Name: Anonymous 2011-09-28 4:02
Easy really... there's only four people using /prog/ so the only task is to find out who writes what posts.
Start by scraping /prog/, create a 24H timeline and put a marker on every moment there's been a post. You'll see clusters of posts close together, these were made by the same poster. This is because most /pro/grammers tend to write multiple posts every time they load up /prog/.
Now that we have clusters of messages all we really need to do is run a bayesian classifier on them. First train it to recognize messages within the same cluster, then spread it out to include all of them. Done. You can name clusters and new messages can be tagged with that name (or none if none match (yet)).
So yeah, anonimity is dead.
BTW this took me two hours in Haskell + another couple hours to write the userscript that tags posts. I'd like to see the low level primates barge in here and say GC sucks: fuck off and go write me a device driver so I can do the *actual* work you fucking plumbers.
Start by scraping /prog/, create a 24H timeline and put a marker on every moment there's been a post. You'll see clusters of posts close together, these were made by the same poster. This is because most /pro/grammers tend to write multiple posts every time they load up /prog/.
Now that we have clusters of messages all we really need to do is run a bayesian classifier on them. First train it to recognize messages within the same cluster, then spread it out to include all of them. Done. You can name clusters and new messages can be tagged with that name (or none if none match (yet)).
So yeah, anonimity is dead.
BTW this took me two hours in Haskell + another couple hours to write the userscript that tags posts. I'd like to see the low level primates barge in here and say GC sucks: fuck off and go write me a device driver so I can do the *actual* work you fucking plumbers.