Arvados Licensing FAQ

This FAQ is designed to answer questions about licensing for the Arvados open source project.

Software licenses

The Arvados API server and the core Arvados system services, Keep and Crunch, are distributed under the AGPLv3 license. Workbench, a browser-based visual tool that calls the Arvados APIs, is distributed under the AGPLv3 license. The Software Development Kits, the code samples in the Arvados documentation, and the tools for writing Crunch scripts, are distributed under the Apache 2 license. For more details on the licensing of each file in the codebase, see the COPYING file.

The Arvados documentation is licensed under the Creative Commons Attribution-Share Alike 3.0 United States license.

Why were these licenses chosen?

Arvados has several design goals. The OSS licenses were chosen to specifically support three of them:

Open Standards: With Arvados, we hope to create an industry standard around which data scientists can coalesce and organize. We believe that this is only possible when developing such a standard in the open and while listening to the needs and wants of the industry. The AGPLv3 license for the core stack prevents an industry stakeholder from attempting to move the stack in favor of a particular vendor.

Federation: Federation is the ability to author a pipeline in one Arvados deployment and run it in any other Arvados deployment, whether that deployment is on an on-premise cluster or on a public cloud provider. To do so requires a uniform set of APIs and environments, and fragmentation of the Arvados core code base increases the risk that an analysis will break for no reason other than that the two installations came from a fork in the code base. Licensing the core stack under the AGPLv3 drastically limits the opportunity for fragmentation of the code base, preserving federation capabilities for everyone who deploys Arvados.

Flexible Authoring: Arvados was always intended to be used by authors of both open source and proprietary analyses, pipelines, tools, and applications. Therefore, the SDKs and code samples are distributed under the Apache 2 license, which allow authors to develop open source or proprietary solutions on top of Arvados.

What are the implications of these licenses for my own code?

According to the AGPLv3 license, all contributions to and modifications of the core Arvados code base that are distributed, or made available for use over a network, must themselves be available to recipients under the same license. This ensures that all those who deploy Arvados are entitled to the same improvements in underlying infrastructure, and that the core of Arvados does not fragment.

According to the Apache 2 license, derivative works are not required to be redistributed under any license. Therefore, developers can choose to build proprietary pipelines and client software for Arvados, and those works are protected under international copyright law.

MongoDB’s licensing also follows this paradigm, and it is summarized in their blog on the same topic.

Updated by Ward Vandewege almost 6 years ago · 6 revisions