Reply #15: Doesn't work like that [View All]

[Prophet 451]

I work for the biggest spirituality site on the web, we tried this idea but couldn't make it work.

The problems:
Firstly, your spam filter scans the header of your mail while it is held in a seperate location (the mail server). This is why those annoying ads for slimming pills or Viagra often use l33t-speak in the header and plain text in the body of the message. Having to scan the entire message (which comment filters would have to do) ratchets up the workload and since comment boards don't (usually) have an intermediate "holding" stage, it would have to be done in realtime.

Secondly, think of the sheer amount of racial epithets around. I could probably think of a hundred off the top of my head, I'm sure you could too and several of those only become offensive in a particular context. For example, if the comment filter sees the word "gorilla" in my post, am I talking about Obama's racial heritage or his work on envirmental issues?

Thirdly, now multiply all those epithets by the number of variations that can be done in l33t (the substitution of numbers or symbols for letters). The variations are almost limitless. I know guys who can conduct whole conversations purely in l33t which are perfectly understandable to anyone who gives it a second's thought. For example, if you write the N-word as "n1gg3r", a comment filter won't catch it but anyone reading knows exactly what you're saying.

It can be done but it requires a lot of processing power and when you think about the number of hits that a news site gets each day, you're talking about massive amounts of processing power. Whether the site should have expended that amount of time, effort and manpower (these things are not simple to code) is open to question, I'm just pointing out that it's a lot more difficult than your spam filter.