[munificent]

I'm on the Dart team and I agree with pretty much all of this.

Personal opinion time:

Lars and Kasper, the original leads and creators of the language came out very confidently with this mission to get Dart natively supported in the browser based on the assumption that the language was so good and the VM would be so fast that users would clamor it.

They had the best of intentions — they really did want to make a delightful, productive, fast language. They intended to move the entire web forward the same way V8 had when it first launched, and they believed deeply that Dart would enable that.

But I think they really underestimated how vastly different designing and marketing a language is from simply implementing an already-successful one.

It's not enough to just have a good product. The way you present it is often more important. And you don't even have the luxury of defining "good product" — you must be mercenary in letting your users' needs override your personal preferences. All of that was a real struggle for them.

For what it's worth, there's been a lot of staff changes over the years since Dart first launched. Lars and Kasper have left to try the startup thing again (which I think is a better fit for their skills and desires than evolving a big open source language). The team we have now, I believe, is much better aligned with what you're saying.

It sucks that we do have this baggage, but I hope we can improve our reputation over time. I'm hopeful we can — it took Java several tries before it found its footing.

[kbd]

I think a lot of what people found off-putting about Dart is that Google already recently came out with a new language that has become very successful. Announcing another one, which at first glance looks like yet another curly brace language not that far from Java, was a step too far for people to be interested.

Nowadays, the only thing driving Dart seems to be Flutter, which again is head-scratching. Like, "Flutter sounds interesting but in addition to learning Flutter I need to learn a brand new programming language for it that's really only used for Flutter?" People only have so much bandwidth for new things and there's already so much to constantly learn in our industry.

[munificent]

I could be wrong, but I don't believe the existence of Go was much of a problem. Go and Dart are very different languages aimed for very different use cases. If anything, I think Go helped us because we could point to it as a successful language that Google hasn't cancelled.

> not that far from Java

This is something that hurt us, I think. You can write Dart code that looks a lot like Java, and many of our early public examples did. To make matters worse, for no good reason, Dart 1.0 didn't do any type inference, so even though you could use "var" for local variables, doing so would give you a worse user experience.

But you don't have to write Java in Dart. You can write really elegant, clean code in Dart in a way that Java doesn't enable. You don't have to stuff everything inside classes. We've always had nice terse lambdas, higher-order functions, collection literals, etc.

Here's a random little program of mine:

    import 'package:hauberk/src/engine.dart';
    import 'package:hauberk/src/content.dart';

    main() {
      var content = createContent();
      var save = content.createHero("blah");

      while (true) {
        var watch = Stopwatch();
        watch.start();

        // Generate a dungeon at each level.
        var count = 0;
        for (var i = 1; i <= Option.maxDepth; i++) {
          var game = Game(content, save, 1);
          for (var _ in game.generate());

          // Read some bit of game data so the JIT doesn't optimize the whole
          // program away as dead code.
          if (game.hero.pos.x >= -1) count++;
        }

        watch.stop();
        print("Generated $count dungeons in ${watch.elapsedMilliseconds}ms");
      }
    }

It's not the most beautiful code in the world, but I do think it's a good bit simpler and cleaner than it would be in Java.

[sbjs]

The thing is, this is almost identical to well written JavaScript. In fact you could do almost all of this in JS even before ES5. So the non-Java-like features you're using here were already present, which again suggests that Dart isn't really getting past the "Good Enough" problem.

It's also something you could do in Java, with the exception of having to wrap main with the obligatory EntryPoint class.

I think this just goes to show that good code can be written in almost any language (except Perl of course) because writing good code almost always just means writing obvious code, which necessitates avoiding cleverness and obscure features (literally all of Scala), twisting control flow (I'm looking at you State monads and call/cc), or adding too many indirections due to limitations of the language (Java).

[munificent]

> In fact you could do almost all of this in JS even before ES5.

Well you do get static type checking if you write it in Dart, but I see your point. :)

[lizmat]

Why was "except Perl of course" needed here? Do you have any data to back up that you can not write well written code in Perl? Or are you just simply venting your prejudices?

[andrewbinstock]

> I could be wrong, but I don't believe the existence of Go was much of a problem.

It wasn't so much Go per se, but the fact that Google was pushing two new languages at the same time. It created the perception, for me at least, that Google was pushing out new languages to see which ones stuck and kill those that didn't. Google had/has the reputation for killing off projects that it lost interest in. And I wasn't willing to learn new a new language and its ecosystem only to be marooned there a few years later.

So, while I was definitely interested in alternatives to JS, there was no way I'd pick up on Dart.

[kbd]

Totally agree. Also, Go was refreshing. Intentionally limited, yes, but simple-yet-powerful. Then they show off Dart and it looked like yet another 90's-style curly-brace language, and I'm sure I'm not the only one who asked "why bother?".

