Fork in the road: Is open source relicensing outrage driving developers to new tools?

Fork in the road: Is open source relicensing outrage driving developers to new tools?

Updated September 9:

In a fun bit of serendipity, on August 29, 2024, just days before this newsletter was set to publish, Elastic announced that it had would return to its open source roots by “adding AGPL as another license option next to ELv2 and SSPL in the coming weeks.” Founder and CEO Shay Bannon wrote that the original relicensing decision had succeeded in “removing market confusion” by forcing AWS and others to build their own distinct fork rather than repackaging Elastic-branded code.

Did community pressure cause Elastic to reverse course? According to Shay, the company “had always hoped that enough time would pass that we could feel safe to get back to being an Open Source project - and it finally has.” Others suggest that the slow but steady growth of Opensearch, clearly shown in the pipeline data we shared, was what pushed Elastic toward re-embracing the open source ethos (albeit via the less permissive AGPL).?

No matter the causes, this decision by Elastic could mark a turning point in the open source relicensing debate. As companies experiment with more restrictive licenses, Elastic's adoption of the AGPL – if it succeeds – may swing the pendulum back toward more open models. Only time will tell.


The digital world is built on open source software. According to a 2024 report by cybersecurity firm Synopsis, 96% of all codebases feature at least some open source code. Researchers estimate the current economic value of open source software usage at nearly $9 trillion.

For open source advocates, these numbers are more evidence that open source has won out against proprietary, closed source software distribution models.?

But for all its successes, there is a fundamental tension at the heart of open source, pitting the community ideals of open access and collaboration against the economic pressure for commercial viability and profitability.

The latest example of this conflict is the uproar that followed high-profile relicensing decisions by major open-source software vendors. In a bid to secure more sustainable financial models, open source companies including Elastic, Terraform, Redis, MongoDB, Confluent, Sentry, and others have shifted to "source available" licenses, giving users access to the source code but restricting how it can be used commercially.

These decisions drew widespread criticism from open source proponents, who used terms like “disingenuous,” “rug pull,” and “bait and switch” to describe them. In some cases, contributors even launched competing forks of the original projects that maintain the original license terms.

But is the uproar all bark and no bite? Has it actually translated into large-scale migrations away from the relicensed tools?

In this issue, we’re looking at two years of CircleCI pipeline data to discover the impact of open source relicensing on developer tool usage. Has there been a mass migration away from source-available projects to their more authentically open source forks? Is the community appeal of open source enough to overcome the inertia of entrenched tool choices?

Let’s dive into the data to find out.

TL;DR

  • Open source companies use relicensing to protect themselves from competition, drawing ire from open source advocates.?
  • Contributors to relicensed projects often respond by creating open source versions, or forks, of the original project.
  • Despite negative press around relicensing, there is very little evidence of a large-scale migration to open source forks.?

What is open source?

Open source software is software with source code that anyone can inspect, modify, and enhance. It is developed in a collaborative, public manner and distributed with a license that allows anyone to view, use, modify, and share the code.

Open source software offers significant benefits for both developers and businesses. Developers gain access to a wide range of projects and the chance to improve their skills through collaboration with a global community. For businesses, open source can reduce software costs, accelerate time to market, and increase trust among stakeholders.

We’ll dive more into the history and philosophy of open source below, but first let’s define a few key terms to help set the stage.

Key terms in open source:?

  • Copyleft: A software licensing term that requires all modified and extended versions of a free program to be free as well.
  • Fork: When developers take a copy of source code from one software package and start independent development on it, creating a distinct and separate piece of software.
  • Free software: Software that grants users four essential freedoms: to run, study, change, and distribute it and its modified versions.?
  • Open source software (OSS): Software that is made available with a license that allows anyone to view, modify, and distribute the source code, typically adhering to the four essential freedoms of free software.
  • Permissive license: A type of open source license that allows software to be freely used, modified, and redistributed, including for commercial purposes, with no copyleft requirements or other restrictions.
  • Proprietary software: Software that is owned by an individual or a company, with usage, modification, and redistribution restricted by the owner.
  • Relicensing: Changing the license under which software is distributed. This can be controversial, especially when moving from more permissive to more restrictive licensing.
  • Restrictive license: A type of license that imposes strict conditions on how software can be used, modified, and distributed, often used to protect proprietary interests and control over software.
  • Source available license: Software whose source code is made available for viewing, and sometimes for modification and redistribution, but does not necessarily grant all the freedoms associated with open source or free software.
  • Source code: The human-readable version of a program, written in a programming language, which can be compiled into a binary executable.

The relicensing controversy: How did we get here?

