Choosing an open source license is one of the most important decisions you can make for a new project. Most developers default to popular permissive licenses like MIT, BSD or Apache. These are great choices for non-commercial projects, as they reduce user resistance toward project adoption.
When applied to a commercial open source project, permissive-by-default licensing schemes may impact your ability to capture value for your creations. To successfully commercialize, you will have to outcompete others building on top of your project in addition to your own free offering. Many successful OSS companies like Redis, Elastic and MongoDB have confronted this challenge by adopting more creative and/or restrictive licensing schemes.
For entry level projects, it is commonly feared that choosing more restrictive (i.e. copyleft) licenses, like GPL will deter potential users from adopting your software. This is a completely valid concern – organizations like Google, for instance, will ban use of AGPL3-licensed software entirely. If no one uses your software, then you've failed before you've started.
Although widespread adoption is undoubtedly the top goal of any new open source company, it's not really clear what your license should optimize for. If the goal is to maximize value, should your license prefer low adoption resistance or high value capture?
Turning to the Data
To answer this question, let's turn to OSS.cash, an index of $100M+ Revenue commercial OSS companies. As of Nov. 2018, this sheet tracks roughly ~$140B of commercial OSS outcomes.
First, we can do a brief breakdown of license choices against % makeup of valuations:
While most commercial OSS companies are permissively licensed, GPL and copyleft licenses seem to occupy the majority in value. This isn't surprising. Intuitively, restrictive licenses should allow you to capture more direct value from your creations.
If we consider a breakdown of average valuation by license type, the prominence of permissive licenses seem to diminish considerably:
In comparison, copyleft-licensed software shows a significantly higher average valuation compared to permissively-licensed software. In fact, even the AGPL, a significantly less popular license, averages a similar valuation to the permissive Apache 2, and even eclipses BSD-licensed companies.
Finally, let's take a look at business models by license:
This graph is really interesting – first, it advises which license choices are best depending on business models. However, what's more interesting is that the license choices can be used as a signal on how directly the company can commercialize its OSS offering.
For example, if you can monetize by selling support and a light enterprise offering (open core) on top of your OSS project, then you are very directly commercializing what you have. Alternatively, if you have to funnel money into building, selling, maintaining and supporting a separate proprietary SaaS offering, your commercialization becomes more indirect. Naturally, a direct model relies on restrictive licensing as a way to control monetization ability as there is no proprietary counterpart to sell separately.
Should we use copyleft-by-default?
It's still far too early to begin recommending copyleft licenses by default for every new project, or even for existing business models. This analysis has its limits:
- Our data is very limited, open source outcomes are still very new -- oss.cash only tracks ~38 companies
- The GPL was a lot more popular among older, more mature companies which have had more time to create value (pre-2007: 40% copyleft, 2007+: 25% copyleft)
- Outlier exits among mature companies (i.e. Red Hat) skew this data to make strong copyleft licenses seem more dominant. In reality, it seems like Weak Copyleft licenses (MPL, LGPL, CPAL) strike the best balance.
Despite these limits, I think it's safe to say that copyleft licenses are commercially viable. While we can't claim restrictive licensing will make your business succeed, it has helped plenty of companies break $100M in revenue and account for the largest outcomes in open source.
- Commercial copyleft OSS is very viable – despite more companies being permissively licensed, the majority of OSS outcomes come from copyleft-licensed software
- Copyleft licenses are more dominant among business models that rely on direct monetization of the core project (i.e. open core / support = direct, SaaS = indirect)
- License choice can impact, but won't determine the viability of your project. Focus on creating great software, and the business model / licensing scheme will follow.