As opposed to writing a new engine from scratch (Servo), I wonder how feasible it would be to fork one of these existing engines and modularize its contents? The ability to start from (close to) zero and then pick and choose which features you want in your engine would be pretty attractive to lots of developers. The application that never expects to play any sort of media or access any peripherals could exclude the respective components from the build entirely.
I've been using Safari more recently and it really is a lot better than Chrome, and especially Firefox. I don't know how Apple does it, but they somehow made a Porsche in an SUV world. Thanks for the ultralig.ht pointer, looks quite interesting!
I'm really curious how you would fork a proprietary project.
> If someone did that, I think we'd have a really ideal setup for cross platform apps.
Buy safari isn't cross-platform.
You’d just need to build a UI.
Projects like ultralight[0] also show there's even much higher room for improvement. That, and libraries like React starting to leverage WASM, and SharedArrayBuffers sharing memory and potentially moving all application logic off-thread (worker-dom is a cool project, I was able to get a medium-size React app fully running with it).
Finally, you could fork Safari and remove quite a few features. A lot of legacy things, simplifying the DOM quite a lot, etc.
If someone did that, I think we'd have a really ideal setup for cross platform apps. A much faster, lighter, smaller bundled app that supported 99% of modern web apps out of the box.
[0] https://ultralig.ht