I was going to title this "Stop Using Modern JavaScript" but decided to call it "Stop Using React". I can't agree with you more, the problem is SPAs in general. I would like someone to explain to me why SPAs need to exist actually. (Preact is an okay framework for speed, but I would still prefer writing HTML...)
I think SPAs are mostly useful for companies who need to organize their code...right? I don't get it.
HTML already has modularity and templating built in, if copy and pasting is what grinds people's gears.
Comment hidden by post author - thread only visible in this permalink
I'm a self-taught dev focused on websites and Python development.
My friends call me the "Data Genie".
When I get bored, I find tech to read about, write about and build things with.
SPAs fit into the Progressive Web App movement which is about bringing native experience to web. So websites feel like mobile apps that have instant load time.
It can also involve service workers to fetch data in the background and also cache it esp like a feed of posts or photos.
What I like about a SPA is it can be served as a static site on GH Pages or Netlify (if you use a CI build step). So no speed and hosting and security issues that go with having a Node Express app. Just fast static assets served.
I recently moved a project from templating with Mustache to templating with Vue and the move meant I didn't have to think about low level DOM operations to find values on form elements and push the results to an output section. I get to focus on the variables as they are bound to HTML elements.
I do agree with your sentiment. Don't make a web app into a desktop app and don't break if JS can't run.
JS was designed for a philosophy of progressive enhancements. I have a JS book I read which emphasises this at the start. Yet many sites are blank with a warning to enable JS. And sometimes just a blank page! The obsession with JS means sites are unusable if you have an older browser or the developer used only the newest syntax and didn't use babel to get output that works pre-ES2016/ES6
I like using Jekyll for templating in general to avoid JS on the page. Or to just add JS to add sort or filter on say a table.
The main reason I guess I like JS whether SPA or not SPA is that there is no page load when you submit a form or like a photo etc. There is white screen flicker.
I'm a self-taught dev focused on websites and Python development.
My friends call me the "Data Genie".
When I get bored, I find tech to read about, write about and build things with.
For interest, I added Vue to my jquery mustache site using frontend code only. So adding a script tag loading vue.js and then using a script tag on my HTML page.
This was light as there was no build step and the site was not a SPA yet, just HTML pages with JS added on top. And I could have using jekyll to add consistency so I have Vue load in the head of each page.
And I could have broken my JS into standalone JS files so I can run tests and linting against them.
The thing about the SPA style is that it means you manage your JS deps using a package file and you then import that in each of your scripts. You treat them like modules like in a server side app like an API or CLI. The problem with the browser approach is there aren't imports, just script tags loaded in a certain order.
Having said that there is a newer syntax where a script tag is imported as a module and you can use import x as * from 'jquery' in a JS file and then use it on the frontend! I am have just not tried it yet. But it means better server side feel of development with the frontend usage without making it a SPA.
Instead, we should concentrate on the unique web selling points: its reach, which, more or less by definition, encompasses all native platforms, URLs, which are fantastically useful and don’t work in a native environment, and its hassle-free quality.
The problem with the browser approach is there aren't imports,
Besides type="module" bundlers were introduced to enable "modular JavaScript" - not to support SPA's. If you are using React/Vue your choice (webpack) is essentially made for you - but for anything else I find rollup.js a lot saner especially as it is built around ES2015 modules (esbuild is faster).
I'm a self-taught dev focused on websites and Python development.
My friends call me the "Data Genie".
When I get bored, I find tech to read about, write about and build things with.
The part about not using Vue template on the frontend might make sense in certain cases like if you are making tables or you have SSR. But vue supports it and it means you can have a touch of vue to your site without rebuilding it as a SPA so it might still be worth it. Same goes for React and Preact especially. A SPA structure comes with safety but it also has a cost which is the point of the post.
What I meant by the module syntax is that I can use that to get the experience of writing JS tests and imports like in a SPA but without actually structuring as a SPA.
I have heard of Eleventy and Gatsby and others but haven't really tried them
But vue supports it and it means you can have a touch of vue to your site without rebuilding it as a SPA so it might still be worth it.
When using DOM HTML as a template you have to ship and execute runtime + compiler (93kB) rather than just the runtime (65kB) on the client.
When you use a build step to compile the template into a bundle you only need to ship the vue runtime to the client - this isn't the same as building an SPA. At the core of the original article is the notion that it should be a goal to ship and execute less JavaScript on the client.
I'm a self-taught dev focused on websites and Python development.
My friends call me the "Data Genie".
When I get bored, I find tech to read about, write about and build things with.
I'm a self-taught dev focused on websites and Python development.
My friends call me the "Data Genie".
When I get bored, I find tech to read about, write about and build things with.
