I understand that, but the idea that HTML + CSS can only be sped up by some hugely parallel structure requiring a whole new programming language and browser engine is really quite shocking. I think once webassembly is established and in use, we will start to see steps in the direction of replacing HTML + CSS altogether, with something that has performance in mind.
It seems like a valid point that many of the details of HTML and CSS, while you can use all kinds of fancy techniques to make them a little faster, impose limits on performance.
It's true that the DOM is a bottleneck on many apps (I don't think anybody is disputing that), but "a little faster" undersells what Servo is trying to achieve: order-of-magnitude speedups on GPU-parallelizable tasks, and at least factor-of-two speedups on CPU-parallelizable tasks, all while providing stronger guarantees WRT user security than any contemporary browser.
If you mean throwing out the DOM entirely for something canvas-based, I'm sure that will happen in some cases but I'm very skeptical that it will ever be anything close to the dominant paradigm for the web.
If you listen to talks about the design architecture of Servo you will understand the HTML CSS is not the problem. It actually is an advantage when you want to build an engine that works like a high performance game engine.
