Now that's a fine language for a server. It combines the type safety of Ruby, the memory safety of C, and the terseness of Java.
(I'm joking, mostly... Actually I was a big fan of Obj-C for desktop apps. Fond memories of times when I didn't have to care about servers and ever-changing web frameworks.)
> It exposes all your frontend source code for everyone. If you don't want to open source your frontend, you might want to remove those source maps from public access.
I wonder how much difference LLMs today have on being able to turn minified JS into something easily readable? JSNice already worked pretty well and I guess that was comparatively naive. You won't really stop anyone motivated to reverse-engineer it by not providing source maps, but you'll definitely stop at least some curious people from understanding how websites work. Your frontend also doesn't suddenly turn "open source" just because you shared the original source via source maps, that part sounds kind of FUD.
Is there any reason sourcemaps are a genuine problem? I'm out of touch with the JS world, but I wonder if code is shared between server and client and server code may show in sourcemaps.
Often though, Javascript is hard to read not because it's been obfuscated, but because its been transpiled and/or minified for smaller network payloads.
I can understand why some don't want to ship their sourcemaps to prod, but also it really doesn't matter all that much.
Some sites want to ship small bundles to the client by default, sourcemaps enables that + you get to introspect it because it's downloaded only when requested. Literally best of both worlds :)
To elaborate on your comment, if you just ship sourcemaps in production, that means you can ship minified code and track down what _actual_ source that you _aren't_ shipping to users is getting called, is in stack traces, etc.
I remain convinced App Store Connect is the project they put interns on. It also explains why they keep redesigning / reimplementing it, then losing interest and leaving it part-finished and incoherent. It’s because the interns working on it go back to school.
Svelte files look like HTML+TS files. You aren’t learning some abstraction to HTML, you are just using HTML. But it adds the modern bits you need: reactivity, loops, components, routing, etc. Nothing react doesn’t have, but the devex is great.
Other benefits:
- your app is compiled. You don’t ship the framework to clients, they just get a minimal compiled app.
- The rendering modes are pretty great. Any page can be server side rendered, or client side, with per page flags. You also can easily setup SSR for the first page, and CSR for later pages - both the fastest option. It will even pre-fetch the next page when you hover a link, making most nav instant.
(Not a user, just evaluated it previously. Please correct what I got wrong.) They compile the reactivity statically, so instead of tracking effects at runtime, they generate code for it. I'd guess it means slightly more JS to download, but less initialization in runtime.
However, they recently added runtime reactivity to be more flexible, so it seems to me they are becoming VueJS.
Copyrighted content is copyrighted content. It doesn’t matter if/how Apple published it, others don’t have a right to redistribute it without a license.
Unsurprisingly there are many frameworks/initiatives that end up falling by the wayside over the years, e.g. MacRuby was being lined up to supersede Objective-C for app development at one point.
Didn’t SproutCore become Ember [1]? Just my vague recollection, not sure if that was the case though. Anyway, Ember is still used and maintained, despite not being very well known.
Dumb question but Apple’s apps are buttery smooth. I just assumed they were using swift and not a web stack to render their UI. Am I completely wrong?!
Apple Music is not buttery smooth and was just a web view for a long time. I feel like I read that this changed a few years ago. This didn’t change the fact that it’s very slow.
There's also some parts of System Settings that were always web views, which I always found surprising for a company trying to make the case for native apps.
It's pretty clear to me that JavaScript is becoming the de facto standard for UI/UX programming, regardless of platform, and regardless of web vs. native targets. Even GNOME has JavaScript bindings. [0]
The problem is performance... requiring a web browser to draw a UI takes a LOT of CPU and memory, and not all devices have enough power to deliver a smooth experience across all potential workloads.
I worry that every year we keep increasing our processing requirements and bloat without good reason for it.
Why should every Windows release require a faster and faster CPU, and more and more RAM?
The recommended amount of memory for Windows 95 was 8 megabytes, and for Windows 11 it is 8 gigabytes. Why is this not horrifying?
My small Linux system with openbox GUI barely cracks 100MB memory usage in 2025.
What makes the browser slow and inefficient is the fact that it's not a UI framework. It's a system to display text and a couple of images on a 2D plane where every element depends on every other element.
Almost every single interaction and change requires the browser to recalculate the layout of the entire page and to redraw it. It's basically Microsoft Word, with nearly the same behaviors.
And there are no proper ways to prevent that behaviour. No lower and low level control over rendering. Awkward workarounds and hacks that browsers employ to try and minimize re-layouting and redrawing. Great rejoicing when introducing yet more hacks for basic things: https://developer.chrome.com/docs/css-ui/animate-to-height-a... etc.
In none of them text is primary and all other incidental?
> What UI frameworks don't do this?
In which UI framework actions like "set focus on an element" triggers a full page re-layout?
Also, in which UI framework there's even a discussion of "try to not trigger re-paint/re-flow"?
And yes, I know about immediate mode UI where the entire layout is re-calculated every frame. But then they can usually render thousands of elements at 60fps.
> In none of them text is primary and all other incidental?
This is a pretty outdated way of thinking. If we were only speaking about (old) HTML then maybe. But these days HTML and CSS are basically inseparable and go far beyond "text is primary".
> In which UI framework actions like "set focus on an element" triggers a full page re-layout?
I don't see why a browser would need to re-layout for just a focus change. Unless, of course, the CSS changes require a re-layout after the focus has changed (`:focus` selector).
Every UI framework can run into performance issues with layout and/or painting if they're complex enough (or poorly made). There is fundamentally very few differences between HTML+CSS and other UI frameworks.
