> Things have become so complex in the > last 10 years that it is difficult to > follow, what's left for development.
Maybe a modularized approach could solve this. You know, once "lp" was a huge software, with print management and printer drivers lumped into a monolithic package. Then lpr-ng came about, IIRC it was the first print management software that separated drivers to their proper place and thus was able to focus on the print management job itself. Nowadays each printer has its own driver, readily usable by any print management software you'd choose.
I suspect similar dissection might be possible in web browsers too. Ofcourse this must entail a strong standardization regarding how each piece of a browser interact with each other (or at least with the main trunk). In such a hypothetical structure, a browser would be no more than a slim, main trunk, perhaps outsourcing even the basic HTML rendering to an external module ("driver" in lpr analogy). So, browser trunks (which are quite manageable in size) compete with each other on their own merit, likewise "drivers" for a specific function do so between them. This would both promote natural selection process among the components, and would - hopefully - free browser development "privilege" from a few tech giants.
But then it would be a different story to choose a browser. You would need to choose the main browser and external modules separately. Maybe a new wave emerges from this, where people start making "browser recipe"s and package them together. And then we would have perhaps hundreds of recipe packages to choose from.
I don't know if it makes much sense. Just brainstorming.