I can't believe I haven't put the words "Gnu" and "Facebook" together in my mind before....
Of course, something like this should be done. There are important questions about the technical architecture (I only ask one thing -- DONT USE XML...), especially because of privacy, but there are also opportunities to explicitly and collectively decide processes (Yay democracy!) -- if the users of FB could vote on issues of data sharing, things there might be really, really different.
The above post links out to http://autonomo.us/, where Stallman talks about Software as a Service and its anti freedom implications. It is interesting reading, but I think collective projects can be Free (as in freedom) when they are collectively controlled and open; this is the important difference between Wikipedia (Free) and Google Docs (not Free). I don't think the important difference is technical, as RMS seems to (whether or not an application "belongs" local or remote).
"Collectively controlled" is the hard problem. It is probably impossible to solve cleanly without hard tradeoffs, whether in a society (do we let people shout fire? consensus versus majority vote? one vote per share or one vote per person?) or in a computer project (do we sanction bad editors? Are certain topics off limits? etc).
So... I think a Gnu FB is actually an awesome challenge, both technically and socially.
I like the link posted in the first article comment: http://onesocialweb.org/ (except their code base is not up on github yet - hopefully soon) Written in Java, uses XMPP, and comes with an Android client - at least it sounds good.
The article has some interesting points but ideas without code for a prototype implementation to back them up is not very useful.
As stated this is far too ambitious a goal. But let's say we don't worry about traffic analysis and just protect the content. What would a minimal "git with privacy" look like?
Git with privacy is a bit hard. Imagine that you determine that 123456abcde is a the SHA-1 for a bomb-making cookbook. Well then you just monitor the net for requests for that hash, and you see who's making bombs.
Tor-type privacy is also possible; but in this case I figure just delegate to people who seem competent and also aware of the problems.
Btw it's all totally way too ambitious. Who knows what will happen, though?
While impractical and reminiscent of 90s crypto-wankery (oh no clipper chip, et al), it's still far more worthwhile than what Stallman's been pushing lately about cloud services.
Stallman seems to think that users give a shit about software, as if everyone using the AGPL would somehow fix everything — except that thet's total bullshit — people only care about access to their data, the code serving it up is completely irrelevant. You wouldn't be able to run your own instance of Facebook on localhost anyway, even if that was somehow useful.
Stallman's refusal to use any cloud services leaves him woefully out of touch with the reality of their implications.
Stallman has done lots of lovely work. Now he does some OK thinking. Just let him be :)
In my case, one must recognize that the GNU manifesto was also impractical. This totally might not work, but I think the concerns are valid, and the implementation at least sounds fun to hack on.
The OP's post was sort of dumb, but he has a more general point: there is a danger that the people who get involved will see it as a purely technical exercise, ignore the user component, and quite possibly magnify complexity when they should be choking complexity down as far as possible. We don't need a concept car (Gnu Hurd), we need a Datsun B210 (Linux).
Linux was written by someone who wanted to use a kernel that took advantage of the MMU in his new 386. One did not exist, and he couldn't practically fork Minix because of draconian licensing issues, so he wrote his own for himself to use. Linux was written by and for a user.
Hurd, on the other hand, was designed for several orthogonal ideological reasons, and never really written. There was never a user within sight of it, including its creators.
Linux distributions (Redhat and Ubuntu especially) have very much focused on being user friendly, even if their idealized users are more sophisticated than those of other consumer OS makers. Re Hurd -- umm, I was setting up a contrast...
I actually understand the risks and importance of software freedoms. I just think this isn't marketable or practical at all. And if nobody wants to do it then you don't have a social network.
The networking concerns (Metcalfe's law and all that) are definitely a problem. I don't think that an autonomous network could start with facebook; it would have to build up to that kind of application through iterative development and refinement, accompanying deployment increases. The typical scaling problem, in the end, but not insurmountable, given the demand.
Of course, something like this should be done. There are important questions about the technical architecture (I only ask one thing -- DONT USE XML...), especially because of privacy, but there are also opportunities to explicitly and collectively decide processes (Yay democracy!) -- if the users of FB could vote on issues of data sharing, things there might be really, really different.