56
A Rant about Front-end Development
(blog.frankmtaylor.com)
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
this post was submitted on 29 Jun 2024
56 points (100.0% liked)
TechTakes
1437 readers
133 users here now
Big brain tech dude got yet another clueless take over at HackerNews etc? Here's the place to vent. Orange site, VC foolishness, all welcome.
This is not debate club. Unless it’s amusing debate.
For actually-good tech, you want our NotAwfulTech community
founded 1 year ago
MODERATORS
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
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?
facebook used to lie about react being faster than native on first load and navigation, in spite of that being impossible by both lived experience and as measured by benchmarks. supposedly templating is just too heavyweight for servers to handle at the mythical Amazon scale literally nobody reaches except Amazon but every shitty manager needs us to be ready for
and now that react can do server-side rendering I guess we’re doing templating again, but in node and much less efficient and with extremely unclear semantics around when it switches to client rendering, and also weird bugs when things render differently under SSR
also it’s still measurably much slower than old school server templating
Ah yes, Amazon, the company with literally the shittiest front-end of all in existence. AWS is downright unusable outside of the CLI, but hey, at least they scale??
you know, for as much poison’s been poured into my ear about how everything must be Amazon scale, there’s no way in fuck they use react for their storefront or AWS, is there? I think the only reason react is considered an Amazon-scale frontend (besides Facebook, which also has a shitty UI, though not as bad as Amazon, and notoriously uses PHP for everything) is how hard they push it as part of AWS Amplify, a toolchain they say will help you reach their scale (but from experience: it absolutely will not, it’s just a set of technologies that increase your AWS bill and perform like shit, which is why Amazon doesn’t use it for anything of value themselves)
the only case I can immediately think of of a very major site going from server rendering to react is GitHub (which used to use Ruby on Rails and Erlang, apparently) and it’s been an unmitigated disaster — none of the new features that supposedly require react are good, the performance fucking sucks now, and the thing keeps breaking (I get weird renders with broken styling every few refreshes and apparently I’m not the only one). the fucking thing even hijacks the keyboard shortcuts I use and has become an accessibility nightmare, all in the name of pointlessly turning it into a react SPA and vscode wannabe.
Hey, GitHub might be shitty in the browser, but did you consider that it's also shitty as an Android app?
I remember seeing an argument on reddit between a css dev that understood the depth of the responsive design philosophy and a dismissive Reacter that shut them down by calling them an old "list-aparter"
i keep finding so much that's done with JS that there's perfectly good CSS for
it turns out browser rendering is fucking fast in 2024
fucking exactly! I’ve been doing a lot of CSS-only work for the sneer archive rewrite, and it’s shocking how fast everything renders without JS, and how much functionality you can retain with a good enough CSS framework and careful markup
I’m also working on a JavaScript library and associated rant named fuckery because it turns out you can’t use Web Components without some utterly unnecessary JavaScript, because the W3C decided to do a fuckery
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.
as a developer my favourite thing about react componentisation is how it makes me and my team more readily replaceable
fuck, this quip’s better than my seeing-red rant
I love when someone argues against something that is arguing against everything they use in their argument
I checked and they do the “well maybe it’s ok in your personal projects” bit a lot, which is very funny because the code for my personal projects usually isn’t garbage
That's such a weird notion. My personal projects are the cutest, most groomed pieces of code I write, cause I do it out of my own volition. The code at work? Just any shite that passes the review so that I don't have to look at that codebase or think about it lest eldritch worms consume my sanity.
thank fuck neither myself nor this instance have employees, turnover, or shitty little project managers that get heartburn when the stack’s HTML5, CSS, and a non-shitty templating language instead of HTML5, react/angular/svelte/whichever frontend framework the market decided is in demand this quarter, a CSS in JS library, an ORM, webpack, and whichever npm clone tweaks your nipples the most
and you’d better hope you chose “right” on all of those pieces of the stack, cause you’re infantilizing your devs so much you think it’s impossible for them to learn a new frontend framework, or how to do modularity or maintainability in a basic fucking backend templating language. do they also have to ask your permission to take a piss?
but why are you posting here? it’s almost Monday and you’ve got an hour-long, unproductive standup to preside over
praise the circumstances that enable the scourge of b2b saas products imposed on employees at the collaboration factory
if we keep this up, the CEO might positively mention the name of our project briefly during an all-hands, then two weeks later vastly reduce our headcount because the good job we’ve done proves we don’t need to waste money on all these developers
I remember when we used to write our name in our css files because we wanted to, not because our ssh key enforced it for auditability
I know this sounds like old man shit, but I'll die on this hill. It's a significant fundamental attitude shift
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)
Magnificent.
As depicted in the paintings of Heironymous Bodge.
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.
Maintainability my ass. Most "reusable" components I inherit on a react project are worse than worthless, they're counterproductive.