> What makes a browser so much more inefficient vs. other UI frameworks?
The fact that each app carries their own copy of the browser engine.
Teams, Chrome, Steam - that's at least three Chromium engine embeds that all take up hundreds of megabytes each. Not to mention Steam is in the background and has no windows visible, and yet it has the Chromium helper processes gobbling up RAM. WTF is this shit.
Life used to be easier in the Windows 98 days with OCX, you just dragged a webview in the VB6 application designer and that was it, and IIRC it was even possible to embed Firefox in the same way for a while...
> The fact that each app carries their own copy of the browser engine.
This has nothing to do with the browser itself. Ideally everything would use the same browser instead of shipping their own. On Windows it's already possible with WebView2 but developers need to choose to go that route. Teams should already be using it.
> Why should every Windows release require a faster and faster CPU, and more and more RAM?
I don't know. But does it? It doesn't seem like you verified that yourself - you're comparing stated recommended specs of Windows to actual usage of Linux.
Have you used other ones? Not a dig, I've primarily used HTML/CSS for UIs and have been playing around with Compose recently and haven't made up my mind what I like more.
GNOME has its own interpreter, kinda how React Native does it for mobile. But performance all boils down to the layout engine. Most native UI components take shortcuts with text which is the most difficult thing to render. And the widget tree is simpler.
And there’s the whole inspector in web browser, meaning that the layout is not done once and forget. There’s various sub components still present for whatever features. Great in the browser, not great for standalone apps.
A choice of tech stack can never be enough to prove anything. It only establishes a lower bound on resource usage, but there is never and upper bound as long as while() and malloc() are available.
In case you want to save sources with the ability to fetch all possible lazy chunks, last year I made a tool to do exactly that:
https://github.com/zb3/getfrontend
(note it won't work on apps.apple.com because apple has removed these sourcemaps)
Honestly the site[1] is very basic and pretty damn slow. When I click into a different category there is a noticeable delay of 1-2 seconds before the new page loads. I don't want to replicate this in any of my own projects.
I wanted to write, 'They have to load the entire catalogue for a category, including all the images. What did you expect?' However, having looked into it a bit more, it seems that they didn't particularly care about optimising load times.
That's what this type of SPA architecture leads to unfortunately. Routers should immediately display the navigated to route with place holder content / skeletons, but instead all the frameworks basically wait for all the data to load before transitioning. You can technically stream the data in but even a single awaited promise will block the navigation until it succeeds. And it's not an issue that shows up in dev because typically the data loading is instant.
The flashes signify actual changes. It's a secondary signal to resume paying attention to the page.
What I truly hate are animated skeleton boxes or element level spinners. Why are you trying to hold my attention on something that's not even loaded yet? We all understand the UI paradigm and implicitly understand network delay, you don't need "comfort animations" to keep me happy. I'd rather use the time to look at any of the other tabs or applications across my screens. Then the flash of content actually means something.
The loading state is indistinguishable from the page crashing. Did the JavaScript fail, or is the connection just slow?
> animated skeleton boxes or element level spinners
Good news! Browsers have low motion settings. Any programmer worth their salt will respect this and the skeletons won't be animated.
> Then the flash of content actually means something.
On the contrary, if the content is loaded in multiple parts (in my own application, I split the loading into multiple requests: one is cacheable across multiple pages, one is cheap, one is expensive), you either need to not render anything until everything is loaded (bad: the user can't interact with the parts that loaded first), or the page jumps around as content loads in. Skeletons fix this. UI elements in the middle don't end up being 0px tall and moving the stuff below them around nearly as much. How annoying is it to nearly click on something and the page jumps violently and you mis-click?
It honestly sounds like you just don't like lazy programming. That's very fair. Skeletons done right just mean the page is the correct layout while they're partially loaded. Without that, the content is literally unusable (you can't read it interact with things that are jumping all over the place).
The point of skeleton loaders is to prevent the page from jumping around furiously, which would force the user to re-parse the layout (possibly) multiple times.
In my experience it's just amateur UI design that causes this. Your display areas shouldn't change size unless the browser changes size. There should be nothing that is "content fitted." That's a historical mistake of early HTML but it's something easily overcome. You really do have to get the HTML+CSS to work like a desktop app before you layout your SPA.
Worse still, applications like microsoft outlook on the web, use the skeleton boxes with comfort animations. What they don't do is pre layout their icons. And different icons will appear in different contexts. I often get the case where I aim for one icon, something will load in, create a new icon, and push my intended target out of the click.
Skeleton loaders are a bad kludge over an initially ignored problem.
Which is fine. Nothingness, or a generic spinner actually don't lie to me.
Skeletons lie by making an impression that the data is just about ready. So there's this failure mode where data is NOT ready because of a slow app/network, and I end up staring at a fake. Even worse, sometimes skeletons also break scrolling, so you end up even more frustrated because your controls don't work.
It far and away beats the alternative which is clicking on a link and nothing happening. Feedback should be within a frame or two of latency, not seconds...
The web version of the App Store? It's always been web and webview based, there used to be a preferences/default command to enable web inspector for App store, Music and more Apple apps on MacOS.
A DMCA takedown is inappropriate, as no copyright was circumvented. It was freely distributed (albeit briefly) on Apple's own website. A DMCA takedown at this point is entrapment: https://en.wikipedia.org/wiki/Entrapment.
That is not how copyright nor entrapment work. Somebody putting something on their website does not grant anybody a license to distribute copies. Filing a DMCA takedown is not entrapment.
https://en.wikipedia.org/wiki/WebObjects
reply