Hacker News Reader

Comment by alanfranzoni | Hacker News Reader

I hope that, at least, the new technology is better than the old one. But I have some objections:

- why can't old libraries/frameworks/things evolve, instead of forcing a breaking change? In many other languages evolution seems to preferred to revolution; in JS, evolution seems to be the norm.

- Is that "better and smarter way of working" THAT MUCH BETTER AND SMARTER? When working JS, it seems that a complete overhaul of an application (or of an ecosystem) is required every very few years. Ultimately, software development should lead to business value; does the improvement in generated business value justify the costs associated with such a tech twirl?

View on HN

>> Is that "better and smarter way of working" THAT MUCH BETTER AND SMARTER?

Absolutely yes. ReactJS is an application architecture that is a complete rethink of how to structure client side applications, and it is an architecture that is being applied outside JS and the browser because it is so much better. The fundamental concept of one way data flow from parent to child through the application requires a start from scratch approach, and would be severely compromised were it to be implemented as an "additional" evolution of some existing application framework such as AngularJS. Yes - that much better and smarter.

ES2015/ES7 are radically better languages that make hugely important changes to JavaScript. Yes - that much better and smarter.

Transpilers make it possible to run ES2015/ES7 even on browsers that support only ES5 - that is amazing - yes - that much better and smarter.

Build tools such as minifiers and packers and uglifiers perform magic to shrink and optimise code for a range of benefits including end users performance - Yes - that much better and smarter than sending a chunk of raw JS to the browser.

In short, yes, JavaScript is experiencing a continuous wave of really major and significant improvements that justify ditching the previous generation entirely.

I will say however that breaking changes give me the shits abut for the most part I am willing to endure them - I simply must assume that the changes are for the better because I don't have time to really understand every single change in every library that goes into my application - no JavaScript developer does - nor indeed does any Python developer try to understand every change to all the Python libs that they use.

If you're willing to use transpilers, why stick to JS at all, when you have a wealth of languages available?

Scala, Purescript, Haskell, Elm, ...

Because JavaScript

1. By far, has the largest ecosystem.

2. If you use a linter, a typechecker (TS or flow), and transpile your code (TS or babel or both), JavaScript is actually a pretty nice language to work in.

Sure it's a chore to set up all the initial tooling, and learn the ins and outs. However, once you're past that, it's usually not worth it to switch.

1. The ecosystem is available via bindings too (to most transpiled languages)

2. It's still far from the state of the art in PL design and missing:

* Nice static typing with type inference

* Nice guarantees about null access

* Nice guarantees about interop between client and server

* ...

>It's a chore to set up...and learn...

>However, once you're past that, it's usually not worth it to switch.

That is somewhat circular and can be said of virtually anything. That is, for every minute you invest in a tech, the degree to which it is worthwhile to switch to another tech diminishes. I'm speaking primarily about tech like frameworks that is aimed at making us more productive.

The exception might be a truly revolutionary new tech that sharply boosts productivity, but also has a very low learning curve. However, that scenario seems particularly unicorn-ish where JS frameworks are concerned.

The closer the transpiled language to JS the lower the impedance mismatch. Whatever language I write in, I may need to dive into debugging it as JS. Whatever language I write in, I likely have to use/adapt JS code samples from libraries and dependencies written in JS. More often than not that's a paper-cut; Haskell often has a huge impedance mismatch from x86 assembly but you can still debug it pretty well and FFI into assembly libraries with varying levels of ease based on library documentation.

But especially when you are working with other developers and a mixture of skill levels, impedance mismatches are useful paper-cuts to try to minimize.

That argument can be used to justify using assembly because you would sometimes have to drop down to assembly level when debugging native code.

I mentioned that. There are still programmers that do justify using assembly that way, and many historically. There are many programmers that just C/C++ over C#/Java/et al for very similar reasons.

I agree, it's not a strong argument, but it is an applicable argument to the discussion. There are trade-offs in every decision, and this is a consideration in those trade-offs.

These are good points. In my experience other platforms have the same tooling as JS has been getting over the past few years.

Look at Android for comparison:

- ReactJS: React Native, also any overall architecture like MVVM or MVP.

- ES2015/ES7: JDK 7/8

- Transpilers: Kotlin

- Minifiers/packers: proguard and others

- Gulp and co.: gradle

- Other libraries: android has a number of libraries to handle network requests, image loading, animations etc.

What I'm trying to put across is that JS has transitioned from a language that would show and hide elements of the DOM to a serious language used everywhere.

A good developer takes the time to understand the tooling around their platform and language.

There's a very good point here - JS is being used for way more than just showing and hiding DOM elements anymore. It is becoming a full application language, and to that end requires the features that full application development requires, not just on frontend, but in Node as well.

I have just inherited a project written in 100% ES5 and I had forgotten how obscenely verbose JS <= ES5 really was. Every extra line of code is a liability and will slow down app development. Language features support precision and conciseness and that in turn promotes enterprise-level development. A toy language (old JS) can never support a full in-browser application.

Well, to be fair, the difference is that with Python (or basically any other langauge) a change in a library may introduce a bug, but you won't have to throw away the library or even substantially change your code, save perhaps for some temporary work-around- and the bug will most probably be fixed eventually.

With javascript, as far as a I can tell, "change in a library" means you do indeed throw out the library and adopt a new one, then write your entire code again from scratch and repeat the cycle next year. To me, that sounds like madness and just makes me happy I don't have to deal with all this.

Well the problem lies in exaggerations of not-actually-using-javascript crowd. If the "cycle" would be five years, would it still be that bad? If the previous framework would be of the very first generation of actual workable JS frameworks, and new one brought major improvements, would it still be that bad?

Reality is very different from these discussions.

Some of that can be compared to trying to find innovative ways to eat the soup with a fork.

That's so true. There was an article on HN recently on how advanced desktop app development tools were many years ago; how quickly you could build a fully functional and quite advanced application. When you have that background, the current state of JS resembles C on DOS in the early nineties. With the exception that there are far, far more complications that could have been be abstracted so that developers can focus on developing the app and not the underlying stack.

Nostalgia is overrated.

You can with a 10 line HTML file and linking a few CDNs make an interactive application that can reach 1 billion people with no install and no buffer overflows. That's magic. The 90s wish they could do that.

I could write a CAD system in the 90s with 10 lines of C and linking a few libraries :)

Well, given sufficiently good libraries everything can be written in 10 lines of C.

It depends what you create. If it's business apps for your customers, they might not care about availability to 1 billion people, but will appreciate several other things (such as rapid modifications) that more often than not are not so trivial to implement in modern web apps.

Could you point us to said article? A few keywords to search, at least. Thanks.

I can't find it either, just spent over 30 minutes searching.

I'm sorry, now when I'm re-reading the said article, it looks like what I meant were some related comments: https://news.ycombinator.com/item?id=13650730 Several people complain that the voids space after Delphi[1] has never been filled in with something that would come close in the speed and easiness of development of desktop applications. [1] not that it doesn't exist, but it doesn't count anymore, mostly because of its exorbitant price

Reply on news.ycombinator.com