Top-30 GitHub projects with highest Truck Factor

Robert Cutts — Flickr: A bus waiting at a pedestrian crossing in Lima (via Wikipedia)

Truck Factor (a.k.a bus factor or lottery number) designates the minimal number of developers that have to be hit by a truck (or quit) before a project is incapacitated. It is a measure of the dependency of a project on a number of key members.

GitTrends is a web-based tool that estimates the Truck Factor (TF) of software projects, using an algorithm we recently proposed (for more information, see this preprint or this paper). GitTrends also tracks the popularity of these projects, as measured by their number of stars (more info, in these papers: paper #1 and paper #2).

http://gittrends.io

Here is a list with the top-30 GitHub projects with the highest TF values, as estimated by GitTrends:

  1. torvalds/linux: 163
  2. caskroom/homebrew-cask: 160
  3. robbyrussell/oh-my-zsh: 74
  4. github/github-services: 44
  5. fzaninotto/Faker: 32
  6. android/platform_frameworks_base: 24
  7. composer/installers: 26
  8. edx/edx-platform: 25
  9. mozilla-b2g/gaia: 25
  10. odoo/odoo: 22
  11. stympy/faker: 20
  12. freebsd/freebsd: 19
  13. qemu/qemu: 19
  14. php/php-src: 18
  15. uxsolutions/bootstrap-datepicker: 18
  16. saltstack/salt: 18
  17. openstack/nova: 17
  18. v8/v8: 14
  19. webscalesql/webscalesql-5.6: 14
  20. instructure/canvas-lms: 14
  21. apache/spark: 13
  22. jhollingworth/bootstrap-wysihtml5: 13
  23. docker/docker: 13
  24. silviomoreto/bootstrap-select: 13
  25. FFmpeg/FFmpeg: 12
  26. mongodb/mongo: 12
  27. git/git: 12
  28. jedi4ever/veewee: 12
  29. kubernetes/kubernetes: 11
  30. rails/rails: 11

Important notes:

  1. The results provided by GitTrends are an estimation, based on code authorship measures. As recommend to any software metrics, TF results should not be used without proper interpretation (in fact, to help on this task, GitTrends also provides historical data about TFs).
  2. Good software engineering practices (documentation, tests, etc) certainly contribute to overcome truck factor episodes (see again our previous paper). However, as an automatic tool, GitTrends does not consider these practices to estimate TFs.
  3. In case the TF developers leave a project, this does not necessarily mean the project will be discontinued, but that its maintenance and evolution will be in trouble. For example, bugs will take more time to get fixed, new features will take more time to be implemented, etc.
  4. GitTrends estimates TFs considering all the projects’ code available on GitHub. In the case of Linux, it includes source code that implements device drivers, for example. When we removed drivers, Linux's TF reduces to 57 (see our previous paper). The same happens with caskroom/homebrew-cask. After excluding the files in folder Library/Formula, HomeBrew’s truck factor reduces to 2.
  5. To create the list of 30 repositories, we only considered repositories with code. We removed repositories containing only documentation, books, tutorials, etc.