Slashdot Mirror
Q&A With MIT's Nicholas Negroponte
Lisa Langsdorf writes "Thought you might be interested in this interview between Nicholas Negroponte and BusinessWeek Online's Steven Baker.
In it, Nicholas says that peer-to-peer is his prediction as to which new products or services are likely to make the biggest splash, he says:
Peer-to-peer is key. I mean that in every form conceivable: cell phones without towers, sharing leftover food, bartering, etc. Furthermore, you will see micro-wireless networks, where everyday devices become routers of messages that have nothing to do with themselves.
Nature is pretty good at networks, self-organizing systems. By contrast, social systems are top-down and hierarchical, from which we draw the basic assumption that organization and order can only come from centralism.
"
You should realize that this Nick Negroponte is the SAME GUY that whored himself to Swatch to promote their ridiculous "Internet Time" initiative.
Funny... but misses the point. When you hit a web page from home, all the computers (routers, proxy servers, etc) that the data passes through have been built, configured and installed for the central purpose of moving data. In that sense they have *everything* to do with routing your web page.
What Negroponte means is that your phone will pass data for other clients like a router does, but it will also be your mobile phone (a helpful, interactive, personal device). So instead of having a fairly strict division between client, server, and message-passing machines, each device will contain the transport functions and also do something individualistic.
This architecture, it seems to me, will imply encryption throughout -- somehow, people are more concerned by the idea of their data passing through other individuals' devices (what if they look at it?!) than they are sending the data through the hands of a few mega-corporations. I would say this is a good thing...
When someone from MIT says peer-to-peer is a good thing, he's talking about peer-to-peer as an architecture. He does not mean "KaZaA 0wnz!! fr33 pr0n = 1337!!!!111oneoneone." People are interested in peer-to-peer for reasons other than file-sharing because they're scalable architectures that can handle load balancing very well, and have no central point of failure.
Most peer-to-peer research in universities regards creating better, faster Distributed Hash Tables, or DHTs for short. Typically, for N nodes on an overlay network connected by a DHT, insertion and queries come at log(N) cost. MIT has one of the best, called Chord. Some DHTs are very fragile and their routing topology can "break" when under extreme churn (when a flash of nodes suddenly join or leave the network), or malicious nodes attempt to manipulate other nodes' routing tables by creating fake identities (see the Sybil attack) -- Chord has been shown to be very resistant to both. Other notables are Kademlia from NYU (which is under the hood of eMule), and Pastry from Rice (Microsoft collaborated).
MIT has done some pioneering research in DHTs, and they have a lot of great minds on it. I'm making my own peer-to-peer program (hopefully it will be ready in a few months) and it will incorporate quite a few of the ideas they've developed. One of their ideas that I find particularly interesting (and I think should be incorporated into BitTorrent, because it seems like the perfect application) is called Vivaldi. You can read for yourself on how it works, but when applying it to BitTorrent, basicially the tracker would give you peers it thinks you have a low ping time to, as opposed to a random list which may be sub-optimal.
They're also involved in Project IRIS, which aims to develop a decentralized Internet infrastructure using all the latest DHT technology. It's funded indirectly through -- gasp -- the government via the NSF.
So yeah, don't just think that MIT is jumping on the bandwagon. They've been on the bleeding edge for some time.
- shadowmatter