Wenn man sich die Gespräche in den U-Bahnen heutzutage anhört
meint man, dass manche Menschen die Warnung
"bitte zurückbleiben" etwas zu wörtlich nehmen.

The decentralized social network Diaspora which was even mentioned by the German Pirate Party now works together with the startup hub Y Combinator. I tried to install it once, but I gave up. It was too complicated, and it seemed too much like a stack of hacks. Probably I would have been able to install it, but for what? Nobody I know would use it. It was hard enough to convince people to use Jabber, and I still use ICQ and MSN with many of them.

People knowing enough to install it are usually familiar with other, more common technologies. People who would, in theory, use it have Facebook and stuff, and are unable to run an own server. I see only a very small niche for Diaspora, and this is the main problem: The most precious resource a social network has is its users.

An alternative that I liked better was Google Wave, which has now become Apache Wave. However, this is also hard to install. And complicated.

And both of them, and Facebook and Google Plus and whatever else, have one common problem: They solve already solved problems in new proprietary ways. What I want to see is a software solution that uses well-established protocols and formats, and just gives a simple layer to combine them.

Firstly, there are already enough protocols for chatting. There is no need to add yet another one. Just use IRC or XMPP. There are already plenty of web-frontends for them, there is no need to reinvent the wheel again. Same goes for notifications. To notify persons about timelines and events, there are common protocols. You can use RSS or Atom feeds, or IMAP. There are web-frontends and libraries for all of them. Again, no need to reinvent the wheel.
It gets a bit harder with discussions, as there is not yet a common format for such discussions in the web. However, there have been fora since a long time, there is enough expierience with these. And outside the web, there was the Usenet, maybe one could adapt the NNTP protocol.
For decentralized authentification, maybe GPG or Tripcodes are the best way of not having to store account specific information on every server, and still verify their origin. This problem is a bit harder, but it is not unsolvable.
"Personal pages" with a lot of design and stuff. Well, there is HTML and CSS. And there is bbcode and wiki-code. It is not like this was not a solved problem.
Facebook was, as far as I read, experimenting with video chats. Well, I already wrote about that problem. For audio-calls, TeamSpeak might be a practical solution, which also allows group chats, but of course is a proprietary protocol. I have heard of Mumble, but never tried it. For video calls, well, there is Ekiga, but there need to be good and stable versions for Windows and Mac. This is probably an open problem. There are protocols, but there is no software.
Desktop sharing is the other thing that is often mixed up with video calls. Well, cooperative working and desktop sharing can be done through VNC, but the applications must support it. But still, this is a problem with many solutions.

I do not say that it is not a lot of work to put all these pieces together, make them stable, and user friendly. I just think that doing it this way will have better results than creating something completely new.

And one final tipp for Diaspora: Packages for the common Linux distributions are a must!