On 1 May 2025, Redis announced that Redis 8 would be available under the GNU Affero General Public License v3 (AGPLv3) — an Open Source Initiative-approved license — alongside its existing source-available options. A little over a year after Redis abandoned its permissive open-source license in favor of source-available terms, the project returned to genuinely open-source licensing. It is one of the more striking reversals in recent open-source history, and it carries lessons that extend well beyond Redis itself.

To understand why the reversal matters, it helps to recall how the original change played out.

How We Got Here

In March 2024, Redis moved away from the permissive BSD license it had used for years, adopting a dual source-available model under the Redis Source Available License and the Server Side Public License. The stated motivation was commercial: the company wanted to prevent large cloud providers from offering managed Redis services that competed with its own commercial offerings without contributing back. Source-available licenses permit viewing and modifying the code but restrict the commercial uses that the company wanted to capture.

The community response was swift, and it followed what had by then become a recognizable pattern. Within days, a fork emerged. Backed by the Linux Foundation and supported by several major cloud providers and contributors, the fork — Valkey — preserved the permissive BSD license and continued development from the last open-source Redis release. We traced that sequence and its relationship to earlier license-change forks in our coverage of the Redis license change and the rise of Valkey.

That coverage placed the Redis episode in a lineage: a company relicenses a widely used open-source project for commercial protection, the community forks under neutral foundation governance to preserve the open version, and the market then chooses between the two. The pattern had been rehearsed with other infrastructure projects, and Valkey appeared to be following the script.

The Reversal

What makes the May 2025 announcement notable is that the script did not run to its expected conclusion. Rather than the source-available original and the open fork settling into a long coexistence, Redis reversed course and added an OSI-approved open-source license to its own offering. The company’s own framing acknowledged the community reaction to the 2024 change as a factor.

The AGPLv3 is a deliberate choice. It is a strong copyleft license: it not only requires that modifications be shared, but extends that requirement to software offered over a network. An organization that runs a modified Redis as a network service must make its modifications available. This is precisely the provision designed to address the original commercial concern — it ensures that a cloud provider offering modified Redis as a service cannot keep its changes proprietary — while still qualifying as genuine open source under the Open Source Initiative’s definition. In effect, Redis found a way to pursue its commercial protection goal within an open-source license rather than by abandoning open source altogether.

What It Means for the Forks

The obvious question is what the reversal means for Valkey, which exists precisely because Redis stopped being open source. Several observations follow.

First, the fork achieved its purpose. Whether or not one views the reversal as a direct consequence of Valkey’s existence, the credible open alternative changed the strategic calculus. A relicensing that produces a thriving, foundation-backed fork carries a cost the original project may not have fully priced in: a permanent competitor with the moral and legal high ground of openness, momentum among contributors, and the backing of the cloud providers whose adoption decisions move markets. The reversal is, at minimum, consistent with the fork having mattered.

Second, the reversal does not dissolve the fork. Valkey now has its own contributors, its own roadmap, its own users, and the backing of a neutral foundation. Projects with that kind of institutional footing do not simply fold when the original changes its mind. The two projects are likely to coexist, with the AGPL license of Redis 8 and the BSD license of Valkey representing genuinely different value propositions — AGPL’s network copyleft is more permissive for some uses and more restrictive for others than the unrestricted BSD terms Valkey carries. Organizations will choose based on their own licensing posture.

Third, the episode refines the fork playbook rather than ending it. The lesson for both companies and communities is that a credible fork is a powerful corrective. For a company contemplating relicensing, the Redis experience is a cautionary one: the commercial benefit may be undercut by the loss of community goodwill and the rise of a competitor that did not exist before. For a community facing a relicensing, the lesson is that the fork is worth doing — not only because it preserves the open version, but because its mere existence can shift the original’s behavior.

The Broader Pattern

Stepping back, the Redis story is one instance of a recurring tension in open-source infrastructure: the conflict between the commercial interests of a company that develops a widely used project and the expectations of a community that has come to depend on it being open. Several infrastructure projects have moved through some version of this conflict, and the outcomes have varied — some relicensings have stuck, with the company retaining control and the fork remaining marginal; others have produced forks that displaced or rivaled the original.

The variable that seems to matter most is governance. Projects developed and owned by a single company are exposed to that company’s commercial pressures; the license can change because one entity controls it. Projects developed under neutral foundation governance — where no single commercial actor can unilaterally relicense — offer a structural guarantee against this kind of reversal. This is the deeper reason the fork-under-foundation pattern recurs: the foundation is not merely a host but a commitment device, a way of placing the project beyond the reach of any one company’s future commercial needs.

This is a theme with long roots in distributed-systems and open-infrastructure history. The most durable shared infrastructure — the protocols and platforms that organizations are willing to build their operations on — tends to be the infrastructure that no single vendor can capture. Neutral governance is what makes that durability credible, and the Redis episode is a vivid illustration of why communities increasingly insist on it before committing to a project.

Operator Takeaways

For teams that depend on Redis or are evaluating in-memory data stores in 2025:

  • The reversal is good news for openness, but read the license carefully. AGPLv3 is genuine open source, but its network copyleft provision has real implications for organizations that modify the software and offer it as a service. For most users who run an unmodified data store, it changes little; for those who modify and redistribute, the obligations differ meaningfully from both the old BSD terms and the source-available licenses.
  • Valkey remains a credible option. The fork is not going away, and its unrestricted BSD license may suit organizations that prefer permissive terms. The choice between Redis 8 and Valkey is now a real one, made on technical and licensing merits rather than forced by a relicensing.
  • Governance is a procurement criterion. The episode reinforces a practical lesson for infrastructure selection: a project’s governance structure is a legitimate factor in whether to build on it. Foundation-governed projects carry a lower risk of disruptive relicensing than single-company-controlled ones.

Further Reading

Back to Blog