In the long run, GPL code becomes irrelevant (2015)

Original Article ↗ Hacker News Discussion ↗

Core thesis and its limits

  • Article’s claim: in the long run, permissive code wins because corporations will rewrite around GPL; GPL projects get sidelined.
  • Many agree this matches recent trends in popular stacks (MIT/Apache everywhere, GPL largely in niches or “enthusiast” projects).
  • Others say this is overstated or outdated: several big GPL projects (Linux, Git, Blender, Krita, QGIS, MySQL, etc.) remain central.

Corporate behavior and incentives

  • Permissive licenses lower legal friction, so large firms ban GPL entirely to avoid “license contamination” of proprietary code.
  • Economic argument: upstreaming patches once is usually cheaper than maintaining a private fork, so permissive projects often get substantial corporate contributions.
  • Counterpoint: when modifications are strategically valuable, companies will keep them private regardless of license, or rewrite from scratch if GPL-encumbered.

User vs developer freedom, and “fairness”

  • One camp: GPL centers user freedom and creates a “ratchet” that prevents proprietary capture, even if that annoys proprietary developers.
  • Another camp: most users care more about convenience and quality than license fairness; they’ll pick the best product, even if proprietary.
  • Some argue rewriting GPL code just to avoid sharing is wasted human effort; others see making bad actors “pay the full cost” as a feature.

Examples and case studies

  • GCC vs LLVM/Clang and EDG cited both as evidence that permissive code can displace proprietary, and that corporate-funded permissive projects can overshadow GPL alternatives.
  • Linux vs BSD: some see this as GPL’s triumph; others attribute it to historical timing and network effects more than license.
  • Web engines: permissive+weak-copyleft engines (Chromium/WebKit/Gecko) and the demise of proprietary engines (IE, EdgeHTML) are seen as supporting the article’s “complexity economics” story.

MPL/LGPL as middle ground

  • Multiple commenters highlight MPL/LGPL as a practical compromise: code remains open, but they don’t force entire downstream applications to be copyleft.
  • Browser engines and some libraries (Qt, GEOS, Servo) are used as evidence that weak copyleft may have the best long-term “survival characteristics”.

Cloud, relicensing, and AI

  • Redis, Elasticsearch, Terraform: permissive cores later relicensed to “source available” to fight cloud providers; community permissive forks (e.g., Valkey) then emerge.
  • Some see this as evidence permissive licensing invites corporate capture; others note the open forks remain viable.
  • A few argue AI makes all licenses weaker by making “clean-room” rewrites cheap, shifting the freedom battle from source code to model weights.