[wmfiv]

That is nice. It's also almost indistuishable from Java 10. The string interpolation at the end is the last big syntactic nicety Java is missing - at least relative to this example.

[pjmlp]

Yes it is a little nicer than Java 11, but not so much to really make a difference, while throwing away all the Maven central goodies.

    import engine.*;
    import static content;

    import static java.lang.System.out;

    public class Demo {
        public void main(String args[]) {
            var content = createContent();
            var save = content.createHero("blah");

            while (true) {
                var start = System.currentTimeMillis();

                // Generate a dungeon at each level.
                var count = 0;
                for (var i = 1; i <= Option.MAX_DEPTH; i++) {
                    var game = new Game(content, save, 1);
                    for (var ignored : game.generate());

                    // Read some bit of game data so the JIT doesn't optimize the whole
                    // program away as dead code.
                    if (game.hero.pos.x >= -1) count++;
                }

                var stop = System.currentTimeMillis();
                var elapsedMilliseconds = start - stop;
                out.printf("Generated %d dungeons in %dms\n", count, elapsedMilliseconds);
            }
        }
    }

[kbd]

> I do think it's a good bit simpler and cleaner than it would be in Java.

Is it really? My Java is rusty, but now that Java has type inference it looks nearly exactly like what you'd type in Java (minus `public static void`, `new`, etc.)

[cutler]

Java 10 only has type inference for method-scoped vars.

[pjmlp]

Java 11 is around the corner and has improved upon it.

[nailer]

I removed everything that seems redundant. Less syntax means less to break, and less for new programmers (who are 10x the volume of existing programmers) to learn:

    import 'package:hauberk/src/engine.dart'
    import 'package:hauberk/src/content.dart'

    content = createContent()
    save = content.createHero("blah")

    while (true) 
    watch = Stopwatch()
    watch.start()

    # Generate a dungeon at each level.
    count = 0
    for (i = 1; i <= Option.maxDepth; i++) 
        game = Game(content, save, 1)
        for (_ in game.generate())

            # Read some bit of game data so the JIT doesn't optimize the whole program away as dead code.
            if (game.hero.pos.x >= -1) count++      

    watch.stop()
    print("Generated $count dungeons in ${watch.elapsedMilliseconds}ms")

[aoeusnth1]

how are the semicolons in a for loop redundant? Without them it's extremely difficult to parse what's going on.

[nailer]

Typo when removing the others. Thanks for the spot. Maybe a forEach or map might be more appropriate for all those future programmers anyway though.

[emmelaich]

While you're there, you may as well indent everything after the "while(true)"

[laxd]

Separating variable declaration and reassignment at a glance makes code more clear. Indenting loop bodys makes code more clear, and the curly braces won't suddenly break. Keeping lines short makes it nicer to work with in various editor setups.

[nailer]

> Separating variable declaration and reassignment at a glance makes code more clear.

Hrm, maybe. Python does well not using them - if you reassign something, it's being reassigned. Maybe, since declaration is more common, make declaration the default and reassignment explicit?

> Indenting loop bodys makes code more clear

Glad you agree.

> curly braces won't suddenly break

They do all the time, everytime indentation (how people read coe) gets out of sync with braces. Hence avoiding redundancy.

> Keeping lines short makes it nicer to work with in various editor setups.

Displaying code to match the screen is the editors job. Sometimes your screen is 30 characters, sometimes it's a lot more.

[Daegalus]

It's a shame, I was a HUGE dart for backend proponent. I made libraries and even middleware for API servers and such. I was constantly fighting to get ANY focus on the backend standard libs. In the end, I had to give up and stopped fighting. I still maintain my uuid, base32, and otp libraries, just upgraded them for 2.0, but Its a shame because Dart could have so easily competed with Node.js and it was a major replacement for it for me.

[sbjs]

A language by itself is almost never useful, every language needs a killer app, a concrete tangible thing that that language can do better 100x better than any alternative. Like how Ruby had Rails or JS had browsers or Python has ML. That’s not to say you can’t do other things with these, just that their competitor is good enough and has momentum and mind share. It sounds like Dart now has Flutter and I think that’s going to be a huge deal moving forward.

[hota_mazi]

> A language by itself is almost never useful, every language needs a killer app,

I don't understand why this myth is still around.

C++ never had a killer app to start its momentum.

Neither did Javascript. Nor Java really (or maybe applets, back in the days?).

Languages can succeed on technical merits alone if they come out at the right time and fix real problems that programmers experience on current mainstream languages.

[munificent]

I think you need at least one of three things to get a new language off the ground:

1. Extreme compatibility with the current entrenched language so you can incrementally migrate. C++ from C. CoffeeScript and TypeScript from JavaScript. Kotlin from Java.

2. A killer app (well, framework). Rails for Ruby. WinForms for C#. Applets and J2EE for Java (later Android).

3. A new platform where you must use the language to target it. C for UNIX. Objective-C for iOS. JavaScript for the browser.

