GoodNews Requirements:

Basic Description:
- GoodNews will be a user-run user-edited news site.
- We pay users for their efforts, and collect a small monthly fee from everyone
  to cover it (as well as operating costs of the site)
- Users submit news items (or interesting links they find)
- Users tag news items to assess their qualities
- Tags are either tri-state (present, absent, or present and negated, e.g.
  "Funny", "not Funny", or no funniness tag at all) or scaled -10 to 10 (e.g.
  "Funny +7.5" or "Funny -8"), or have an argument, like a URL or other
  story (e.g. "DuplicateOf(Story from Last Week)").
- Users get a personalized newsfeed from the system, based on the likelihood
  that the individual user will appreciate that story.
- Users provide feedback on their personal feed by giving each story a
  thumbs-up or thumbs-down.
- The quality of each individual's newsfeed should improve over time
- The system will consist of a simple web interface for all users, and a
  database and computational backend (the computation is for finding good
  news stories)
- For a simple start, the system can use Bayes' Rule to predict enjoyability
  of stories

Usefulness:
- Well, other news sites suck
- Other "personalized" news sites don't motivate high quality editing
- Tagging allows for communities of like users to form automatically
  through common interests; there's no need to be explicit about grouping.

Users:
- Everyone who uses the internet for news (lots of folks, including my parents
  who know nothing about computers)
- Obsessive editors and taggers from places such as wikipedia, photo sites,
  etc.
- Well-respected and prolific members of current news sites who could be
  raking in 10s, 100s or 1000s of dollars for their contributions, but instead,
  get "internet cred", which clearly isn't as valuable.

Scale/Scope:
- As big as the internet

Divisibility:
- Individual pages can be changed independently, and don't really rely on
  each other
- Various correlation engines can analyze stories in the background and offer
  suggestions; they are also entirely independent
- The 2 major portions, the web UI and the news analyzing backend, communicate
  by simply dumping entries into a database in the right fashion, so they
  may or may not even share any code

In C++ on Linux:
- Not likely; there are so many better language choices, it's almost funny
- Still on Linux, of course

Possible Extensions:
- Statistics reporting
- Other ways to find good stories besides Bayes' Rule
- Other kinds of tagging
- Analysis of link content itself
- Automating various tasks like duplicate-finding