian researchers asked whether open source is a viable model for the space sector, looking at CubeSats, at UPSat, at SatNOGS. It was nice to see the work we did at Libre Space Foundation show up in a policy discussion on the other side of the Atlantic.

Their conclusion, though, is the one I keep hearing everywhere. Open source is great for education and for the “less sensitive” layers of a mission: telemetry, simulation, payload control. The critical parts, security and navigation, should stay proprietary or under restricted access. They call it a hybrid model, and present it as the balanced, grown-up answer.

I don’t think it is. I think it’s a habit. Inertia, really. Space has been closed by default for most of its history, and that default has less to do with engineering than with where the industry came from. It’s time we noticed, and let it go.

Born classified

The first satellite went up on an R-7, the Soviet Union’s first intercontinental ballistic missile (UMD Aerospace). Most of the American launchers of that era were converted military missiles too: Redstone, Thor, Atlas, Titan (Britannica, US Space Force history office). For decades the main customers for space hardware were defence ministries and the agencies that grew up next to them. Reconnaissance satellites were secret: CORONA, America’s first spy satellite programme, flew from 1960 to 1972 and was only declassified in 1995 (NRO). Launchers were missiles with a different nose cone, and the whole field was shaped by people whose job was to make sure the other side learned nothing.

That culture made sense for what it was. But it didn’t stay where it belonged. Civil and commercial space grew up inside the same institutions, with the same suppliers, the same contracting habits and the same instinct: if it flies, it’s sensitive, and if it’s sensitive, nobody outside gets to see it.

The clearest example of what that instinct costs is the American satellite industry. In 1999, after a scandal over US companies sharing technical information with China while investigating failed launches, Congress reclassified commercial communication satellites as weapons. They moved from the Commerce Department’s lighter export rules to ITAR, the US regulations on arms exports (Freethink, Orbit Codex). A TV broadcast satellite was now regulated like a machine gun.

Before the change, US manufacturers held on average 83% of the commercial satellite manufacturing market. Within a few years it was down to about 50% (The Space Review). Not all of that was ITAR; the market also had too many manufacturers chasing too few contracts. But European companies saw the opening and started designing satellites with no US components at all, so customers could skip the licensing delays. France launched the first “ITAR-free” satellite in 2005 (Satellite Today). Most commercial satellites only came off the munitions list in 2014, fifteen years later (US Office of Space Commerce).

The point isn’t about the US. It’s that closing things off, in the name of protecting an industry, shrank that industry. The ones who gained were the ones who were easier to work with.

Europe has its own version, with better manners. ESA has an open source policy, and on paper it even defines open source the way the Open Source Initiative does. Then it carves out a kind of “open source” whose licensed territory ends at the ESA member states, and requires committee approval before ESA-owned software goes anywhere else (ESA Open Source Policy). ESA’s standard contracts only allow open licensing as an option, so closed stays the default. And a licence that stops at a border isn’t open source by any definition the rest of the world uses. It’s the same old reflex, dressed up as openness.

We’ve seen this before

If you worked in software in the late 1990s, you remember the arguments. Linux was a hobby. Serious companies ran serious workloads on licensed operating systems, with a vendor to call when something broke. Open source was fine for students and for the edges, not for anything that mattered. In private, Microsoft knew better: its internal memos, leaked in 1998 as the “Halloween documents”, called open source a direct threat to its server business (Halloween Documents). In public, Steve Ballmer was still calling Linux “a cancer” in 2001 (The Register).

Sound familiar? It’s the hybrid argument, almost word for word, twenty-five years earlier.

Today Microsoft says that more than 60% of customer cores on Azure run Linux (Phoronix). The company that fought Linux hardest now runs a cloud where most of the paying workload is Linux. Nobody converted Microsoft with an ethics lecture. Customers wanted it, it was cheaper to run, and the ecosystem around it was bigger than anything one vendor could build.

The numbers on the wider economy point the same way. A 2024 Harvard Business School working paper estimated what it would cost if open source disappeared and every company had to rebuild the open source it uses: about $8.8 trillion. Firms would be spending 3.5 times more on software than they do now (Hoffmann, Nagle & Zhou). A 2021 study for the European Commission put open source’s contribution to EU GDP at €65–95 billion a year, from roughly €1 billion invested in 2018. It also estimated that 10% more contributions would add 0.4–0.6% to GDP and more than 600 new tech start-ups (European Commission). You can argue about the methodology of either study, and people have. You can’t really argue about the order of magnitude.

Open source won in operating systems, the web, the cloud, databases, machine learning frameworks. It didn’t win because it was nicer. It won because it was the better business model for the shared layers: the ones everybody needs and nobody makes money by owning.

It’s already happening in space, quietly

Space has started down the same road. It just doesn’t talk about it much.

When NASA’s Ingenuity helicopter flew on Mars in 2021, its navigation computer ran Linux and its flight software was built on F Prime, a framework JPL had released as open source. JPL’s engineers said it was the first time the lab had open-sourced flight software, and that they were flying an open source operating system, an open source flight framework and parts you could buy off the shelf (NASA/JPL, GitHub). The first powered flight on another planet ran on code anyone can download.

Orekit is a less famous story and a better business lesson. CS (now CS Group), a French company, started building a flight dynamics library in-house in 2002. In 2008 they released it under the Apache licence. By traditional logic, they gave away the crown jewels. Today Airbus Defence and Space uses it in its flight dynamics software, Thales Alenia Space uses it for a new geostationary platform, and CNES chose it in 2011 as the basis of its next generation of flight dynamics systems (Orekit). CS Group still sells products built on top of it. They didn’t lose their business by opening it. They became the centre of it.