I it was pointed out earlier in this thread but there is a major misconception that SPAs are PWAs. A SPA can technically be a PWA, but honestly there is no justification for SPAs in the context of a well written PWA.
The service worker can eliminate the network latency, which means you can ditch the fat, out of shape, slow JavaScript that are popular fast food frameworks. love2dev.com/blog/pwa-spa/
SPAs are actually useful. It is justified when you trade a longer loading time for faster interactions in the future. Think a webmail client, like GMail. Sure, you can load a lo-fi version and keep reloading on every click, but the experience of having email load super-fast and be able to answer them fast is just nice. You usually spend a couple of minutes browsing them and then the time consumed on initial load is saved.
Other areas where SPAs are good are live applications, such as betting ones. You always want to see the current stakes, so you need the data pusher from the server to update what you see in the frontend. No one would probably bet on a static site (unless those are bet for something static too, for example "which year a human will land on Mars").
And of course, SPAs are a must in web applications imitating interaction from regular desktop applications, such as Google Docs.
Pity they are also used in blogs (!), contact forms (!!) or just informative website about the small business.
My opinion is that React should be a last-resort option considering all the negatives, and even then, should be used with adaptive loading and every web performance optimization possible. Right now it seems like React is the default for every app and no one considers the negatives.
I also know that Facebook, who developed React, is aware of these negatives and serves a different version of Facebook to users on feature phones (a version of Facebook with less JavaScript to download) than the version they serve to users on iPhones. The developers of React know that React is flawed.
Complex apps that actually require JavaScript intensive SPAs usually have adaptive loading in place (Gmail has the basic HTML version for slow connections) but the average developer doesn't think about any of this.
Comment hidden by post author - thread only visible in this permalink
Tech Lead/Team Lead. Senior WebDev.
Intermediate Grade on Computer Systems-
High Grade on Web Application Development-
MBA (+Marketing+HHRR).
Studied a bit of law, economics and design
Location
Spain
Education
Higher Level Education Certificate on Web Application Development
It depends on many factors but you generalize a problem for everyone. We've different analytic tools so we know our customers devices, resolutions and so. Why I must take care about 2G connection guys when the slowest connection we get is with 3G or 20Mb ADSL and on a <5%?
You need to know your market before choosing technology, isn't it?
If my customers are using high speed internet connections I MUST NOT care about those who are with 2G, because you put extra-efforts and engineering for it to run as nice as possible, even building a different version (check the cost of that in time and money) instead on giving the best experience to YOUR customers.
I see all that post and most comments some kinda romantic programming where things are idealized like "welcome all 2G users, welcome 56Kb modem users, let's push accessibility" but on the real world companies do what their customers need and no more for obvious reasons. And if you are a programmer you either will work on a company or being the boss of one so you need to take this in mind (the third option is being so bad you've no job).
If you already thank on that, you probably are on a country with very bad internet speeds then ok, change the title and say "Stop using react if your business market is on [country_name]" and will be zero misunderstanding.
This sounds like you really believed that most of decisions about using technology X are preceded by careful market evaluation, not just jumping on a hip technology bandwagon. I feel that software development is plagued by this kind of thinking:
Most of my users are on a fast internet, so I don't care about the bundle size
Most of my users are not blind, so I don't care about accessibility
Most of my users have names longer than three characters so it's okay if I put the validation like that in my registration form
The thing is, if you think about it from day one, you probably won't spend huge resources on adjusting the project for people who are not a majority of your market. But if you did not think about it, sure, you have to spend time on coming up in post-factum justifications like that ;)
Of course, there are some cases when using SPA is fully justified (see my previous comments) - but let's be honest, in most cases it's not.
Tech Lead/Team Lead. Senior WebDev.
Intermediate Grade on Computer Systems-
High Grade on Web Application Development-
MBA (+Marketing+HHRR).
Studied a bit of law, economics and design
Location
Spain
Education
Higher Level Education Certificate on Web Application Development
I must be biased about business side thinking but that's what you're taught at college, doing before thinking makes you redo several times so it's inefficient 🤷
When you open up a new project to a market there must be a previous analisis, not doing that and complain about the results is something dumb, don't you think?
SPAs fit well on tones of applications while does not fit on tones of others, that's right, the same for microservices architecture vs services or vs monolithic.
That's just the point, there's a big nonsense on recommend to stop using something "as is", your needs will depend on what do you need to achieve (requirements), knowledge about your market, seek the pros and cons for each option and decide which one fits better for your product, that's all
Well put. Some people keep saying "stop using Laravel" but if you don't have crazy concurrent requests, Laravel can solve almost all of the problems you would typically need & can get a lot done using this framework in a limited amount of time. So the context is important. 2G,3G,4G these are relative to the business of the website is catering. Remember we had to support that shitty IE ? Now we don't because the market cap is insignificant for the older versions of IE. There's always a tradeoff according to your customer base.
