Words to Avoid: OOS

(see GNU’s Words to Avoid) The distinction between “open source” and “free software” has become more than an academic debate, it’s a battleground for software freedom itself. While these terms are often used interchangeably, their conceptual differences have created a dangerous loophole that corporations exploit to appear freedom-friendly while maintaining restrictive control over their products.

The Wikipedia entry for “Open-source software movement” explicitly warns: “Not to be confused with Free software movement.” This distinction matters profoundly. Software that provides source availability, while certainly superior to closed-source alternatives, does not automatically guarantee user freedom. The “open source” label has become a marketing tool that obscures the more fundamental question: does this software respect user autonomy and control?

Telegram exemplifies this deceptive practice. The company markets itself as “open-source software,” yet only portions of its client code are available, and even that availability is questionable. Compare this to Signal, which provides complete software freedom: users can examine, modify, and distribute the entire codebase. Telegram’s selective transparency creates an illusion of openness while maintaining control over critical components.

Meta’s approach to AI represents perhaps the most sophisticated example of this deception. The company releases what it calls “open source” Large Language Models (LLMs), providing training code to create an appearance of transparency. However, this selective disclosure conceals the most crucial elements; no training datasets, no raw data access, no processing scripts, no data samples, no reproducibility mechanisms. This strategic omission allows Meta to claim “open source” credentials while maintaining proprietary control over the foundation of their AI systems. The community’s response celebrating Meta as a “leading AI company” for these limited releases demonstrates how effectively this strategy manipulates public perception.

Several major database companies have exploited “open source” terminology while fundamentally restricting user freedom, MongoDB pioneered this approach by switching from the genuinely open AGPL license to their proprietary Server Side Public License (SSPL). MongoDB submitted the license to the Open Source Initiative (OSI) for approval in 2018, but the OSI rejected it precisely because it fails to meet open source standards. Elasticsearch made a similar move in 2021, moving the Open Source portions of Elasticsearch and Kibana source code to non-OSI approved software licenses, SSPL and Elastic License v2. These companies continue marketing themselves as “open source” while using licenses that explicitly violate open source principles.

Companies like GitLab and Nextcloud employ “open core” models where basic functionality remains open source, but essential enterprise features require proprietary licenses. While technically more honest than completely closed alternatives, this approach fragments software users can examine and modify basic functions but remain dependent on proprietary components for full functionality.

Corporations selectively release components that appear generous while withholding elements that would enable true user freedom.

Only by maintaining these distinctions can we prevent the complete erosion of software freedom under the guise of corporate-friendly “openness.”