Zig and Rust

Agree. Also agree that it’s possible to write performant software in the languages you mentioned.

> where a native-compiled GC language would be a clearly better long-term choice.

I don’t think this is the case, and I don’t think you’ve made the case for it. I think there’s some implicit assumption that the GC gives a better developer experience and that’s why it’s a better choice? In a vacuum maybe, but the dev ex in Rust is better than Nim for a number of reasons.

It’s easy to see if you’re right or they are. In 5 years I’m willing to bet that the tooling for the JS ecosystem is all written in Rust, Zig or Go. Nothing in Nim or D.

You might see that and think “oh, it’s because they made suboptimal choices”. I see it differently though.

I would only caution you about assuming that people are following the zeitgeist instead of making sound technical decisions. For example there’s only one language that prevents race conditions at compile time. Maybe they write more multithreaded code than you did and really prize this feature.

Maybe they like the libraries that make the ecosystem rich. Does nim have anything as polished and performant as clap and serde? Maybe they like the generally high quality of documentation in the Rust ecosystem? I could point out several other legitimate axes on which Rust is a superior alternative.

It doesn’t matter, because you’re aware of most or all of these. What’s important is for you to avoid falling into the fallacy of “what’s popular must be bad, because the average person isn’t as smart as me”. Lots of smart people fall into this trap. It’s especially bad to fall into this when you’re thinking about platforms because the utility of a platform is directly proportional to its popularity.

Language X might have so many flaws, but if it’s popular then the community around it could find myriad ways to mitigate the flaws. They’ll release a million useful libraries that make our lives easier. And if the language gets 50% more popular, it gets 50% more useful to us because the ecosystem gets richer. Better ecosystem means more companies adopt it => more jobs available => it’s a sound choice to learn this language in your free time => more devs available with this skill => it's a good idea for more companies to adopt it. It’s a fly wheel of adoption. Java benefited from this, even if it wasn’t a perfect language. Rust benefits, but to a lesser extent because there’s so many good alternatives now (including Java!)

Even if Rust doesn’t do it for you in 2023, I’d counsel taking a look again in 2026. The growth around the language (80% per year in crate downloads, 30% per year in crate authors) means that it’s possible that you might find most of your pet peeves might have been addressed by then. Either in the main language or in a library.

"Polished" and "high quality" are more subjective/implicitly about adoption, IMO. "Performant" has many dimensions. That said, I just tested the Nim https://github.com/c-blake/cligen vs clap: cligen used 5X less object file space (with all size optimization tweaks enabled in both), 20% less run-time memory for large argument lists, and the same run-time per argument (with march=native equivalents on both, within statistical noise). cligen has many features - "did you mean?/suggestions", color generated help and all that - I do not see obvious feature in clap docs missing in cligen. The Nim binary serde showing is unlikely as good but there are like 10 JSON packages and that seems maybe your primary concern.

More to add color your point than disagree (and follow up on my "adoption") - your ideas about polish, quality, docs, etc. are part of feedback loop(s) you mentioned. More users => Users complain (What is confusing? What is missing? etc.) => things get fixed/cleaned up/improved => More users. Besides "performant" being multi-dimensional, the feedback loop is more of a "cyclic graph". :-) Or maybe this is just a sub-graph of "ecosystem richness". While I probably prefer Nim as much or more as @netbioserror, I am not too shocked by the mindshare capture. It seems to happen every 5..10 years or so in prog.langs.

While many of your points are not invalid, tech is also a highly hype-driven & fad-driven realm. In my experience, the more experience with this meta-feature that someone has, the more skeptical they are of the latest thing (more rounds of regret, etc.). Also, that feedback graph is not a pure good. Things can get too popular too quickly with near permanent consequences. ipv4 got popular so quickly that we are still mostly stuck on it 40 years later as ipv6 struggles for penetration. Whatever your favorite PL is, it may also grow features too fast.