I think part of the problem these days is that the most creative and innovative efforts are going towards the browser. And, indeed, if browsers keep growing and

I think part of the problem these days is that the most creative and innovative efforts are going towards the browser. And, indeed, if browsers keep growing and start to provide an credible, sandboxed, near-native-app equivalent(e.g. pNaCl + local caching, low-level input, sound and video APIs and full network capabilities) the desktop could potentially be stripped all the way back to the programmer-and-sysadmin-centric environments we had in the mid-80s, and usability focus could be on supporting the browser experience instead, as ChromeOS is attempting. As it is, the Linux desktop is still working towards "yesterday's experience," since we don't have good enough browsers and there are tons and tons of legacy apps. Unfortunately, supporting yesterday's experience is harder for usability since it's conceptually richer, and most of these concepts are dueling at the bits-and-bytes level, where they're hard to reason about and present in digestible ways. The browser can blackbox a lot of stuff behind the standard APIs, something which the OS does not enjoy. Another problem, just as large, is that there continues to be lots of new hardware, and correspondingly new drivers to support and specs to reverse engineer(since many manufacturers remain unwilling to assist). This is a problem that can only be solved by maintaining a hardware taskforce roughly proportional to the amount of new hardware released, but it's not particularly interesting or sexy, and it's not a core goal of most of the Linux ecosystem to create premium-grade drivers for all hardware.