Dual licensing is the oldest working answer to the question the Collaborative Source Initiative was asking: how do you keep the code open and still charge for it? The trick is to publish the same codebase under two instruments at once. To the community, the software is free under a strong copyleft license, usually the GPL. To a company that wants to embed the code in a proprietary product, the same software is sold under a conventional commercial license that lifts the share-alike duty.

One codebase, two doors. The model worked well enough to build companies, and its history explains a great deal of how the licensing shelf is organized.

How the two doors work

The mechanism depends entirely on ownership. Only the holder of all the copyright can offer the same code under different terms, because the holder is not bound by the license it grants to others. A company that writes every line itself, or that collects signed contributor agreements, can release version 1.0 under the GPL for the world and simultaneously sell version 1.0 under a proprietary license to a customer who cannot accept copyleft.

MySQL made this the textbook case in the 2000s. The database was GPL for everyone, which spread it through the open source world, and commercial licenses were sold to manufacturers and vendors who needed to ship it inside closed products. Qt, the application framework, ran a similar arrangement for years. The beauty of the mechanism is that the two doors share one codebase, so every improvement paid for by the commercial side lands in the commons, and every fix contributed under the GPL improves the product being sold.

What the GPL side is really for

The copyleft side of a dual license is not charity. It is the distribution engine. Releasing under the GPL costs the company nothing it was not already giving away, spreads the software through every distribution and package manager, recruits users and contributors, and crucially makes the code unusable to a competitor who wants to close it. The free edition builds the market; the paid edition harvests the customers for whom copyleft is a deal-breaker.

This is why dual licensing almost always pairs with a strong copyleft text. A permissive free edition would let competitors take the code commercial without paying, and the second door would have nothing to sell. The copyleft is not the ideology of the model; it is the moat that makes the paid door worth walking through.

The contributor agreement underneath

The model has a hidden precondition: the company must own or control every contribution. That is why dual-licensed projects historically asked contributors to sign a contributor license agreement or copyright assignment, granting the company the right to relicense the work commercially. The clause is reasonable on its face and it has always been contested, because it means the community contributes under open terms while the company alone can sell exceptions to those terms.

Where the model shows its seams

Dual licensing carries a structural tension. The community edition must be good enough to spread, but the commercial edition must be better enough to pay for, whether through features, support, or freedom from copyleft. Hold back too much and the community calls it crippleware. Give away too much and nobody buys the second license. Companies walk that line differently, and the market judges each choice.

Who used it and who moved on

The model also produced one of the licensing shelf's odder artifacts: the contributor who writes a patch for the GPL edition knowing the company may sell it under the commercial one. The community accepted that arrangement when the value flowed both ways, and rejected it when the flow became one-directional, which is why the dual license's golden age is remembered as a bargain rather than a trick.

Beyond MySQL and Qt, the model passed through Sleepycat, Berkeley DB, and a long list of infrastructure projects. Its great era was the 2000s, when distribution meant shipping software. The cloud weakened it: a service provider can run GPL code without distributing it, so the copyleft door no longer forces anyone to buy the commercial one. That erosion is part of why the industry moved toward source-available and delayed-release texts.

The honest reading

Dual licensing is the cleanest version of the contribute-or-pay idea: the payment is not for using the code but for exemption from sharing back. Filed beside the Jahia license, it shows the same instinct expressed with better mechanics. The binder keeps it as the model that proved a company could charge for openness without closing the code, even if its golden age is behind it.

The dual license is also the model that taught the community to ask who owns the code. Before it, a contributor's signature was a formality; after it, the signature was the mechanism that made the second door possible, and the question of who holds the copyright became part of reading any project at all.

What the text grants

The same code under copyleft for the commons and under a paid license for those who cannot accept the share-alike duty.

What a reader should check

Who owns the contributions, since dual licensing only works where a single holder can relicense the whole.

The limit of this reading

The model thins once software is delivered as a service, where the copyleft trigger rarely fires.