And then there’s our own corner. On 18 May 2017, UPSat was deployed from the ISS: a 2U CubeSat built by the University of Patras and Libre Space Foundation, with every subsystem designed in the open, hardware and software. Shortly after deployment, its telemetry was being picked up by SatNOGS stations around the world (LSF). SatNOGS is a ground station network built and run mostly by volunteers, on open hardware and software, and more than four thousand stations have been registered on it over the years (SatNOGS Network). No proprietary ground segment could have been rolled out that widely, that cheaply, by that many different people.

Not only a matter of principle

At LSF we started from principles. We believe space is a commons, that the technology to reach it should belong to everyone, and that knowledge paid for by the public should go back to the public. I still believe all of that.

But I’ve noticed that when I make that case in a room full of industry people, eyes glaze over. Ethics sounds like a cost. So let me make the other case, the one I’d make to a CFO, or to a ministry deciding where to put its space budget.

Open lowers the entry fee. A new player in space today, whether a small company, a university, a non-profit or a country starting its first programme, has to pay for the same things everyone before them paid for. Flight software. A ground segment. A comms stack. Mission control tools. Most of that is not where anyone’s value lies. It’s plumbing. When the plumbing is closed, every newcomer pays for it again, either by building it or by licensing it from someone who did. When it’s open, they start from where the field already is. The Brazilian paper makes this exact point about licensing costs in countries with tight budgets, and then stops one step short of the conclusion.

Open is how you disrupt incumbents. What protects a traditional space company from newcomers is accumulated, closed heritage: years of flown designs nobody else can see. Open components take that advantage away, because the same building blocks become available to everyone. That’s exactly why it’s disruptive, and exactly why incumbents won’t lead it. It happened to the proprietary Unix vendors. It can happen to proprietary platforms and software stacks in space. If you’re a new entrant, open is not a sacrifice. It’s your best weapon.

Open is more sustainable. Missions outlive vendors. A satellite designed today may be operated for ten or fifteen years, and the company that wrote its ground software may not be around for all of them. With open code, maintenance can be shared, forked, picked up by someone else. More importantly, every closed mission throws its lessons away. The fixes, the test cases, the things learned the hard way stay inside one company and die with the programme. In an open stack, heritage compounds. Each mission that flies an open component makes it more trustworthy for the next one.

And there’s still value to be generated. Orekit shows you can open a core asset and still sell products, integration and expertise around it. The real value in this industry is generated in services, data, operations and the payloads that actually make a mission different. Nobody’s business case is “we wrote our own telemetry parser”.

“But what about security?”

This is the objection the hybrid model rests on, so it deserves a real answer.

Cryptographers settled this in the 1880s, with what’s known as Kerckhoffs’s principle (Kerckhoffs, 1883). A system should stay secure even if everything about it except the key is public. Security that depends on nobody seeing the design isn’t security. It’s a delay.

Space already has the counterexample. On 24 February 2022, the morning Russia invaded Ukraine, an attack on Viasat’s KA-SAT network knocked tens of thousands of satellite broadband modems offline across Europe. The attackers used the network’s own management system to push malware that wiped the modems’ memory (SentinelLabs, BleepingComputer). As collateral damage, Enercon lost remote monitoring and control of 5,800 wind turbines in central Europe (Euronews/Reuters). That was a closed, proprietary system run by a serious company. Being closed didn’t protect it. It just meant the people who might have spotted the weaknesses weren’t looking.

The things that actually make space systems secure all work at least as well on open code: threat modelling, verification and validation, signed updates, keeping a full inventory of every component in your software, penetration testing. Many of them work better, because more people can check.

I’m not saying a military reconnaissance payload should be on GitLab. Defence will keep its secrets, and that’s fine. My argument is about the default for everyone else: the civil, scientific and commercial missions that picked up defence’s habits without having defence’s reasons.

Casting off the inertia

Inertia doesn’t go away by itself. Somebody has to push. Here’s where I think the push should come from.

  1. Public money, public code. Software and hardware designs paid for by space agencies should be open by default, under standard OSI-approved licences, available worldwide. Not limited to a territory, not waiting on a committee.
  2. Procurement that rewards reuse. Tenders should give credit for building on open components and contributing improvements back, instead of quietly favouring whoever owns the most closed heritage.
  3. Count open heritage as heritage. When an open component has flown, that flight history should count for the next mission that uses it, whoever integrates it.
  4. Companies: open your plumbing. Keep what really sets you apart. Open the rest, and become the place everybody else builds on. It worked for CS Group. It’s also how we’ve worked at LSF from the start: so many CubeSat teams use SatNOGS precisely because anyone can build on it.
  5. Keep export controls narrow. Control what is actually sensitive, and stop treating ordinary satellite parts as munitions. We’ve seen what the alternative costs.
  6. Newcomers: build on the commons, and give back. Every fix that goes upstream lowers the entry fee for the next team.

None of this is radical. It’s what the software world did, slowly and painfully, over twenty years. Space can do it faster, because we already know how the story ends.

Closing

Space is about to have more players than it has ever had. Small companies, universities, non-profits, countries launching their first satellite. The question is whether each of them has to pay the same entry fee separately, rebuilding the same plumbing in private, or whether they can start from a shared foundation and spend their money on what’s actually new.

The first option is how space has always worked. The second is how every other technology field that scaled ended up working. I know which one I’d bet on, and not only because I think it’s right. It’s cheaper, it’s faster, and it makes the whole field bigger.

Space was born classified. It doesn’t have to stay that way.

Sources