There are some weak points for web. One of them is performance, hopefully it'll be resolved by Webassembly, so you could write performance-sensitive portions using C++ or Rust. Another one is using hardware and that's resolved by evolving web standards like WebUSB, WebGPU and so on. I don't think that replacing HTML, CSS or JS with something different would solve any actual problem in the web. They are not ideal, but they are good enough and that's all that matters.

Well, technically you could replace JS with any other language compiling into JS or Webassembly. Then you could just use one full-screen canvas and draw anything you want there, so you could use Qt or something like that. That does not require any new APIs, but I don't think that's the good idea in the long run.

To be fair, CSS is making big strides in this regard with Houdini[0]. It's not yet complete but it's a much better alternative to composing divs and spans in weird ways to get the look you want.

0. https://developer.mozilla.org/en-US/docs/Web/Houdini

However, I think the need to support the current web stack will continue for the foreseeable future since it would be a terrible waste to simply abandon a platform that so many applications are written for.

So much tech today is being built just to put a band-aid over this stuff, instead of addressing the core issue (the browser is no longer used to just deliver static documents). I can't be the only one that wishes for a new language/browser. WebAssembly might be a saviour, although I'm unsure how issues like accessibility will be addressed.

If it was just a matter of a fresh start with a new language/stack, then Android/iOS/macOS dev would be compelling experiences. But they're not. And they're changing just as much, bandaging over themselves so much that you end up credentializing in a whole lineage of idiosyncrasies worse than just putting a <select> in a modal, except unlike the web you're just building something that works on an iPhone.

Clientdev is hard.

Client Dev has regressed since those days, now it's in a state where devs can't even imagine anything better, but we clearly had that...

Try Sciter then, it was designed precisely for running desktop UI using HTML/CSS as UI declaration languages.

I was native UI developer initially, then Web Frontend Architect who returned back to desktop development taking HTML/CSS with me to Sciter.

Am I the only one who thinks the sooner the web goes back to just delivering static documents, the better?

As a dev if have to choose between a flawed but still very capable open platform and proprietary walled gardens where corporations wield absolute and arbitrary control I’ll take the open option and so should you.

I don't do native software development but just looking at things, it seems like software is eating the world, but also at the same time that web technology is eating software development.

I don't think the web is the best development tech on any OS, but it's proving to be the best cross-platform tech that runs on any device. It will continue to evolve to be more suitable, but the idea that it's too weak or wrong is exaggerated.

Looking at what web browsers are both capable of and expected to do today I think what we really want is an "internet terminal" which would be able to contact resources over the internet, negotiate both local and remote resources for an intended purpose, and then share state bi-directionally (stateless is just zero state).

To just conjure a couple of quick examples you can imagine the differences required in providing a service that produces a simple JSON feed versus a full interactive experience that mutates a shared database. The JSON feed wouldn't require a display and could easily be consumed by autonomous agents while the other might require a minimum display size and want to use a widget set to produce controls and toolbars, etc. If my client agent is allowed access to my video display, mouse, and keyboard it could offer these to either service and allow it to respond in whatever way it feels appropriate. The JSON feed could provide a simple UI for viewing recent events, for example.

Now we do all of this today with HTML, CSS, JavaScript, etc. but none of it feels organized and all you see is re-invention after re-invention of essentially these same concepts. The fact that we even call them "applications" when they are "documents" is really what seals it to me. I personally would like to see something that is more true to the reality of computing and internet connectivity today and looks at it from a device-to-service POV. Locally I might have a graphical display, audio, and inputs but I might also have access to an SQL database or some other network-attached service. Assuming some level of trust and if negotiated properly there's no reason a remote application couldn't make use of these local resources using credentials and security profiles under my control. The can of worms here is that this is basically RPC which is the other never-ending disaster...

Internet is hard. Let's go shopping.

Scoped CSS works really well in conjunction with web components. Languages like SCSS make everything more sane, especially in combination.

I'm currently messing about with Yew[1], which is an entire SPA framework in Rust/WebAssembly. There's about 5 lines of bootstrap JS involved (and fiddling about with bundlers, which is much more sane since Rollup[2]).

[1]: https://yew.rs/docs/ [2]: https://rollupjs.org/guide/en/

The web is a mine-strewn no-mans-land between competing factions, which accounts for a lot of the strangeness.

(Ironically, the tech is leaking in the other direction as web apps become more popular on both mobile and desktop)