Some comments have been hidden by the post's author - find out more
For further actions, you may consider blocking this person and/or reporting abuse
We're a place where coders share, stay up-to-date and grow their careers.
I was going to title this "Stop Using Modern JavaScript" but decided to call it "Stop Using React". I can't agree with you more, the problem is SPAs in general. I would like someone to explain to me why SPAs need to exist actually. (Preact is an okay framework for speed, but I would still prefer writing HTML...)
I think SPAs are mostly useful for companies who need to organize their code...right? I don't get it.
HTML already has modularity and templating built in, if copy and pasting is what grinds people's gears.
This confuses me greatly.
Are you referring to
<template>element?SPAs fit into the Progressive Web App movement which is about bringing native experience to web. So websites feel like mobile apps that have instant load time.
It can also involve service workers to fetch data in the background and also cache it esp like a feed of posts or photos.
What I like about a SPA is it can be served as a static site on GH Pages or Netlify (if you use a CI build step). So no speed and hosting and security issues that go with having a Node Express app. Just fast static assets served.
I recently moved a project from templating with Mustache to templating with Vue and the move meant I didn't have to think about low level DOM operations to find values on form elements and push the results to an output section. I get to focus on the variables as they are bound to HTML elements.
I do agree with your sentiment. Don't make a web app into a desktop app and don't break if JS can't run.
JS was designed for a philosophy of progressive enhancements. I have a JS book I read which emphasises this at the start. Yet many sites are blank with a warning to enable JS. And sometimes just a blank page! The obsession with JS means sites are unusable if you have an older browser or the developer used only the newest syntax and didn't use babel to get output that works pre-ES2016/ES6
I like using Jekyll for templating in general to avoid JS on the page. Or to just add JS to add sort or filter on say a table.
The main reason I guess I like JS whether SPA or not SPA is that there is no page load when you submit a form or like a photo etc. There is white screen flicker.
Progressive Web Apps are great!
For interest, I added Vue to my jquery mustache site using frontend code only. So adding a script tag loading vue.js and then using a script tag on my HTML page.
This was light as there was no build step and the site was not a SPA yet, just HTML pages with JS added on top. And I could have using jekyll to add consistency so I have Vue load in the head of each page.
And I could have broken my JS into standalone JS files so I can run tests and linting against them.
The thing about the SPA style is that it means you manage your JS deps using a package file and you then import that in each of your scripts. You treat them like modules like in a server side app like an API or CLI. The problem with the browser approach is there aren't imports, just script tags loaded in a certain order.
Having said that there is a newer syntax where a script tag is imported as a module and you can use
import x as * from 'jquery'in a JS file and then use it on the frontend! I am have just not tried it yet. But it means better server side feel of development with the frontend usage without making it a SPA.That is a misconception that dates back to the App Shell Model.
The core technology for PWAs is ServiceWorker using the Cache to retain the essential parts of an (possibly MPA) web site for offline usage.
Developing Progressive Web Apps (PWAs) Course.
"bringing native experience to the web" is a fools errand.
Web vs. native: let’s concede defeat (2015)
Why Progressive Web Applications Are Not Single Page Applications
FYI: Eleventy was inspired by Jekyll but uses JS instead of Ruby.
That approach isn't recommended:
Why You Should Avoid Vue.js DOM Templates
Besides type="module" bundlers were introduced to enable "modular JavaScript" - not to support SPA's. If you are using React/Vue your choice (webpack) is essentially made for you - but for anything else I find rollup.js a lot saner especially as it is built around ES2015 modules (esbuild is faster).
I don't really understand. The website just tells good solutions, but not really against it.
Thanks for the links.
The part about not using Vue template on the frontend might make sense in certain cases like if you are making tables or you have SSR. But vue supports it and it means you can have a touch of vue to your site without rebuilding it as a SPA so it might still be worth it. Same goes for React and Preact especially. A SPA structure comes with safety but it also has a cost which is the point of the post.
What I meant by the module syntax is that I can use that to get the experience of writing JS tests and imports like in a SPA but without actually structuring as a SPA.
I have heard of Eleventy and Gatsby and others but haven't really tried them
When using DOM HTML as a template you have to ship and execute runtime + compiler (93kB) rather than just the runtime (65kB) on the client.
When you use a build step to compile the template into a bundle you only need to ship the vue runtime to the client - this isn't the same as building an SPA. At the core of the original article is the notion that it should be a goal to ship and execute less JavaScript on the client.
The "Developer Experience" Bait-and-Switch
Any opportunity to complete work at build time that reduces the amount of code needed at runtime should always be taken.
Time and time again Rich Hickey's observation proves to be accurate: Programmers know the benefit of everything and the tradeoffs of nothing.
