I'm going to throw this idea out there: companies can succeed independent of type checking.

> Dynamically typed languages for large code bases are terrible

was the original argument. Success/failure of entire companies has somehow been brought into it.

That's really going out on the limb though...

But you're right. Moreover I noticed obsession with a particular language is often a hallmark of a junior engineer or a recent grad. This this idea of "If you don't use Go/Node.js/etc you'll totally fail".

HN skews young, I'm sure they'll grow out of it.

    Prove how?

By means of natural selection.

    They could have succeeded in spite of
    dynamic typing, not because of it.

You think the otherwise smart and successfull founders of these projects all made the wrong choice when it came to selecting the right tool for the job?

    Facebook defined a new programming language

No, they wrote a faster runtime.

    to add static typing to HHVM

No, they added type annotations.

    There’s a lack of good statically-typed languages

No, in the past statically typed languages were the norm. C, C++, Java...

It's just that dynamic languages outcompeted them.

> It's just that dynamic languages outcompeted them.

All "trending" languages are statically typed (Rust, Go, Swift, Kotlin, ...) and all the popular JavaScrip-improved-languages are statically typed as well. Even further, many dynamically typed languages have pushed in recent years to add type annotations or even static typing (e.g. mypy).

Dynamic typing is the brainchild of the 90s for rapid initial development. Sustainability has been a big issue from the outset.

Dynamic typing is from 1958 Lisp. Your history is off a bit there, by some forty years.

The 90's also saw much commercial experimentation with visual environments (e.g. Visual Basic, Delphi, JBuilder) with round-trip conversion between the layout form and underlying code. They were promoted for R.A.D. (rapid application development -- remember that acronym?) but, like dynamically typed languages, they also had sustainability issues.

Of course, like dynamic typing and Lisp, visual environments also had Smalltalk from way back, but they were before big business became infatuated with the ideas.

How do products like Delphi have sustainability (maintainability ?) issues ? Delphi/Object Pascal is statically-typed, so you can't even build an application that has issues with the code-behind for a form.

You can see errors at run-time from mismatches between the code-behind and the forms, but this has nothing to do with the language. It's just the way that certain products like Delphi decided to handle such mismatches. You'll see these types of issues with any code that performs run-time serialization/deserialization of class instances. Our product, Elevate Web Builder, uses a similar architecture with its WYSIWYG form designer/forms (they use JSON instead of a custom key-value format), but it simply ignores properties in forms that don't exist in the actual code.

Lisp has nothing to do (in many ways) with the late 80s and 90s wave of dynamic languages.

JavaScript was definitely partly inspired by Scheme.

likewise ruby by lisp

I’d say Ruby’s closest sibling is Smalltalk personally.

People say this mainly because of the brace-enclosed blocks of code being functions. However, that is really just a syntactic sugar for lambda (and can even have explicit arguments: (lambda (x y) ...) <--> {|x,y| ... }.

there's the "everything is objects and messages sent to them" bit too, which is definitely from smalltalk. ruby took the good bits from several languages.

I've still not seen a single "sustainability issue" in Python code bases which seem to stem from dynamic typing as opposed to static typing.

I think Elixir is trending, and it's not statically typed, right?

That said, I do agree that static typing is in vogue.

Smart answer successful founders of Wikipedia took an existing wiki engine. Their aspirations were not in the software development area.

Highly dynamic languages like Python or PHP or Perl are easy to start with, so they're great for prototypes. Then either a serious rewrite follows, using a statically checked language, or scaffolding is erected to continue to make do with the dynamic code. Microservices help!

Erm, so splitting big Python projects into microservices is proof that dynamic typing is unsustainable?

No; but a single microservice is a smaller code base that a large monolith, so the problem . Also, you normally control the microservice's interface more strictly, and change it less often, and care about compatibility and versioning more, than for a typical internal API.

If type annotations support a static verifier, they implement static typing (which, after all, is just having a static verifier that uses some combination of annotations and inference to validate types.)

