CNN drops the link
Friday, August 25, 2000 by Dave Winer.
Three pieces in one day 
I can't believe I'm writing the third piece in a single day.
Shortly after my last piece ran, CNN dropped the link from the page I pointed to.
As promised, here are the screen shots.
A big one, showing that it is indeed a link on the CNN site.
http://www.scripting.com/images/fullPageCnnScreenShotDeCss.gif
A smaller one, zeroing in on the link in question.
http://www.scripting.com/images/cnnLinkingToDeCssSites.gif
Upside already has the story.
http://www.upside.com/News/39a6fef00.html
Other reporters are working on it.
But wait there's more!
I got an email from a student at MIT with some very welcome news. Here's the text of the email.
"A friend of mine tipped me that Scripting News has been covering the ever-more-bizarre digital music distribution controversy, so I thought you might be interested in hearing the latest from the MIT bomb-throwing contingent.
"Basically what we're working on is an 'audio fingerprinting algorithm.' Given a few seconds of audio, we can compute a unique fingerprint that doesn't change even if the audio is compressed, played back at a different volume, distorted slightly, etc -- ideally we're shooting for an algorithm that's robust to attack, ie, it's impossible to change the fingerprint substantially without distorting the music beyond anything you'd want to listen to. (There are all kinds of fun psychoacoustics happening, just like in MP3 compression.)
"Then, based on these several-second 'subfingerprints', we can match an entire music file against a database in a way that is robust to having the beginning clipped off, the middle 25% being replaced with a different song, etc, by looking at the match offsets between successive subfingerprints.
"Once you've found the song in the database you can do all kinds of fun things. You have authentic artist, title, and copyright information. You can do collaborative filtering based on what people listen to and suggest other music the user might like. You can rip music automatically in bulk off the radio or stream servers. You can build realtime 'netplay' popularity charts based on what mp3's are playing all around the world *right now*. You can find where samples within a track came from (well, if we get performance that far up.) You can link to lyrics or the band's webpage. You could have a one-click 'give this artist $.10' button (see www.fairtunes.com.) You can stop people from renaming mp3's of their own band (or people screaming in German) to 'unforgiven.mp3' and posting them to Napster. And lots more stuff. There are some more ideas on our webpage (www.tuneprint.com) and I'm sure there's lots of stuff we haven't thought of yet.
"We're trying to be open-source, but we need to get someone to sponsor the server cluster too.
"If you know other people who might be interested, or if you have any tips on how we can get the word out, we'd of course love it if you passed it on :)"
A great fit
So far I haven't talked about Radio UserLand in this column, I wanted to wait until it's ready for prime time. Many of the ideas that Geoff talks about above are already working in our software, but we've been looking for what they promise, a way to know what song you're pointing at, so we can tell if you have it in your music collection. That's key for community and collaborative features, and eventually for linking up to the commerce systems that we hope the music industry is developing. I'll keep you posted as we learn more about what the MIT students are doing.
I sincerely hope this is the last DaveNet for today!
Dave