logo

drewdevault.com

[mirror] blog and personal website of Drew DeVault git clone https://hacktivis.me/git/mirror/drewdevault.com.git

Complicated.md (3472B)

    

  1. ---
  2. date: 2017-09-08
  3. layout: post
  4. title: Killing ants with nuclear weapons
  5. tags: [philosophy]
  6. ---
  8. Complexity is quickly becoming an epidemic. In this developer's opinion,
  9. complexity is the ultimate enemy - the final boss - of good software design.
  10. Complicated software generally has complicated bugs. Simple software generally
  11. has simple bugs. It's as easy as that.
  13. It's for this reason that I strongly dislike many of the tools and architectures
  14. that have been proliferating over the past few years, particularly in web
  15. development. When I look at a tool like Gulp, I wonder if its success is largely
  16. attributable to people not bothering to learn how Makefiles work. Tools like
  17. Docker make me wonder if they're an excuse to avoid learning how to do ops or
  18. how to use your distribution's package manager. Chef makes me wonder if its
  19. users forgot that shell scripts can use SSH, too.
  21. These tools offer a value add. But how does it compare to the cost of the
  22. additional complexity? In my opinion, in *every case*[^1] the value add is far
  23. outweighed by the massive complexity cost. This complexity cost shows itself
  24. when the system breaks (and it will - all systems break) and you have to dive
  25. into these overengineered tools. Don't forget that dependencies are fallible,
  26. and never add a dependency you wouldn't feel comfortable debugging. The time
  27. spent learning these complicated systems to fix the inevitable bugs is surely
  28. much less than the time spent learning the venerable tools that fill the same
  29. niche (or, in many cases, accepting that you don't even need this particular
  30. shiny thing).
  32. Reinventing the wheel is a favorite pastime of mine. There are many such wheels
  33. that I have reinvented or am currently reinventing. The problem isn't in
  34. reinventing the wheel - it's in doing so before you actually understand the
  35. wheel[^2]. I wonder if many of these complicated tools are written by people who
  36. set out before they fully understood what they were replacing, and I'm *certain*
  37. that they're mostly used by such people. I understand it may seem intimidating
  38. to learn venerable tools like make(1) or chroot(1), but they're just a short man
  39. page away[^3].
  41. It's not just tools, though. I couldn't explain the features of C++ in fewer than
  42. several thousand words (same goes for Rust). GNU continues to add proprietary
  43. extensions and unnecessary features to everything they work on. Every update
  44. shipped to your phone is making it slower to ensure you'll buy the new one.
  45. Desktop applications are shipping entire web browsers into your disk and your
  46. RAM; server applications ship entire operating systems in glorified chroots; and
  47. hundreds of megabytes of JavaScript, ads, and spyware are shoved down the pipe
  48. on every web page you visit.
  50. This is an epidemic. It's time we cut this shit out. Please, design your systems
  51. with simplicity in mind. Moore's law is running out[^4], the free lunch is
  52. coming to an end. We have heaps and heaps of complicated, fragile abstractions
  53. to dismantle.
  55. [^1]: That I've seen (or heard of)
  56. [^2]: "Those who don't understand UNIX are doomed to reinvent it, poorly."
  57. [^3]: Of course, [*"...full documentation for make is maintained as a GNU info page..."*](https://xkcd.com/912/)
  58. [^4]: Transistors are approaching a scale where quantum problems come into play, and we are limited by the speed of light without getting any smaller. The RAM bottleneck is another serious issue, for which innovation has been stagnant for some time now.