Back to tech.

I think I've learned more in the last month than in my entire extended tour of duty in college. I've got ideas piling up in a draft post, don't have time to get it all into a legible form.

Had an idea on the way home from work today while listening to 1984 on CD...

• User clicks a dowload link that is mirrored by Universities, misc other bandwidth benefactors.
• Server checks to see if the user has a bittorrent client available.
• If no torrent client exists, the data is sent via a traditional FTP like system. If it does exist, it can take advantage of the aggregate bandwidth of the active neighborhood.
• All downloads would work via the bt:// protocol similar to how you can download a file now in IE or Firefox via ftp using ftp://server/filename.exe
  

Here's the good part though. The hybrid servers are able to dynamically shift load priority based on demand. Not demand for the files on that particular server though, the server is aware of demand on all of the servers in it's neighborhood. A neighborhood could be a bunch of Universities in collaboration. Now if there is a sudden spike in demand for lets say a 3GB breakthrough video documentary the server that is getting swamped by requests can send a copy of the video to one or more of its network neighbors and then add the other server(s) to the "tracker" so that there would be a seamless increase in bandwidth. So how's that different than just adding a torrent client to all of the servers? Well the difference is that the seed files would be handled separately based on an awareness of load. A file could be added and removed almost instantaneously to the other servers. Its a client server system but weird in that the servers are actually the main part of the distributed network because of their huge bandwidth. If the server has surplus bandwidth then it doesn't care if the client isn't sharing bandwidth unlike the current incarnation of BitTorrent. Initial client installation would be provided by the java wrapper being used by bannedmusic.org.

The beauty of it is this: You don't need overworked sys admins to upload files to University servers. The file would propogate (like DNS) to all "neighborhood" mirrors within an hour or two of the release of the MHBF (must have big file).

I was thinking of using the theory behind DHCP to create flexible work schedules for businesses with limited parking availability or even based on home locations and traffic data. Maybe I need a hobby.