- cross-posted to:
- programming@programming.dev
- cross-posted to:
- programming@programming.dev
A masterful rant about the shit state of the web from a front-end dev perspective
There’s a disconcerting number of front-end developers out there who act like it wasn’t possible to generate HTML on a server prior to 2010. They talk about SSR only in the context of Node.js and seem to have no clue that people started working on this problem when season 5 of Seinfeld was on air2.
Server-side rendering was not invented with Node. What Node brought to the table was the convenience of writing your shitty div soup in the very same language that was invented in 10 days for the sole purpose of pissing off Java devs everywhere.
Server-side rendering means it’s rendered on the fucking server. You can do that with PHP, ASP, JSP, Ruby, Python, Perl, CGI, and hell, R. You can server-side render a page in Lua if you want.
it’d be very nice to have a progressively enhanced static frontend instead since there’s really nothing about any of this that should require JavaScript (and something like unpoly would give us react SPA style functionality strictly as an enhancement on top of plain HTML)
this might be a cool project for someone to pick up once work on Philthy gets going; most of the alternative Lemmy frontends still have an unnecessary JS framework dependency, or are lacking features for essentially no reason
The main reason companies use frontend frameworks is it’s easier to continue development through employee turnover. If your app was written in react or angular you just have to hire someone who knows how those work and they can get up to speed pretty quickly. Modularity also allows for code reuse. It increases maintainability. Labor isbtye major cost of software development, so making things easier and faster to develop and maintain is better from a business perspective than ensuring your app can run on a 15 year old iphone.
If you wanna go frameworkless, JS-less, or whatever on your personal projects then fine. If you insist on it in a professional team environment, you’re making everyone’s lives more difficult.
I’ve heard this exact same bullshit spun defending choosing golang too, and it’s just as bullshit there as it is here
and that’s not even touching on the aspect of this being based on the extremely toxic “oh yeah just burn them up and find the next one” mentality that has become far more prevalent in the world under the umbrella of zirp-funded bayfuckery gaining international traction
I beg you to go consider whether this is your actual position, or some shit you picked up from someone else. to consider what the effects of this stance are, not just today but in 5/10/15y+. it should be quite easy to see both how it helped us get into the collective pile of shit we now do have, as well as why it won’t ever be good
another thing to think about is how this was not invented by frontend frameworks. We did it fine pre-SPAs and pre-preprocessors. It was part of the architecture and strategy. The hard work that allowed us to essentially reskin entire, very complex, projects every couple of years
i’ll put myself out there - here’s a receipt from 06~07 https://web.archive.org/web/20070512035940cs_/http://www.toyota.com.au/toyota/main/css/elements.css
we were a team of 5 devs including me. We weren’t tribed off into separate areas of concern, we all knew the whole project back to front, and (maybe not the most clever move) managed without version control by always being aware which part we were working on. Cos, ya know, communication is easy when you are 5 people sitting in a group.
Don’t give me shit about the complexity of the UI in modern apps either. We were dealing with a huge collection of brochure style pages that had plenty of variations. We kept all that css under 500kb. We could achieve the bland flatness of modern uis under 100kb easily. No fucking doubt.
I built sites as large and larger than Toyota with a team of 4-5 devs. Even with some of them being very junior devs, we still managed to keep the CSS under 500kb.
Lots of front-end devs don’t understand the difference between complicated and complex.
Complicated means it’s difficult to do and hard to understand. Complex means it’s got many parts.
All it takes is a little bit of maturity and planning, and most any modern UI could be achieved in under 100kb of CSS. You put on your big kid pants and think about what you’re going to write before you write it.
CSS isn’t some deep, level-10 arcane magic. You literally gotta roll an occasional persuasion check against a browser.
Thanks for sharing the article, BTW
Thanks, Frank!
Killer line. THIS is what DESIGN is. The lost art of knowing what you want to do and deliberating over how to do it. The tech industry reversed it and now everyone is figma-ing about like children lost in the forest.
I’m a backend developer and jesus fucking christ, 500KB? That’s around the weight of code in an actual programming language for a mid-sized project, and front-end needs that much just for CSS? And CSS isn’t even that verbose.
The whole Rust compiler is like 10MB and that’s a huge codebase, including all the documentation and shit.
One more reason front-end work never clicked with me I guess
Yeah. That’s one of the many reasons I wrote a damned rant about how fed up I am with front-end. It’s insane to think web sites telling you about cars or coffee need 1MB+ to do so.
But it happens because front-end tooling makes it so fucking easy to write complicated, bloated code.
Most of my time and energy these days is deleting shit. I delete more than I write. But my proudest achievement was like 7 years ago when I was asked to add a Japanese font.
Took me 6 weeks. I deleted like 15% of the CSS in the process.
Why did it take so long, and why did it require deleting 15% of the code?
Because the dorks on the project used Sass for literally every line. So I had like 400+ instances of someone applying a font-family spread across 80 Sass files. Not a single.fucking.HTML.element received a style. Not. One.
There was no font-family declared anywhere that it could be inherited down.
So over three sprints I had to move the font-declarations into mixins, then onto raw elements, then actually delete the mixins on the classes. Eventually I got it down to like 18 instances of declaring the font family.
and then I wrote another 18 instances where I could apply the Japanese font.
Adding a Japanese font reduced the size of the CSS by 15%.
front-end work doesn’t click with most front-end devs, either.
I’ve been forced to do react for years and I still don’t like or understand it. Most times plain JavaScript is easier and quicker to write and quite maintainable if people can resist the urge to take the piss with nested anonymous functions.
I honestly can’t get my head around the idea that people can hit the ground running with react, but can’t write unabstracted JavaScript. It’s like a MotoGP rider not being able to ride a push bike.
I have, in a previous age, unfortunately been the first one to suggest react at work. it’s declarative! the mental model makes sense! it’s kind of like functional programming! why, Facebook is surprisingly good at CS, maybe we should look at graphql too since that seems like such a good fit for react
this venerable house, opulent and imperial, is a festering abomination. as soon as you run into any performance issues or edge cases with react (or far more quickly with graphql, where the edge cases include shit like authentication and API versioning), you’re going to start burning out developers doing the most counterintuitive bullshit ever invented to torture a development team. and react is structured such that performance issues will accumulate in web apps; it’s just a matter of time (and not even that much time) before they do.
that’s why the advice now is to dodge performance issues with server-side rendering, almost like your site should have been fucking static html in the first place, except SSR won’t fire up without a gigantic bundle of JavaScript affixed to it, and in general it’s another source of bugs and weird performance regressions that you now have to debug in two places
and for what? react’s DX is better than HTML and CSS until you hit a wall, then it’s much worse. you can get a fairly react-like set of functionality out of plain HTML with Web Components… except Web Components requires fucking JavaScript for no reason but to not threaten existing frontend frameworks (see our sister community FreeAssembly soon for the gigantic rant and JavaScript library I’m writing about this shitty situation)
we used to strive for minimum possible front-end payload, and it was an embarrassment to do anything with JS that wasn’t backed up by a non-js default. Will never forget how suddenly React removed all those things from front-end team meetings.
They were solid industry-wide concerns that just… disappeared
Remember when our industry cared about loading times?