[machinelabo]

This is the epitome of a cliche JS project:

Emojis in docs? Check

Piles of abstraction layers? Check

No tests? Check

Already at version 2, but has breaking changes? Check

Stacking up on yet-to-be-proven-framework (tailwind)? Check

We just need a BLM support message, donation link in the npm install stdout and some more JS shenanigans like support for yarn, gulp, webpack, npx, degit, and the next gen packaging tool soon™ to be released.

Sorry if this is offensive, but this ecosystem offends me and my sanity. This project is going to last 6 months and they'll move onto something else. I think I am finally ready to jump on React since it has been a few years and appears to be stable.

Edit: My ramblings are unfair. There are tests.

[neya]

I am not attacking any individual developer or project here, but in recent times, there's been a trend I've been observing and what you're pointing out. I call it the Hipster Driven Development phenomena* and it's just getting worse with every year. People just start by discarding established best practices in the name of being unique and cool and then wind up to the inevitable conclusion - the established best practices were there for a reason.

There's been many such trends to name a few - discarding years worth of stable codebases to try out some shiny new framework, throw away proven technologies to trade for some perceived developer convenience (Eg. PostgreSQL to MongoDB), labeling known terminologies to sound cool (Hacker vs Growth 'Hacker'). All these hipster moves come with a huge cost - they cost morale, time, money, teams, relationships. It just feels so unnatural to accept this as the norm.

In comparison, I just feel like some old dude from the 1950s rambling though these things have started to happen only the last half of the decade or so. Where did we go wrong?

Edited for OT

[efdee]

I was with you for the first two paragraphs, but I feel you invalidate everything you said by pretending separation of technologies and separation of concerns are equivalent. That reeks of dogmatism more than anything else.

[odiroot]

> I call it the Hipster Driven Development phenomena* and it's just getting worse with every year.

My impression is, that began, when highly political people started to join the JS community (and Web engineering in general).

[dotancohen]

Its funny that you mention that. After Eich was forced to quit, I've been noticing that everything related to client-side development is highly influenced by liberal mindsets, many of which are also political. I'm also a minority, but I don't want (even if I need) anybody advocating for me.

Of course, both you and me are going to be downvoted because these types oppose (downvote, "cancel") anyone who mentions their presence.

[sod]

Thanks for hipster driven development then. Frontend development - or really any software environment - would be very sad if everyone was scared to try something new or mix exiting ideas when a field seems to be already saturated by an established framework.

[olieidel]

Agree, an interesting observation! I wonder: Junior developers nowadays who learn JS and React as their first, go-to stack for web development, they haven't ever experienced the raw productivity of "just write php and upload it via ftp" or "just write it in rails/django and render stuff server-side".

They come to expect that all development is as slow as theirs, including frontend <--> backend apis (graphql), huge js complexity (webpack), breaking changes (react-router), etc.

I see lots of companies nowadays wanting to hire "React devs". It seems we're collectively forgetting how fast you can ship stuff if you just use proven technology. Makes me wonder where we're headed.

Maybe it'll be a competitive advantage to use old (= proven) technology in the future?

[cutler]

You gotta hand it to the Rails team who came up Stimulus [Reflex] as an antidote to all the SPA abuse.

[sbergot]

Sorry to burst your bubble but Tailwind is anything but "yet-to-be-proven-framework". It is a very solid and powerful abstraction that has existed for 3 years.

Speaking of piles of abstractions, please enlighten me. What is not a pile of abstraction today? Even x86 instructions are implemented with microcode.

Speaking of tests, anyone with some frontend experience will recognize that jest is a testing library.

I would say that your comment is the epitom of a HN frontend thread ie mean criticism by someone with 0 experience in the field.

[machinelabo]

Oh please.

Tailwind’s entire premise stands on these pillars which are shaky at best:

- Developers don’t want to write css so they add insane amounts of nuisance in templates. Impossible to read the design language.

- The original author of Tailwind complains about naming classes in the css but Tailwind ironically names things that are “tx—-bb-4-2” or whatever the hell that means. No thanks, I’ll write “customer-support-card” or something useful. Maybe even more generic - “card”.

- Tailwind’s front page is everything that’s wrong with this. Moving around components with “mr-4” tags and shifting it up. Left. Right. Making it larger. That means that there is no standardized design language and you’ll create design that will have no cohesive unity. One component will have “mr-4” margin and another “mr-3”.

We have the technology. Sass can create compositional components with reusability and logical hierarchy. There is no need for this Tailwind nonsense.

Tailwind is a controversial project, not mature. It’s proponents think that others just don’t get it. No, we do. We’ve used it and it makes it very fast to get something working but that’s more telling about one’s inability to take the time to think about layout/design and writing proper sass.

