Long-term scalability first, mobile/embedded usability second. Failing at scalability would eventually leave Bitcoin in the hands of "the rich ones" (namely Google etc.), which makes for an even worse scenario than banking system. Relying THAT much on processing power or disk space is a doomed solution to me.

Generally speaking, it seems that Bitcoin keeps afloat based on ridiculously optimistic predictions, whereas any software in the world is written with a constant eye on worst-case scenarios.