I completely agree. I'm having trouble imagining it as anything other than some kind of personalized online borders. I think coming up with a more easy to grok name for the issue is worthwhile. Any suggestions?
It seems to me that he's talking about computers personalizing based on their interactions with you passively.
However, I completely agree that reddit and almost all website that brings together like minded people will have this "bubble" effect. That, however, is nothing new. It's the same as similar groups of people going to the same church, or joining book clubs, or gathering in groups for any reason. The scary thing here is sites that everyone uses will begin to passively function in the same way, making it even less likely than ever for someone who isn't seeking out a diverse range of opinions to find them.
Also worth noting is that there is a moral assumption here that being exposed to a diverse range of opinions is a good thing. If you don't believe that, then this is a moot point.
I think that the premise here, that by giving relevant results search engines will filter the diverse range of opinions which exist in the world, is intriguing. In the short term, I think it's unlikely that online personalization will be accurate enough to create the type of bubble Tim Berners-Lee is suggesting. In the long term, however, I do see this as significant and dangerous. Unlike the way current tribes work, in which a person needs to actively seek and follow their 'tribe', this is a passive action - over time web services get to know you and seemingly shield you from diversity. The passive vs active nature required to keep up with one's tribe is the part of his prediction which worries me.
I agree, but I do think it seems to have a limited appeal within urban areas, then tunneling connections between server nodes in other cities over the regular internet. The tunneling between cities is still vulnerable, but when it gets cut off there will still be a way of communicating. The nice thing about this when localized to large urban areas is that the cost per capita shrinks. I'd imagine it wouldn't be too expensive to set up a bunch of routers that weren't online. Quick Google seems to suggest that a router tends to cost around $1-2 per month in electricity, plus an up front cost. Create a Kleinrock ecosystem in a section of a city would not necessarily require a large number of people adopting it, but rather a manageable start up cost plus maintenance, which seems almost reasonable.
Of course, if a few people are paying for it, you'll still need people to actually use your wacky science project, which may or may not happen... If you build it, will they come?
I think that this seems more suited for urban areas in the US, France, and England more than blatantly oppressive countries. I'd imagine that if this caught on somewhere more seriously oppressed by its governments, a few guys with guns would start driving around with a laptop and knocking on doors:
"Excuse me, sir, but there seems to be a wi-fi network originating in your house with the name 'Kleinrock-Node.' Come with us please."