There are a few exceptions here and there, but the above are the typical well-trod paths to success for a language.

[hota_mazi]

I listed two clear counter examples to claim #2, C++ and Javascript. I'm even tempted to put C in that list, because C was already popular before Linux.

What was Python's killer app?

What is Rust's killer app?

Go's?

Kotlin's?

Java is a counter example to your claim #3 since it not only does it run on all OS'es but you can also develop it on all OS'es.

[platinumrad]

"at least one of three"

[realusername]

> What was Python's killer app?

Not a killer app, but the fact you have so many ML frameworks

> What is Rust's killer app?

That's clearly Servo.

> Go?

It was docker which really made Go take off, it was their first big platform.

> Kotlin's?

Android compatibility, before that it was a niche language not being picked up much.

[jay_kyburz]

Python already won before ML was a thing.

[wild_preference]

Killer app is the wrong phrase. Rather every new language needs a competitive upside over existing languages to be successful. At which point your proposed questions have easy answers.

[pjmlp]

C++ and Javascript are definitely not counter examples as I commented.

[flukus]

Outside of the HN bubble python is the only one of those that got off the ground so far.

[pjmlp]

> C++ never had a killer app to start its momentum.

C++ was immediately adopted by all major C compiler vendors as it came from AT&T as well.

Microsoft C/C++ 7.0 for MS-DOS was the very last one to adopt C++ on the toolchain.

It was the lingua franca of all major desktop OS GUI frameworks, leaving C just for their lower layers, MS-DOS (Turbo Vision), OS/2 (C/Set++), Windows (OWL, VCL, MFC), Mac OS (PowerPlant, MacApp), BeOS, Symbian.

If you were doing plain C GUI programming you were doing it wrong.

> Neither did Javascript.

JavaScript owns the browser, there is hardly any other alternative regarding "browser systems programming".

> or Java really (or maybe applets, back in the days?).

Java's portability was a gift compared with C compilers still in the middle of migrating from K&R C to ANSI C, C++ compilers which were silos with their GUI frameworks and a standard still a couple of years away to be fully defined.

[sbjs]

These days it's mostly about apps, because apps are the main thing that facilitate the transfer of $$$. In the 80s and 90s, enterprises were full of $$$ and needed fast and safe languages for writing all their software. C++ came out to solve this need, and Java came out saying it could do it better. Today enterprises are moving many of their internal tools to the web. Enterprises don't need apps per se, but they do need the tech needed to build internal apps and apps to bridge between them and their partners. Consumers are using desktops less, and mobile or web more. Apps are basically the current de facto software "product". But a lot more goes into making apps today than 30 years ago. An example is how most apps need ML for spam protection and recommendation algorithms, or they are instantly going to be considered outdated and behind the times.

[coldtea]

>In the 80s and 90s, enterprises were full of $$$

More than today?

[sbjs]

I haven't had my caffeine yet today cut me some slack :)

[coldtea]

Golang and Rust also didn't have any killer app.

[nwienert]

Rust definitely has Firefox

[cutler]

Just thought I'd remind you that there are currently 0 Rust jobs on Indeed.co.uk by title for the whole of the UK. The HN bubble is not how the world at large operates.

[pjmlp]

Two examples from Indeed.de in Germany, there are other results.

https://de.indeed.com/viewjob?jk=6c49a2deeb389df8&q=rust&tk=...

https://de.indeed.com/viewjob?jk=6823f26122417b2d&q=rust&tk=...

[coldtea]

Firefox is not the kind of app to drive adoption of the language (like Rails or Python scientific libs etc). It's just an app, people using it could not care less what it's written in.

[sbjs]

Go has Docker

[coldtea]

Yes, but it didn't have anything before it, and yet it gained adoption like crazy.

[megaman22]

Go has a bit of a niche, and Rust is trying real hard, but they aren't Javascript or Java, or C#, or Python, or Rails, or PHP, or C++, or Objective-C or... Out in the wider world, they are very much playing catchup, and have a very far way to go.

[billchristian]

Javascript had the web. And, Java had the JVM. A killer 'app' is not necessarily singular & discrete.

[pier25]

Off topic but... are there any plans to adapt Flutter for desktop apps?

[MarkMc]

I'm the lead dev on just such a project: https://feather-apps.com/

[pier25]

Looks great!

I beat the Reversi app on first try. I'm guessing it's not very strong. :)

[MarkMc]

Thanks :)

Haha yes I weakened the AI somewhat compared to the original version created by RedBrogdon. It's not fun to be constantly thrashed by an opponent who takes 0.2 seconds per move! Despite my changes I still lose about 3 out of every 4 games :(

[kyrra]

There is an unofficial implementation for it.

https://github.com/google/flutter-desktop-embedding