Yes, a declarative web API is _far_ harder to get right, but the web community has created an amazing long-term compatibility story. It _should_ be difficult to add things to the web-at-large precisely because it has to live, essentially, forever.
Even if Houdini fails adoption, WebAssembly + Canvas/WebGL/WebGPU will do it.
https://www.qt.io/qt-examples-for-webassembly
https://blogs.autodesk.com/autocad/autocad-web-app-google-io...
With a wasm+web hybrid, there's a clear separation between the imperative wasm bits and the declarative web bits, so you can even combine them in ways that allow both stacks to perform well.
They should make Houdini available to use not just with JS but also with WebAssembly. Then you could use a more fitting language of your own choice.
Like, hey, this URL just points to say, a QT app, and then it runs in this "browser" window over the network. Where have I heard of this before...
Everything else just feels like loads of duct tape on top of an over-stressed document browsing paradigm. Just rip it out completely.
Recent efforts to strip browsers down to their core functionality, and provide a world where the current web is effectively userland code:
* web components
* service workers
* Houdini
Imagine a future where browsers just implement these low level concepts, and HTML is defined as an (open source/community) library that all vendors bundle as part of their browser.
And then, also consider how much overlap there is here with traditional operating systems. Your next device OS could be just another layer on top.
And Facebook uses on their Redis cluster. Design is important but far from the most important.
[1] Working example: https://scratchy.ooer.com/ - click and drag to scratch off the symbols. The code lives at https://github.com/onion2k/scratchy
