INPUT: A number of ridiculously unfiltered feeds ("firehose me!") and a bunch of filter-style rules
OUTPUT: A filtered feed.
- TWITTER : http://hiemanshu.fedorapeople.org/identigator/twitter.xml
- IDENTI.CA : http://hiemanshu.fedorapeople.org/identigator/identica.xml
We have a list of feeds we pull in (autosearches for #fedora and all the tags, etc. we want to look for, and then we run them through a filter...
Sample filter rules:
- Randomly let through 50% of the tweets that match $this_regular_expression
- Let through 0% of the tweets that match $this_regular_expression
- Let through all the tweets that match $this_regular_expression (which may be "come from these usernames" or come with works for me, etc [need to write a beautiful AI for this])
- If things match $this_regular_expression, put them in the "approval queue" (main feature)
- Get a queue with Need more +1 to approve
- Remote DB if possible for queue (why?) (so that we dont have to keep pulling in the same stuff and no two people retweet the same thing [even if it will have once in a 10 tweets])