Open source software originated in the context of collaborative, community-driven efforts where source code was freely available for anyone to use, modify, and distribute. User modifications could be submitted back to the original developers, creating a collaborative cycle of improvement and innovation that would become a hallmark of open source development. Key projects like the GNU/Linux operating system and the Apache web server helped to cement the reputation of open source as a powerful model for software development in the 1990s.?

As the software industry evolved, so did the economics of open source projects. In the 2010s, the original ethos of open collaboration began to clash with commercial interests, particularly as major corporations integrated open source projects into commercially driven products and services.?

Thus began what Bryan Cantrill called open source software’s “midlife crisis”: companies built on open source tools found it increasingly difficult to sustain themselves financially as their products were commoditized by larger entities that could offer the same tools with more robust infrastructure and support systems.?

To protect themselves from these challenges, many companies adopted new licensing strategies. “The open source model has to evolve,” declared HashiCorp CEO Dave McJannet, “given the incentives that are now in the market.” Most often, evolving meant moving away from permissive open source licenses towards more restrictive "source-available" models that allow access to the source code but impose restrictions on commercial use.

Relicensing typically involves moving from open source to source available license types

This shift has sparked considerable controversy. Proponents argue that relicensing is necessary to protect open source projects from being exploited by service providers that contribute little back to the community. Some even suggest that source-available licenses retain all the important collaborative benefits of open source while allowing creators to safeguard their interests and ensure the project's sustainability.

Critics, however, argue that source-available licenses undermine the foundational principles of open source by restricting how software can be used and shared. “It’s deception, plain and simple,” wrote the directors of the Open Source Initiative, “to claim that the software has all the benefits and promises of open source when it does not.”

Notable relicensing decisions

MariaDB founder Monty Widenius first explored the idea of a source-available model in a 2013 academic paper, arguing that a business source license had the potential to “facilitate the generation of income without alienating the open source community.”

MariaDB adopted the BUSL for its MaxScale 2.0 database proxy release in 2016, providing a blueprint that resonated across the industry. Numerous companies followed suit, adopting source-available licenses as a strategy to sustain their business models while maintaining a commitment to the open source community.?

Redis

Introduced in 2009, Redis is an in-memory data structure store that can be used as a database, cache, and message broker.

Initially licensed under the permissive BSD license, Redis quickly became a core offering from many large cloud providers, including AWS Elasticache, Azure Cache, and GCP Memorystore. In 2018, Redis Labs applied the Commons Clause to parts of its software. After a strong negative reaction, Redis retired the Commons Clause in favor of a new Redis Source Available License (RSAL) in 2019. In March 2024, Redis expanded its relicensing to the core product, applying a dual-licensing approach using the new RSAL and the SSPL.??

Days after Redis adopted the licensing change, the Linux Foundation announced a fork of Redis 7.2.4, the last BSD-licensed Redis release. The new project, Valkey, quickly gained the support of major providers including AWS, Google, and Oracle.?

MongoDB

Launched in 2009 as an open-source tool, MongoDB is a document-oriented database designed for ease of development and scalability with a flexible document schema.

Originally released under the AGPL, MongoDB was widely used and offered as a service by a number of large cloud providers. In 2018, Mongo announced its adoption of the SSPL, arguing that many vendors were “test[ing] the boundaries of the AGPL license,” and that “the community needs a new license that…makes explicit the conditions for providing the software as a service.”?

Opposition to the SSPL was widespread and vocal. However, MongoDB's initial use of the AGPL—a license already more restrictive than most open-source licenses—meant there was no significant movement to create direct open source forks. Instead, some service providers opted to build similar offerings with compatibility layers, such as AWS's DocumentDB, which replicates much of MongoDB's functionality but does not use its code. Others, like Red Hat, opted to drop MongoDB outright.?

Elasticsearch

Elasticsearch is a powerful search and analytics engine known for its speed and scalability. It is part of the Elastic stack (formerly ELK stack), which also includes Logstash and Kibana.

In 2021, Elastic NV changed the license from Apache 2.0 to a dual Elastic License and SSPL, limiting how cloud providers can offer Elasticsearch services. “Our license change is aimed at preventing companies from taking our Elasticsearch and Kibana products and providing them directly as a service without collaborating with us,” wrote founder and CTO Shay Bannon.

In response, AWS announced that it would launch OpenSearch, a fork of Elasticsearch 7.10, under the Apache 2.0 license, with support from other companies including Red Hat, SAP, Capital One, and Logz.io.