Many tools are adopted to enhance the "developer experience" while the potential negative consequences downstream are either ignored or downplayed.
I see the solution as quite easy.
Use
*.vuefile, with Parcel. It will internally usevue-template-compiler.Do not put
<template>tag inside Jekyll.I actually followed an approach without a template element so could have skipped the compiler.
I just added v-model etc. to plain html and setup
new Vue(...)at the bottom.Here was the work in progress before moving to SPA
github.com/MichaelCurrin/badge-gen...
I did use vue-markdown tag though which I guess needs the compilation.
Thanks for sharing. I liked the video a lot.
I it was pointed out earlier in this thread but there is a major misconception that SPAs are PWAs. A SPA can technically be a PWA, but honestly there is no justification for SPAs in the context of a well written PWA.
The service worker can eliminate the network latency, which means you can ditch the fat, out of shape, slow JavaScript that are popular fast food frameworks.
love2dev.com/blog/pwa-spa/
SPAs are actually useful. It is justified when you trade a longer loading time for faster interactions in the future. Think a webmail client, like GMail. Sure, you can load a lo-fi version and keep reloading on every click, but the experience of having email load super-fast and be able to answer them fast is just nice. You usually spend a couple of minutes browsing them and then the time consumed on initial load is saved.
Other areas where SPAs are good are live applications, such as betting ones. You always want to see the current stakes, so you need the data pusher from the server to update what you see in the frontend. No one would probably bet on a static site (unless those are bet for something static too, for example "which year a human will land on Mars").
And of course, SPAs are a must in web applications imitating interaction from regular desktop applications, such as Google Docs.
Pity they are also used in blogs (!), contact forms (!!) or just informative website about the small business.
Thanks for the explanation!
My opinion is that React should be a last-resort option considering all the negatives, and even then, should be used with adaptive loading and every web performance optimization possible. Right now it seems like React is the default for every app and no one considers the negatives.
I also know that Facebook, who developed React, is aware of these negatives and serves a different version of Facebook to users on feature phones (a version of Facebook with less JavaScript to download) than the version they serve to users on iPhones. The developers of React know that React is flawed.
Complex apps that actually require JavaScript intensive SPAs usually have adaptive loading in place (Gmail has the basic HTML version for slow connections) but the average developer doesn't think about any of this.
It depends on many factors but you generalize a problem for everyone. We've different analytic tools so we know our customers devices, resolutions and so. Why I must take care about 2G connection guys when the slowest connection we get is with 3G or 20Mb ADSL and on a <5%?
You need to know your market before choosing technology, isn't it?
If my customers are using high speed internet connections I MUST NOT care about those who are with 2G, because you put extra-efforts and engineering for it to run as nice as possible, even building a different version (check the cost of that in time and money) instead on giving the best experience to YOUR customers.
I see all that post and most comments some kinda romantic programming where things are idealized like "welcome all 2G users, welcome 56Kb modem users, let's push accessibility" but on the real world companies do what their customers need and no more for obvious reasons. And if you are a programmer you either will work on a company or being the boss of one so you need to take this in mind (the third option is being so bad you've no job).
If you already thank on that, you probably are on a country with very bad internet speeds then ok, change the title and say "Stop using react if your business market is on [country_name]" and will be zero misunderstanding.
This sounds like you really believed that most of decisions about using technology X are preceded by careful market evaluation, not just jumping on a hip technology bandwagon. I feel that software development is plagued by this kind of thinking:
The thing is, if you think about it from day one, you probably won't spend huge resources on adjusting the project for people who are not a majority of your market. But if you did not think about it, sure, you have to spend time on coming up in post-factum justifications like that ;)
Of course, there are some cases when using SPA is fully justified (see my previous comments) - but let's be honest, in most cases it's not.
I must be biased about business side thinking but that's what you're taught at college, doing before thinking makes you redo several times so it's inefficient 🤷
When you open up a new project to a market there must be a previous analisis, not doing that and complain about the results is something dumb, don't you think?
SPAs fit well on tones of applications while does not fit on tones of others, that's right, the same for microservices architecture vs services or vs monolithic.
That's just the point, there's a big nonsense on recommend to stop using something "as is", your needs will depend on what do you need to achieve (requirements), knowledge about your market, seek the pros and cons for each option and decide which one fits better for your product, that's all
Well put. Some people keep saying "stop using Laravel" but if you don't have crazy concurrent requests, Laravel can solve almost all of the problems you would typically need & can get a lot done using this framework in a limited amount of time. So the context is important. 2G,3G,4G these are relative to the business of the website is catering. Remember we had to support that shitty IE ? Now we don't because the market cap is insignificant for the older versions of IE. There's always a tradeoff according to your customer base.