Do I Look Like a CEO?
businessweek.com
businessweek.com
The idea that people will continue to upload the file after they have downloaded it seemed silly before someone actually did it. Using a http tracker was going against the trend that all the other protocol designers were following at the time.
Then someone made it, and it became a big hit. Just like YouTube - it sounds technologically silly before it is done. NOW it's easy to see how obvious it is, but before it was done, it was not the same way.
Bram Cohen is the polar opposite of Ian Clarke, and that's why his project is so much more successful.
This part had been well-validated before BitTorrent came out. People were already using Napster, Kazaa, and AudioGalaxy, all of which relied on people leaving files available for upload after they finished downloading them.
Using an HTTP tracker was a pretty courageous act though, and I don't think the rest of the protocol would work without it. At the time, everyone was doing desktop clients and trying to make things as anonymous as possible, since they were all afraid of copyright-infringement lawsuits.
I kinda wonder whether Bram really didn't know it would be used for copyright infringement, or if he just says that because he's smart enough not to get sued. Every other P2P filesharing system was used to share music, and the mainstream media was going on and on about how Napster was the death of the music industry. You'd have to be living under a rock not to make that connection.
Sarcasm aside, universal tools are like that - people use them for many things.
You can't predict the future, but you ought to have a pretty good idea what's happening in the present.
I remember arriving at college, where each person had their own pet ways of downloading MP3s. Within a week, my whole dorm had basically standardized on AudioGalaxy.
Plus there is no bandwidth wasted on sharing content other than what you are interested in. Almost every other client wants to make you share all kinds of other stuff to even use the network: emule, freenet, tor, i2p, gnutella, etc.
Finally, pure luck has allowed the laws to progress in a way that hasn't resulted in the immediate shutdown or blocking of popular torrent sites. It should be noted that in certain countries this isn't the case and bittorrent has been a total failure there.
Also, on the majority of American college campuses today, bittorrent is a huge failure because it is easily fingerprinted and will get you banned from the network. Maybe they have finally added some stenography to protect against this but not that I know of. Then, there are the large ISPs that are degrading or blocking bittorrent now.
Of course, everyone knows that the real reason bittorrent beat freenet is because bittorrent was python and freenet was a big steaming pile of java:)
The use of torrent files as a universal link to the downloadable material is also genius. It removes the need for implementing a costly peer-2-peer search mechanism (Gnutella and Kazaa spent quite a lot of resources on search), and a potentially cluttered GUI. Instead, the torrent files can be distributed in any way people find convenient, eg. www, e-mail or RSS.
On top of that, including the tracker-URL transparently in the torrent file is a huge usability improvement. No more searching for lists of super-nodes to enter the network, as was the case with Gnutella.
This is true of many, if not most gamechanging breakthroughs. Skype wasn't the first IP telephony provide - but they made it simple and were brave.
There are two books that I must recommend for any engineer that is in the same situation as Mr. Cohen.
Blink, By Malcom Gladwell. This book explains the concept of intuition and thinking with the subconscious. I think this is something that is lost to a lot of science people that have to rationalize everything in life with concrete details. In my opinion, I think that our subconscious is a much more logical real life processing unit than our conscious because it can link concepts much more quickly. As the old adage says, sometimes you have to trust your instincts.
How To Win Friends & Influence People, by Dale Carnegie. This book goes into incredible depth with a bunch of narratives explaining the proper way to deal with people. Very interesting book, in fact, this is the one book PG recommends EVERY startup entrepreneur to read, and I couldn't agree more.
http://bittorrent.cvs.sourceforge.net/viewvc/bittorrent/BitT...
BitTorrent is an implementation of an idea (group delivery of content using FEC) that was quite popular at the time. There was already an open-source C language startup doing it, and more notably the idea was the foundation for Digital Fountain, which Michael Luby started during the tail end of the boom. The Fountain guys even tried to standardize it, as a protocol called FLID-DL.
The relationship between BitTorrent and the kernel of its idea seems to be approximately the same as the relationship between Napster and IRC XDCC, or between WinAmp and MP3 players. I think that's a compliment, by the way.
IIRC Luby came up with a rateless encoding algorithms while working on overcoming content delivery issues in lossy networks. It had less to do with a group delivery per se, and more with the lossy transmission medium. They did initially focus on a single-sender setup, and in that they were quite a bit different from BT. Only after DF was founded they started looking at alternative applications of their (patented) ideas and resurrected the swarm-casting stuff.
In other words I wouldn't mix DF and BT together. They are certainly related, but they are more different than they are similar.
I founded a VC-funded startup that ran from '99-'01 that was directly in this space; we implemented basically the same system as BitTorrent for group file transfer. We definitely felt like there was significant overlap with Digital Fountain (though we were using off-the-shelf PD coding schemes, like BitTorrent did).
"Lossy transmission medium" is another way to think about multicast congestion, and multicast congestion is another way to think about "how to get content to a large, diverse audience of people", which I think sums up BitTorrent.
The big difference between BitTorrent and DF is that BitTorrent is P2P, and DF isn't. That sounds huge, but realize that "P2P vs. Planned" is an informal way to describe the major approaches to multicast reliability protocols.
Feel like sharing a name of the startup ? :)
Again: not saying XDCC is as big a deal as Napster.
I believe this is the first mention of Bittorrent when still under development.