The questions you pose also apply to centralized services in exactly the same way. There is no reason that personal clouds would be any less secure than centralized services are today. [1]

Even though some people are confused about the difference between browsers and search-engines doesn't mean they don't know how to use either. You'd probably have got the same response if you'd asked about the difference between the Web and Internet. Or any other complex system they don't have deal with day-to-day (internal combustion engine, electricity transmission, etc)

[1] edit: think of things like mass-assignment issue with rails and github or the recent facebook exploit that was posted. Centralized services are just as much of a honey-pot as a personal cloud might be.

No one who steals your Facebook password can get into anything as serious as your entire medical history. The proposal is dangerous precisely because it's centralized — in the sense of centralizing all different types of data on a particular individual. It also, as a side effect of giving the user more control, would give the user more to lose if their account is compromised. Again, with no clear gain for most people.

I feel we're just going to disagree over this but it's probably because neither of us has clearly stated the threat models we're dealing with. Also, it's somewhat hyperbolic to claim that such proposals are 'dangerous'. I could claim the same about the current situation where more and more personal data is handed over to companies, almost by default.