Terraform + Vault

HashiCorp’s Terraform is an infrastructure as code tool that automates the management of both cloud and on-premises resources, allowing users to deploy and manage infrastructure through code rather than manual processes. Its secrets management tool, Vault, complements Terraform by securely storing and tightly controlling access to tokens, passwords, certificates, API keys, and other secrets.?

In 2023, HashiCorp announced that it would shift from the Mozilla Public License to the BUSL first introduced by MariaDB in 2016. According to the terms of the license, vendors who built products on top of HashiCorp code would “no longer be able to incorporate future releases, bug fixes, or security patches contributed to our products.”

In response, prominent members of the Terraform community formed the OpenTF (later renamed OpenTofu) project and published a manifesto pledging to maintain a fork of Terraform under the guidance of the Linux Foundation. OpenTofu became generally available in January 2024. OpenBao, a fork of Vault, was announced in December 2023, but has yet to reach general availability.?

The impact of relicensing on open source fork adoption: A look at the data

It’s clear that the decision to relicense formerly open source projects can provoke a strong reaction from developers. But it remains to be seen if this enthusiasm can translate to broader community adoption.?

To gauge the impact of licensing changes, we analyzed tool usage data across thousands of organizations and millions of CI/CD pipelines run on CircleCI from January 2023 to June 2024, comparing changes in the number of organizations using relicensed tools against their open source competitors, as well as the number of pipelines run for each.?

Let’s start by looking at the most established open source fork in our dataset, OpenSearch. By some measures, OpenSearch is significantly outperforming its source-available rival, Elasticsearch:

  • Since January 2023 (the time window of our query), the number of organizations using OpenSearch has grown by 132%, as compared to 25% growth among Elasticsearch users.
  • During the same period, the number of CI/CD pipeline runs that include OpenSearch have grown by 183%, as compared to a 40% increase in pipelines for Elasticsearch.?

OpenSearch has essentially doubled its market share, growing from 11% of Elasticsearch’s user base in January 2023 to 20% in June 2024.?

Yet despite the momentum behind OpenSearch, Elasticsearch maintains a large lead in both raw adoption and raw usage numbers.?

OpenSearch has been growing at a faster rate, but remains a fraction of Elasticsearch usage

These findings mirror the results of a DB-Engines comparison, which shows OpenSearch growing faster in popularity but still lagging significantly behind Elasticsearch overall:

While OpenSearch has achieved significant, even surprising levels of success, it still has a long way to go before it overtakes its more established predecessor. And that is a story we see playing out among the newer open source forks: while enthusiasm is high, adoption has been slow.

Take Valkey for example. As RedMonk’s Stephen O’Grady points out, the Linux Foundation-backed Redis fork has seen significantly more development activity in recent months than Redis. But real-world usage of the project, at least in the first few months since its April 2024 GA release, has not materialized.

Similarly, the Terraform fork OpenTofu was received with much fanfare, yet the number of organizations utilizing OpenTofu in their pipelines has remained in the single digits since it went GA in January 2024. Terraform usage has stayed essentially flat, at approximately 2,000 organizations, over the same period.?

This data supports the conclusion in OpenLogic’s 2024 State of Open Source Report that “the license change does not seem to have negatively impacted Terraform usage.”?

From this analysis, we can see that most open source forks face an uphill battle against established source-available projects. Despite considerable noise and numerous threats, there is little evidence of an immediate, large-scale backlash against relicensed tools. For now, the bet made by open source companies that relicensing would deter service providers while retaining the majority of other users appears to have paid off.?

Conclusion

Open source is essential to the modern software ecosystem. From mobile phones to internet infrastructure to enterprise software, open source tools are the foundation for most technologies we use today.?

Source XKCD

Support for open source development is driven not just by its practical benefits but also by a philosophical commitment to accessibility and community development. Yet commercial open source stands at a crossroads. As companies struggle to balance the community ethos with the need for financial sustainability, the number of organizations opting for source-available licenses will inevitably grow.

The real question then becomes whether this trend will have any impact on the tool choices for everyday developers. Matt Asay, tech commentator and current VP of developer relations at MongoDB, suggests the answer is no: “[open source proponents] continue to fret about licensing when developers mostly care about use, just as they always have…’Open source’ is a secondary concern to ‘It helps me be more productive, faster.’”?

Our findings support Asay’s conclusion. At least for now. But with major tech companies and open source foundations throwing more and more resources behind open source projects, it may be only a matter of time before forks like Valkey, OpenTofu, and OpenSearch reach critical mass and break through into mainstream adoption.?

Further reading:

What is open source?

Build open source projects with CircleCI

Optimizing open source project builds on CircleCI

要查看或添加评论,请登录