[tvphan]

There’s a new wave of thought in design called “design tokens” where you have a set of predefined values that are considered valid as part of the design, e.g. a website can only have spacing of 4px, 8px, 16px, etc.

One component having mr-4 and another having mr-3 is chill if that was the designer’s intention. If not, then it should be picked up during design review and easily fixed (since you end up really only having to pick from like a few values).

[machinelabo]

Agreed on the tokenization but you can do this via variables in Sass.

And put constraints as well! Comment them! Write for loops to create a full spectrum of tokens!

Then use those tokens to create a card class. Extend it for customer-support-card class!!!

What’s missing here?

[tvphan]

Mate, you gotta chill. You seem very triggered by Tailwind! You don't have to use it if you don't want...

Some comments:

- It's very easy to integrate Sass with Tailwind. You can import the Tailwind classes just like you would Bootstrap: https://tailwindcss.com/docs/using-with-preprocessors#using-...

- You are also allowed to write semantic classes as well: https://tailwindcss.com/docs/extracting-components

I would argue the main point would be you can rapidly build interfaces... like really fast. I worked on a proposal for a client using Tailwind a month ago and we smashed out a whole website in a week. We ended up "extracting components" like buttons and cards and stuff that we found was being reused a lot (kinda like Sass).

As opposed to Styled-Components, the Tailwind classes are very valuable for designers who want to prototype live in the browser.

Also, you were complaining before about design consistency but I would argue that Tailwind allows for very high design consistency. I work with armies of outsourced developers of varying skillsets and I honestly don't trust a lot of them to write clean CSS at all. With Tailwind and Tailwind design tokens, the army of outsource developers are encouraged to work with the design tokens (as opposed to yolo'ing 3px !important everywhere).

[machinelabo]

I am arguing that the speed is a delusion and the reason why people love it so much. It seems very attractive to design things super fast since you have instant control in the same context as HTML, but I question the maintainability, readability and most importantly - design unity that breaks apart in this way of developing things. Especially, if there already exists a great solution - Sass.

I am not sold.

[Vinnl]

I think you might be misinterpreting Tailwind's "pillars":

> - Developers don’t want to write css so they add insane amounts of nuisance in templates. Impossible to read the design language.

It's not so much that developers don't want to write CSS, but that it's largely impossible to delete lines of CSS.

> - The original author of Tailwind complains about naming classes in the css but Tailwind ironically names things that are “tx—-bb-4-2” or whatever the hell that means. No thanks, I’ll write “customer-support-card” or something useful. Maybe even more generic - “card”.

So those are different complaints about naming. Your complain is that names are hard to understand. Tailwind's complaint is that trying to define "generic" (i.e. reusable) classes is practically impossible. Is "card" the right abstraction? Is "customer-support-card"?

> - Tailwind’s front page is everything that’s wrong with this. Moving around components with “mr-4” tags and shifting it up. Left. Right. Making it larger. That means that there is no standardized design language and you’ll create design that will have no cohesive unity. One component will have “mr-4” margin and another “mr-3”.

The idea is that all those margins are part of a set of "constraints" that all compose well together to form a cohesive unity. More about this is written here: https://styled-system.com/guides/why-powers-of-two

> We have the technology. Sass can create compositional components with reusability and logical hierarchy.

This might be the main point: Tailwind means you're not defining reusable styles. All the styles are specific to the component, and the component is the thing you reuse - which we were defining already. So it actually gets rid of yet another tool to create compositional components with reusability and logical hierarchy.

(It's less fast for me to get to something working than e.g. using Bulma is, but I'm expecting that it will rid me of a number of the pains I run to when I'm hitting the limits of its built-in component styles.)

[machinelabo]

I get the compositional aspects of it. I've used similar utility css classes. They're pretty nifty because its instant gratification. No need to switch context and go to Sass.

However, for production, I always come back and refactor everything into a beautiful Sass file that explains with comments what all the stuff is doing.

Now, the designer can come in and change something in the sass without having to worry about the rest of the app. They can easily debug too 'customer-support-class' straight from chrome browser.

I feel like Tailwind is a prototyping tool that's then shoehorned into production environment and in large applications where the design language (despite of tokens and standards) breaks down.

[Vinnl]

I definitely see how that workflow works better in your setup, but... That doesn't apply to everyone? Not everyone works somewhere where they have an independent designer that can and will change the actual SASS, whereas others might even have designers that a perfectly comfortable rummaging around in e.g. JSX or Vue components. The indirection of a separate SASS file might not be worth it in those cases.

All of which is to say: just because it's not the best solution in your use case, that doesn't mean it doesn't solve actual problems for others in production.

[ficklepickle]

My brother. Oh god I thought I was alone. I just kinda got CSS and all these easier, softer methods just get in my way.