> You think the otherwise smart and successful founders of these projects all made the wrong choice when it came to selecting the right tool for the job?

Not at all! I think dynamic typing hurts code, but there are:

- not always alternatives

- many other slightly suboptimal choices that it’s quite possible to succeed in spite of

> No, they wrote a faster runtime.

I’m referring to Hack, which is in fact another language with static typing. (Type annotations are static typing with compromises, and you can see what’s being gone for.)

> No, in the past statically typed languages were the norm. C, C++, Java...

Those are not remotely good. Some examples closer to being good are Rust, a step in the right direction that’s still unfolding, and Haskell, a wonderful language hampered by a slow compiler and a lack of decent packages. (And probably several less popular languages, but popularity is an important factor…)

"Those are not remotely good. Some examples closer to being good are Rust, a step in the right direction that’s still unfolding, and Haskell, a wonderful language hampered by a slow compiler and a lack of decent packages. (And probably several less popular languages, but popularity is an important factor…)"

Wait: so you are arguing that startups should use languages that are either incomplete (Rust) or that have slow compilers and poor library support (Haskell), or possibly some other "good" language that doesn't actually even exist?

I’m saying that dynamically-typed programming languages are the best choice right now, but that it’s not because they’re dynamically-typed; existing fare for statically-typed languages is just worse.

Probably should have clarified that earlier since it looks like I agree with the statement that choosing Python was “a mistake”; sorry. I don’t. Vehemently.

> No, they wrote a faster runtime.

Which includes a statically-typed version of PHP, namely Hack, that they use for their code.

> No, they added type annotations.

Hack's type checking is much more sophisticated than mere “annotations”. That label might be better suited to PHP's type system.

>You think the otherwise smart and successfull founders of these projects all made the wrong choice when it came to selecting the right tool for the job?

Nobody is perfect. Mark Zuckerberg has had his fair share of controversies for example.

>No, in the past statically typed languages were the norm. C, C++, Java...

Not good examples for your argument. These are largely considered poorly statically typed. Compare to ML, Haskell, Ada, Rust, Swift for reference.

"They could have succeeded in spite of dynamic typing, not because of it."

Yes, they could have. Evidence for your hypothesis would be the existence of similarly successful companies that used compiled statically-typed languages from the beginning.

Are there any recent ones?

Are you suggesting that there are no similarly successful companies that used Java, C# or C++?

Are you suggesting there are?

I asked for examples. Can you provide one?

The same can be said about any company to justify your opinions, like Linkedin, they use Java and Scala and their page speed is so slow. Compared to Facebook they are like a 1900' tractor. A facebook page opens in less than a second, a Linkedin page in 5.

I think it's the fact LinkedIn uses Ember. Anything built with Ember is sluggish and bloated

>Prove how? They could have succeeded in spite of dynamic typing, not because of it.

Then maybe the burden of proof is on you?

If you started your career recently, it's easy to be caught up into this "static typing" as some kind of the only way.

And until you have gone back and forth a couple of times, building moderately complex systems; it's difficult to appreciate the advantages that static checking bring to the table. Sure, it's possible to achieve the same effect with unit tests; but having half of those tests written and rewritten for you by the compiler definitely helps. Back in the days, I wanted to buy into the Python/Ruby hype as much as anyone; but having implemented several complex systems in Python/Ruby, I'm not that excited about the idea anymore; once they pass a certain point, agile begins to feel more like Jello.

> Then maybe the burden of proof is on you?

I’m saying that the parent doesn’t prove its point, and I’m not attempting to prove its parent’s point.

> If you started your career recently, it's easy to be caught up into this "static typing" as some kind of the only way.

I didn’t, and the bulk of what I’ve built has been in languages with dynamic typing. See https://news.ycombinator.com/item?id=14675577. Dismissing stuff as “recent starts” is pretty cheap, anyway.