I am still trying to work out why something approximating (2) can't be done through a simple combination of:
- tried-and-tested public key cryptography techniques
- using e-mail as a buffered delivery mechanism (everyone has an e-mail account that can store plenty of messages until they're next on-line)
- borrowing the typical DVCS copy-the-whole-repo approach so you've got distributed back-ups
- writing a native client for whichever platforms you wanted to support.
If someone wants to make money off it, come up with a neat physical way to connect something personal but memorable to keys of sufficient complexity to encrypt everything robustly: mobile app, USB device, whatever.
The only obvious limitation is physical bandwidth and storage capacity, which would make copy-the-whole-repo unsustainable if people kept sharing lots of photos/multimedia content but their friends don't want to "download" all of it. How much this will matter as data storage and communication network capacity increases is anyone's guess.
In the meantime, if you were willing to accept a delay you could have a request/reply system to fetch larger items, or someone could charge a modest sum of real money to people who want to use an actual centralised escrow-like system that is always available to their friends even when they're not on-line. The system doesn't need to be able to see into any content, just to act as a more real-time substitute for the default buffered transmission via e-mail.