I have a saying that summarizes my opinion of Rust compared to Go: “Go is the result of C programmers designing a new programming language, and Rust is the result of C++ programmers designing a new programming language”. This isn’t just a metaphor - Go was designed by plan9 alumni, an operating system written in C and the source of inspiration for many of Go’s features, and Rust was designed by the folks at Mozilla - whose flagship product is one of the largest C++ codebases in the world.
The values of good C++ programmers are incompatible with the values of good C programmers1. Rust is a decent C++ replacement if you have the same goals as C++, but if you don’t, the design has very similar drawbacks. Both Rust and C++ are what I like to call “kitchen sink” programming languages, with the obvious implication. These languages solve problems by adding more language features. A language like C solves problems by writing more C code.
I did some back of the napkin estimates of the rate at which these languages become more complex, based on the rate at which they add features per year. My approach wasn’t very scientific, but I’m sure the point comes across.
- C: 0.73 new features per year, measured by the number of bullet points in the C11 article on Wikipedia which summarizes the changes from C99, adjusted to account for the fact that C18 introduced no new features.
- Go: 2 new features per year, measured by the number of new features listed on the Wikipedia summary of new Go versions.
- C++: 11.3 new features per year, measured by the number of bullet points in the C++17 article which summarizes the changes from C++14.
- Rust: 15 new features per year, measured by the number of headers in the release notes of major Rust versions over the past year, minus things like linters.
This speaks volumes to the stability of these languages, but more importantly it speaks to their complexity. Over time it rapidly becomes difficult for one to keep an up-to-date mental map of Rust and how to solve your problems idiomatically. A Rust program written last year already looks outdated, whereas a C program written ten years ago has pretty good odds of being just fine. Systems programmers don’t want shiny things - we just want things that work. That really cool feature $other_language has? Not interested. It’ll be more trouble than it’s worth.
With the philosophical wish-wash out of the way and the tone set, let me go over some more specific problems when considering Rust as a C replacement.
C is the most portable programming language. Rust actually has a pretty admirable selection of supported targets for a new language (thanks mostly to LLVM), but it pales in comparison to C, which runs on almost everything. A new CPU architecture or operating system can barely be considered to exist until it has a C compiler. And once it does, it unlocks access to a vast repository of software written in C. Many other programming languages, such as Ruby and Python, are implemented in C and you get those for free too.
C has a spec. No spec means there’s nothing keeping rustc honest. Any behavior it exhibits could change tomorrow. Some weird thing it does could be a feature or a bug. There’s no way to know until your code breaks. That they can’t slow down to pin down exactly what defines Rust is also indicative of an immature language.
C has many implementations. C has many competing compilers. They all work together stressing out the spec, fishing out the loosely defined corners, and pinning down exactly what C is. Code that compiles in one and not another is indicative of a bug in one of them, which gives a nice extra layer of testing to each. By having many implementations, we force C to be well defined, and this is good for the language and its long-term stability. Rustc could stand to have some competition as well, maybe it would get faster!2
C has a consistent & stable ABI. The System-V ABI is supported on a wide variety of systems and has been mostly agreed upon by now. Rust, on the other hand, has no stable internal ABI. You have to compile and link everything all in one go on the same version of the Rust compiler. The only code which can interact with the rest of the ecosystem is unidiomatic Rust, written at some kind of checkpoint between Rust and the outside world. The outside world exists, it speaks System-V, and us systems programmers spend a lot of our time talking to it.
Cargo is mandatory. On a similar line of thought, Rust’s compiler flags are not stable. Attempts to integrate it with other build systems have been met with hostility from the Rust & Cargo teams. The outside world exists, and us systems programmers spend a lot of our time integrating things. Rust refuses to play along.
Concurrency is generally a bad thing. Serial programs have X problems, and parallel programs have XY problems, where Y is the amount of parallelism you introduce. Parallelism in C is a pain in the ass for sure, and this is one reason I find Go much more suitable to those cases. However, nearly all programs needn’t be parallel. A program which uses poll effectively is going to be simpler, reasonably performant, and have orders of magnitude fewer bugs. “Fearless concurrency” allows you to fearlessly employ bad software design 9 times out of 10.
Safety. Yes, Rust is more safe. I don’t really care. In light of all of these problems, I’ll take my segfaults and buffer overflows. I especially refuse to “rewrite it in Rust” - because no matter what, rewriting an entire program from scratch is always going to introduce more bugs than maintaining the C program ever would. I don’t care what language you rewrite it in.
C is far from the perfect language - it has many flaws. However, its replacement will be simpler - not more complex. Consider Go, which has had a lot of success in supplanting C for many problems. It does this by specializing on certain classes of programs and addressing them with the simplest solution possible. It hasn’t completely replaced C, but it has made a substantial dent in its problem space - more than I can really say for Rust (which has made similar strides for C++, but definitely not for C).
The kitchen sink approach doesn’t work. Rust will eventually fail to the “jack of all trades, master of none” problem that C++ has. Wise languages designers start small and stay small. Wise systems programmers extend this philosophy to designing entire systems, and Rust is probably not going to be invited. I understand that many people, particularly those already enamored with Rust, won’t agree with much of this article. But now you know why we are still writing C, and hopefully you’ll stop bloody bothering us about it.
Are you a free software maintainer who is struggling with stress, demanding users, overwork, or any other social problems in the course of your work? Please email me — I know how you feel, and I can lend a sympathetic ear and share some veteran advice.
Articles from blogs I follow around the net
This post gives an overview of the recent updates to the Writing an OS in Rust blog and the corresponding libraries and tools. I focused my time this month on finishing the long-planned post about Async/Await. In addition to that, there were a few updates …via Writing an OS in Rust April 1, 2020
Clipboard and drag & drop are arguably one of the most complicated parts of the core Wayland protocol. They involve a lot of back-and-forth communication between three processes: the application where some content has been copied, the compositor, and …via emersion March 26, 2020
Go always comes second to more basic concerns like personal and family health and safety. Around the world, the past couple months have been terrible, and we are still at the start of this awful pandemic. There are days when it seems like wo…via The Go Programming Language Blog March 25, 2020
Generated by openring