Published on May 12, 2025

Recently, I had some discussions on whether or not open source matters and I felt compelled to articulate some of my thoughts in writing.

Would we be better off if all research in the USA was kept private? Would Nvidia be worth more? Would our lives improve? Would we have more or fewer Nobel prizes? Does China have its own private alternative to Apache? Does it have a proprietary version of TCP? What about AI models - do they rely on open data, and are the models themselves open? To me, the answers to these questions are somewhat obvious.

Open innovation represents both a fundamental requirement and a complex challenge in our tech ecosystem. Most scientific research institutions and private corporations rely heavily on open source - from operating systems, programming languages, and communication protocols to applications themselves.

Open systems generate superior innovation outcomes, yet, funding remains problematic.

Why I Think This Way - My Foundations

Open source has been idealized during my education, for good reason. During my bachelor’s in communication networks at the University of Brasilia (2008-2013), all my school coursework was built on open source foundations. The ability to examine implementation details allows for a comprehensive understanding that proprietary systems could not provide. The Internet’s architecture itself was built on open standards.

In parallel, while pursuing CCNA certification, I observed that vendor-specific educational material invariably included promotional elements and technology biases.

The OpenFlow Promise

When I first came to the USA in 2012, I learned about systems research. In 2013 I got an internship at Cisco and even though the technologies seemed amazing, they never felt as accessible as open source alternatives.

2013 was the beginning of the OpenFlow protocol hype, whose main value proposition was the separation of control plane and data plane. This separation would allow for a new wave of innovation as software development proceeded independently from hardware implementation cycles.

Vendor APIs already existed before OpenFlow, all under NDAs. Want to develop a new network protocol? Sign an NDA with Broadcom, or Mellanox or whoever is your hardware provider and then get to work. OpenFlow proponents identified this inefficiency and argued that this feedback cycle was suboptimal. Eventually some vendors started to open their APIs, though I’m not certain of the current state.

OpenFlow and Software Defined Network (SDN) fascinated me, and when I decided to pursue a master’s degree, I was very happy to join Georgia Tech where they conducted substantial research in SDN. The level of innovation achieved with these open source-based technologies was unparalleled, at least from my perspective.

Open Innovation’s Real Impact

Open standards and open source proved to be philosophies that were not only effective but also efficient. Operators such as Google and Facebook used this momentum to drive vendor innovation. The research flourished, with remarkable quality in results obtained using SDN technology and paradigms.

Critics often claimed these results could be achieved with other proprietary technologies. However, Google’s B4 papers documented functional high-utilization networks that remain unmatched by proprietary alternatives. Common knowledge today is still that networks should be over-provisioned - everybody simply gave up on QoS.

My thesis is not that these problems are too complex, but instead the closed systems we have today are too complex. I hypothesize that open systems would facilitate learning and accelerate innovation by reducing barriers to experimentation.

SDN’s Legacy and Market Validation

SDN research drove significant innovation during the past decade. Although production deployments of OpenFlow were limited to Google and select research networks, it certainly created sufficient market pressure to drive incumbent vendors to update their marketing material. It also drove material innovation towards network programmability plus a few startups.

In enterprise environments, SDN transmuted itself into network automation, focusing on practical improvement metrics rather than architectural transformation. Automation proponents were all about the low-hanging fruits - using interfaces and systems that already existed, such as Ansible and SSH. This represented another win for open source. No closed solution has achieved comparable adoption rates, documentation quality, and innovation velocity.

Contemporary Challenges and Predatory Capitalism

At the same time, open source communities are often prey to predatory capitalism systems that attempt to purchase open development ecosystems, purposefully stifling innovation. Take Ansible for example: Red Hat acquired it in 2015, and since then, development increasingly focused on Red Hat’s proprietary products. Community contributions were deprioritized in favor of paid features, slowing innovation—a pattern observed repeatedly across technology markets.

Another example is Terraform, the most used configuration management tool today which was open-source until 2023. I predict a similar development curve to Ansible; it will lose market share to the next truly open-source tool.

Fast forward to today: I’ve heard of networking technologies for Artificial Intelligence that all seem to be closed source. But how good are those really if you have no open benchmarks? The absence creates information asymmetry between vendors and customers, leading to suboptimal resource allocation. It’s essentially marketing teams competing secretly with engineering teams.

In the broader AI domain, while several closed models exist, openness clearly has played a role in its popularization. OpenAI leveraged openness to attract talent and build momentum—hey, it says “Open” in the name. Has it been the bait-and-switch of the decade? Also yes. A non-profit organization transforming into a for-profit worth billions while restricting access to its models directly contradicts its original mission. This represents perhaps the most striking example of economic tension between open innovation and profit maximization.

Scientific Progress and Open Reproducibility

I believe openness is crucial for innovation and science. Closed scientific experiments create huge barriers to reproducibility and falsifiability, foundational principles of the scientific method.

Government-funded research should be required to be open and reproducible to maximize return on public investment. The knowledge spillover effect demonstrates that open research generates higher social returns. This brings me to the issue of intellectual property.

Patent systems function primarily as priority documentation mechanisms enforceable through judicial processes. They exist to protect inventors. While I recognize intellectual property rights, their efficiency in promoting innovation is questionable. Most private innovation and inventions couldn’t exist without the foundation of public knowledge.

Most of us are making small improvements on public knowledge instead of writing our own proprietary software stacks. I’ll admit my own ignorance here—I don’t claim to know the optimal approach to technology licensing. The economic and legal complexities are substantial.

The Funding Paradox

Most companies operationally benefit from open source software through reduced costs and increased flexibility, yet investment structures predominantly favor closed development models. This creates a market failure where technologies with the highest positive externalities struggle to attract adequate funding via traditional investment frameworks.

How about competition? Would any startup ever be competitive if it had to write its own post-quantum cryptography? Should we all just be customers of Red Hat and blame it all on them? I’d much rather support an environment where innovation is required to be open than an environment that pretends innovation is private.

If governments are driving research, they should promote openness. Historically, there are numerous examples of working groups successfully driving open technologies. At the same time, I do get afraid that we might all become like Europe with rigid regulatory bodies that achieve little practical progress.

Conclusion

I believe open source is not only beneficial; but also essential for meaningful technological advancement. Yet the challenge of creating sustainable models that reward contribution while maintaining accessibility remains unresolved. Finding economic models that support innovation while preserving its accessibility represents perhaps the most significant challenge for technologists and policymakers.

I’d be curious to hear your thoughts.

Written by Flavio Castro on May 12, 2025