We had a productive meeting tonight.  First, we heard a report from Eric
on what he's learned about free software at MIT: turns out lots of
groups at MIT making software are lame, but nobody cites a worse reason
than that for not releasing their software output.  Then Christina
described what's been happening up at Harvard Free Culture.
Peter mentioned one project they talked about up there,
and we decided we could actually pull it off.

We call it YouTomb.  Read on to see the plan.
Please join if you're interested in helping out.

Greg


--------

2007 Nov 16

MIT Free Culture

Present: price, mrgordon, ecprice, dsheets, macowell, albertni, jansel,
  Peter Olson, kxu@fas.harvard.edu

New project: Chilling Effects for YouTube.
  We call it "YouTomb".

Videos disappear from YouTube all the time, in response to DMCA
takedowns or to Google's own filtering.  Nobody knows (outside Google)
just which videos, or even how many.  How many are really infringing a
copyright?  How many are fair use?  How many are incidentally
infringing -- background music, or someone singing Happy Birthday?

Let's find out.


Here's several ways we can identify what's getting taken down.

 - regularly crawl YouTube, or a feed on it, maybe limited
   to the most popular/viewed/whatever if resources are short;
   when a video goes missing, we'll notice on the next crawl;

 - if we can't crawl everything, we could focus on search patterns
   ("happy birthday", "mlb", etc) manually chosen, or user-submitted,
   in hopes of getting particularly good results.

 - while we're crawling, download the videos themselves, to
   look at for ourselves to identify outrageous examples;

 - invite web users to tell us about their taken-down videos
   (let's not host their videos, but we could link to them);

 - ask Google to provide (us, or the web) a feed of the metadata
   of taken-down videos, rather like they do for Chilling Effects.

 - ask Google (or try to notice ourselves) what videos successfully
   go through the counter-takedown process.


Here's some things we can do with the data.

 - show a feed of the most-recently-taken-down videos;

 - show a feed of videos we hand-identify as bad takedowns;

 - link to hand-identified bad-takedown videos found elsewhere on the web;

 - automatically identify patterns suggesting likelihood to be taken down

 - in Mac's terms, "provide a nice, ajaxy stack of metadata entries of
   freshly-killed youtube videos, with new entries getting pushed onto
   the stack once or twice a second, a la digg labs: http://labs.digg.com/stack".


The two major tasks we'll be working on first are

 - start pulling stuff from YouTube and its API and feeds
   to get a first version of the data and the website,
   so we get a sense of what and how much is getting taken down
   and can better pick our next move (eg, let's find 100
   copyright takedowns per day);

 - approach Google to ask nicely to work together in the way
   that they do with Chilling Effects.


We have a new, higher-traffic mailing list youtomb@mit.edu
for people interested in working on this project.
Mail me (or freeculture-helpers) if you want to join the list.