Hacker News Reader

Hacker News

Comment by mg | Hacker News Reader

Full thread

I wonder if I will ever use a JavaScript framework.

My projects are usually either in the form of "User enters data, sends it to the server, server sends a new page back"

Example: https://www.gnoosic.com/faves

Or in the form of "User changes parameters on screen, Javascript updates the screen accordingly".

Example: https://www.productchart.com/laptops

In both cases, I don't see the benefit I would get from a framework.

If you have built a project that is public and benefits from using a framework, I would love to see it!

View on HN

If you aren't using a public framework and you use any JavaScript in your app then you are creating your own framework. Nothing wrong with that.

The real value of public frameworks comes into play when you have more than one generation of developers working on your project.

It's much easier to onboard a new developer onto a react app than it is to get them to understand the weird bespoke nuances of the conventions you came up with while hacking on your site at 3am 5 years ago.

Your last paragraph is not a logical conclusion to your first two paragraph statements.

> If you aren't using a public framework and you use any JavaScript in your app then you are creating your own framework.

Sorry but a “framework” used by one project (especially not broken out into a separate subproject) is not a framework. That’s just infra code.

I think your point still stands that the value is developers understanding a framework can more quickly get to work—I’m just triggered by your exceedingly liberal usage of the word “framework”.

> Sorry but a “framework” used by one project (especially not broken out into a separate subproject) is not a framework. That’s just infra code.

Infra code that over time, as your app grows, takes on more and more of the features of a framework

This might hold true when the application in question does a lot of stuff covered by a framework. For example if it does routing, then it is nice to have common ground on how routing is done. If the application mostly does custom stuff, not covered by the framework, then the framework might harm more due to the added complexity.

That is why I proposed a discussion based on real life examples.

The other thing is that frameworks change. You picked React for your example. A currently popular framework. Chances are, the next developer will not come on in 5 years, but in 15 years. And does not know anything about React. Or about React as it is today. Then they might have a harder time understanding React than to understand a well written, minimalistic piece of Javascript.

> Chances are, the next developer will not come on in 5 years, but in 15 years.

That doesn't seem likely to me at all. More likely in the next 6 months than in 15 years. Why are you planning for 15 years out? The company you are working for might not be around in 15 years. Code might not even be recognizable in 15 years. All of our programs might be written by AI in 15 years.

Optimizing for the short term makes way more sense to me. 2 or 3 years out, not 15.

Human labor is expensive. Companies aren’t going to throw away a perfectly good asset they poured tens of millions in, unless they have to. I don’t see why a well-written, time-tested piece of software shouldn’t still be around 15 years down the road.

The initial comment was not about software quality but about "the popular framework now might not be popular in the future"

I don't see any contradiction between creating well written, time tested software and also choosing currently popular frameworks to work in.

If your React app is well written and stable and well tested, I see no reason it shouldn't be available in 15 years. And if you can't find React Developers to hire in 15 years to work on it, that's tough. Pay to have some trained.

It is a similar situation to COBOL is today.

Except that unlike React, COBOL doesn’t bring along 10000 dependencies written by 5000 different authors, randomly abandoned a few months later but teeming with unpatched vulnerabilities.

Sure, somehow I’m going to find React developers 15 years from now to maintain my line-of-business React app but how secure will it be?

One could argue that I can rewrite the framework-dependent parts. But rewrites cost money, too.

I love new stuff but I still think that, depending on the app, it may make good economic sense to plan ahead for it to be useful after a decade or two.

I'm not sure where you are getting 10,000 dependencies from.

https://www.npmjs.com/package/react

Maybe I'm missing something but npm suggests it's only one dependency, which also only had one dependency.

I get that create_react_app pulls in a ton of dependencies, but React itself is not a culprit of dependency hell.

Good point. You’re absolutely right.

> Or in the form of "User changes parameters on screen, Javascript updates the screen accordingly".

Once this gets a bit more complicated you have a bunch of state to keep track of and you have to make sure to re-render only the parts that have changed, and deref that parts that should be GCd. I'll happily take on a reactive UI framework dep if I can avoid writing my own diff pipeline.

Not endorsing react by any means though.

I also started out without frameworks and still prefer to work that way. However, I have found them to come in handy when solo-ing projects that require:

- transpiling JS back into the stone age for compatibility

- accessibility

- seamless localisation

I have found I just can't keep track of all of these things as well as React does.

At the same time, I think React is a huge, dependency-laden mess, very difficult to learn (because it's changed its mind on how to do things about 100x, so no responses to issues on SO apply to your version of React, Redux, React Router, etc), slow as hell and probably overkill for most projects...

Working with other people and agreeing on how to do things is hard. The backend and frontend roles crystalized with a natural boundary between them being HTTP, one side doing HTML stuff, and the other defining URL endpoints (everything else is basically done by the ORM).

I never use frameworks either. In fact the last site I published didn't have any Javascript at all (it's just a static site so shouldn't need it yet that doesn't seem to stop most other sites from abusive amounts of JS).

Sure, probably works, just don't complain about age discrimination when you bring up nonconsensus things that have been superseded, and the team isnt fond of that system design

just wanted to say gnoosic is pretty great if I'm careful about choosing bands as an outer perimeter of a type of music I want to listen more to.

gnoosic.com/faves is cool but the Allmusic.com's related artists section is the ultimate place to discover new bands.

Reply on news.ycombinator.com