I love me some sass scoped to the component. You don't need classes and the structure matches the markup, making it trivial to see what styles apply.

[deergomoo]

> the structure matches the markup

So you’re just writing your HTML twice, essentially. You can’t change your HTML without changing your CSS, so at that point, what’s the point in writing the CSS? May as well just add the styles right on the HTML.

They’re both fundamentally just part of how it looks, and I don’t see a lot of benefit in trying to keep that separate.

[earthboundkid]

Everyone hates Tailwind before they use it. None of your criticisms are new. Actually using Tailwind makes you realize that creating CSS classes for your components is basically a waste, since you only ever use them in the component itself.

- Unless the same class can be used in multiple HTML templates, there’s no point in giving the class a name.

- Only very general CSS can be used in multiple templates because anything specific will have the brittle base class problem.

PS mr-3 is .75rem and mr-4 is 1rem, so you can see with your eyeballs if you’re using the wrong one.

[type0]

Tailwind is just like shorthand, it might not be useful to you but it is useful for some fast prototyping and iteration.

[8lall0]

My experience teaches me that everything born for "fast prototyping" is going to temporary stay for years to come. So tailwind, like bootstrap, just moves the problem from css to html class attribute.

[machinelabo]

Agreed and that’s it’s only selling point.

Fast is not always better. It makes spaghetti code with shit design that lacks an architect behind it.

There are many threads on HN about Tailwind and it’s always very controversial.

I wouldn’t call this thing mature when there is so much controversy.

[bryanrasmussen]

>- Developers don’t want to write css so they add insane amounts of nuisance in templates. Impossible to read the design language.

I don't think this premise is very shaky, unfortunately, if it was shaky I doubt we would be seeing all these solutions coming around trying to help developers avoid writing css, solutions which seem to have uptake in the market.

[sbergot]

Please have a look at https://adamwathan.me/css-utility-classes-and-separation-of-...

With tailwind you can create a very small configuration files where you pick a small set of available options for spacing and colors. This is actually a big help to implement a consistent design system. Many people see tailwindcss as a "design system builder".

Please show me the doc for "tx—-bb-4-2". Does it exist or are you just inventing it for some reason? Is "mr-2" unclear?

Sass has nothing to do with tailwind. It would be like comparing typescript and react.

Tailwind can be controversial and mature at the same time. Both things are not exclusive. PHP is controversial and mature.

I don't know about other people but in your particular case I would say that you are showing sign of a knee-jerk reactions and bad faith justifications.

[bitten]

I see tests in the repo :p

[machinelabo]

Yikes, my bad. I didn't see a test folder. I am just learning JS coming from an enterprise Java background. We've got our own pile of shenanigans as well, probably just as bad as the JS world, don't worry!

[cutler]

Don't worry, you can have your public static AbstractProxyFactoryBuilder with Typescript to make you feel at home.

[machinelabo]

Ok, I am done with this. Anyone wanna start a tea shop in a small town in Asia? I am down. We can bootstrap it, live a simpler life and mingle with the locals. May be even create a little ordering system, postgres v9 in the backend, running on a dingy moth balled Dell PC under the cash register and use jQuery frontend, of course limited to 1024x768 resolution monitor from Compaq with a broken power button.

[idkwhoiam]

Small towns in Asia have millions in population :D

[degurechaff]

yup. my town in java have 1.6 millions people :)

[Brajeshwar]

Can I, please, join too. I will do all the interfaces in XUL and will optimize the heck out of it to work on IE.

[jamil7]

I would unironically love to do this.

[cutler]

Sign me up. Procedural PHP4 preferred or Perl/CGI.pm.

[machinelabo]

I approve of this stack. No version control because we would just do this in one shot and the never add any features. The time saved will be spent with friends at the tea shop. An occasional bug fix for a customer that ordered over 32768 rupees of tea for a wedding party.

[sbergot]

So you are just learning JS but felt legitimate enough to post an unfair rant?

[jamil7]

I picked up a React Native project for next year and am sort of terrified of what I might find jumping back into the world of JS after some time away from it.

[elygre]

What do you mean by “BLM support message”?

[girvo]

A number of JavaScript projects have put support messages for the Black Lives Matter movement into the top of their README or home pages. I think it’s an excellent thing, personally.

[saturn_vk]

I agree with the sentiment that BLM, but I don't see how a message/banner/whatever helps in the matter.

[ficklepickle]

Autist here. I see this with so much human behaviour. Often the purpose seems not to be actual measurable progress but rather social signalling. I suppose it serves a social purpose, perhaps it is a precursor to material change. I feel it is important to remember that humans are tribal apes. If we deny our nature, it is much harder to overcome.

[neilsense]

It's a nonsense thing but, hey, it's easier than doing something of actual value.

[tannhaeuser]

"Black lives matter" I guess.