For an enterprise, yes, analytics would certainly be more important; because you'd want to measure success of whatever...but those are goals more for a company. If I only wish to share with, say, my family and a few important friends (without resorting to email blasts) the types of things folks normally share on a centralized social network, then analytics isn't important for me. I just basically need to host "the party" but don't care as much about measuring the success of "the party". My goals are not necessarily all the same as twitter/facebook, etc. ;-)

No, it is for something bigger than a company, it is for the world: if there is a earthquake or tsunami you want to receive the alert immediately.

I am not pro Twitter, I am just saying that from the computer science point of view you can't find a distributed solution to one of the problems Twitter is trying to solve.

Ah, ok, sorry i think i misunderstood your note because i fixated too much on the term "analytics".

For what i will call "broadcast" events - such as your very valid tsunami, earthquake examples - sure it would be essential for those alerts to get out...But whether you're using a centralized or a decentralized approach shouldn't matter - assuming of course that one of the nodes of the decentralized method gets the needed input about the alert - in order to pass it on to the other peers of the decent. network. Much like email and websites, and to some degree mobile phone carriers, having a decentralized network does not necessarily prevent fast distribution of important/up-to-the-minute/broadcast information such as alerts. In fact, one might argue that having a decentralized approach might help with overall resilience. That being said, as much as i am a proponent of decentralized (and open) social networks, they have many non-trivial obstacles to contend with...but i do hope for one day to not have to exclusively depend on closed and centralized platforms like twitter et al.