When scrmabled text, which was briefly popular in September 2003, meets up with Bayesian mail filtering, you get this:
I finlaly was able to lsoe the wieght I have
been sturggling to lose for years!
And I couldn’t bileeve how simple it was!
Amizang pacth makes you shed the ponuds!
It’s Guanarteed to work or your menoy back!
In other words, spam you can understand, but that your spam filter passes straight through.
I always thought it was a bit arrogant to assume that spammers wouldn’t almost immediately catch on to the limitations of Bayesian filtering. I’m guessing that spammers probably don’t like getting spam to their personal accounts as much as the rest of us do, so they’re probably using Bayesian filtering too. That means they too have access to a list of words with a high spam score, so they’ll know what to avoid.
So here’s a plan I call the Elitist Filter: reject a message if it has a certain number of spelling mistakes. While this would have problems if you received multilingual e-mail, it would have the serendipitous strong incentive for people to improve their spelling. And that can only be good.