Programs should be Small

What are you supposed to do? Find the int main() and then make the program run in your head?

I can make an analogy with a car - I don't know every one piece of it but I can infer from the context. The scale is evident.

So why do larger programs 'win' in the open source world as well[1]? Pop psychology about management doesn't seem sufficient to explain the phenomenon (although I'm sure it is a good way to sell a '101 habits of highly effective managers' book or get paid to give talks about management). Large systems are large, breaking them up into smaller pieces doesn't change that, but it makes navigating the code base harder (although I assume you don't care about that since judging from your blog posts you don't think tooling is important). It makes your interfaces less malleable (can be good can be bad), and moves a lot of communication to places where the compiler can't warn you about mistakes (again, if you don't care about tooling I guess this doesn't matter… but I would argue that this is bad).

It seems to me like systems of many small separate processes is basically dynamic OOP. Everything is late bound, dynamically typed and async. It's easy to make changes and also easy to break things. You can argue that this is better for certain problems, but I don't think it's universally better, and the community seems pretty divided on the issue too: look at the popularity of Go, statically typed and building concurrency into the language rather than using the OS like in the older C world.

Aa an aside; surely the web developer community is eventually going to grow tired of talking about how terrible Java is and how $idea_of_the_moment is good because it's 'not java'? As an outsider the obsession seems extremely unhealthy, and leads you to bizarre places like arguing against automated refactoring or interactive debugging or static type systems just because those things are associated with Java. I guess to maintain credibility I also need to point out that I don't and have never used Java…

[1] Firefox/Chromium vs uzbl, gcc/llvm/clang vs pcc, gdb vs printf debugging, sqlite/mysql vs directories and plain text files, perl vs sed/awk/grep shell scripts, emacs/vim vs ed/notepad etc etc.

I'm actually a pretty big fan of static typing. You don't get static typing's main benefits in C++ or Java, though. You have to use a language like Haskell or Ocaml, or the right subset of Scala, to see the major benefits of that.

Open-source is a bit different because people choose whether they contribute to a project. The quality of code in the active open-source world is leagues above what you find in typical enterprise codeballs, because of survivor bias. No one has the authority to mandate that code be maintained by others, so the messes are cleaned up by people who actually care, not people slogging through it to keep a paycheck coming.

The big-program methodology of the corporate world is the evil. In FOSS, the major projects are an unusual set-- code-quality at a high level just not seen in the for-paycheck commodity-engineer world and large because of success-- rather than the reverse. There's a survivor bias that occurs because the best projects are the only ones people pay attention to.

The corporate world is screwy because projects become large or small based on political reasons that have nothing to do with code quality. In the FOSS world, code-quality problems related to growth will be self-limiting because no one has the authority to "force" the program to grow.