WOSS
The Apache Software Foundation: Open Source's Other Root
Aman Mundra · 2026-07-15 · 5 min read
Contents
TL;DR - The Apache Software Foundation (ASF) is one of the oldest and largest homes in open source: a US 501(c)(3) public charity, founded in 1999, hosting 300+ top-level projects under the permissive Apache License 2.0. It is not part of the Linux Foundation - it is a separate governance root, run on a fiercely community-first model known as "the Apache Way." And it governs the data-infrastructure layer - Kafka, Spark, Airflow - that most production AI quietly depends on.
When people map open source, they tend to draw one tree with the Linux Foundation at the root. That map is incomplete. There is a second root, older than most of the projects hanging off the first one, and a great deal of the software you rely on grows from it: the Apache Software Foundation.
If you have ever streamed events, orchestrated a data pipeline, or run distributed compute, you have almost certainly used something the ASF governs.
What is the Apache Software Foundation?
Founded in 1999 to steward the Apache HTTP Server, the ASF is a nonprofit that has grown into a home for more than 300 top-level projects spanning web servers, databases, big-data engines, streaming platforms, and orchestration tools. Everything it hosts ships under the Apache License 2.0 - a permissive license that lets companies build commercial products on top without the copyleft obligations of the GPL, which is a large part of why the enterprise world adopted Apache projects so readily.
The ASF is a genuine peer of the Linux Foundation, not a member of it. The two are the primary neutral roots of the open-source world, and they are structured differently in one important way:
- The Linux Foundation hosts umbrella foundations - CNCF, the PyTorch Foundation, OpenSSF - which in turn host projects.
- The ASF hosts projects directly. There is no intermediate layer. Kafka, Spark, and Airflow each sit as their own top-level project under one flat, consistent governance model.
The Apache Way: community over code
The ASF's real product is not any single piece of software - it is a governance philosophy that has kept hundreds of projects healthy for two decades. It is called the Apache Way, and its core tenets are worth knowing because they explain why Apache projects rarely die and rarely get captured by a single vendor:
- Meritocracy. Every project is run by a Project Management Committee (PMC) of committers who earned their seats through sustained contribution, not through their employer's cheque. Influence is earned in the open.
- Community over code. A healthy, diverse contributor community is treated as more important than any individual feature. A project with brilliant code but one contributor is considered at risk; a project with good code and a broad community is considered strong.
- Consensus and openness. Decisions happen on public mailing lists, in the open, by consensus. "If it didn't happen on the list, it didn't happen" is a genuine cultural rule.
- Vendor neutrality. No company owns an Apache project. This is what lets fierce competitors - who would never trust a rival to control shared infrastructure - collaborate on the same codebase.
That neutrality is the whole point. When AWS, Microsoft, and a hundred startups all depend on Kafka, none of them wants a competitor to own it. The Apache Way is the structure that makes shared ownership durable.
The data-infrastructure trio that AI runs on
The ASF's most strategically important projects, at least from where we sit building AI and data platforms, form a single chain. A production AI pipeline reads almost verbatim as a sequence of Apache projects:
streaming (Kafka) → orchestration (Airflow) → compute (Spark)
- Apache Kafka - the distributed event-streaming platform that ingests real-time data at the edge of most modern data systems.
- Apache Airflow - the workflow orchestrator that sequences ingestion, processing, and ML-pipeline steps as code.
- Apache Spark - the distributed engine that does the heavy batch compute and feature engineering behind model training.
This trio is the data-infrastructure backbone underneath much of cloud-native AI: the events, pipelines, and feature data that model training and inference depend on. It is not a competitor to the Kubernetes-and-agents world of the CNCF - it is the layer that feeds it.
There is a striking pattern in how all three evolved in their latest major releases, and it says something about where mature distributed systems are heading: each one shed an external dependency to become simpler and more scalable. Kafka 4.0 removed ZooKeeper entirely in favour of its own internal metadata layer. Spark 4.0 decoupled clients from the cluster via Spark Connect. Airflow 3 decoupled task workers from the metadata database via a new execution API. The same instinct - remove the thing bolted on the side, make the core self-contained - runs through the whole ASF data stack right now.
Why it matters
For anyone building on modern data infrastructure, the ASF is not optional knowledge - it is the foundation you are already standing on. Understanding its governance model tells you why these projects are safe to build a business on: they are community-owned, permissively licensed, and structurally resistant to any one vendor pulling the rug. That durability is exactly what you want under the data layer of a system meant to run for years.
