Re: Problems with Bayesian filtering

On Feb 25 2004, der Mouse wrote:

> But then the filters can filter on the javascript itself, just as they
> can on the CSS constructs to set up invisible ink &c.  (Not to mention
> that the more elaborate the spammers get the lower the proportion of
> people that can actually read their message.  People are beginning to
> learn to run with JS disabled already....)

That's why I say it's a losing battle. Where do you stop trying to
understand the javascript and the CSS? 

If you simply scan it without trying to understand what you're
scanning, the noise becomes problematic. Early experiments with
scanning straight HTML tags have shown that indiscriminate scanning of
HTML markup does not reliably distinguish categories of email.

IMHO, picking up explicitly on individual spammer tricks is at best a stop
gap until a unified way of dealing with those is found. That's not to
say it doesn't have value, just that it seems to me too much like the
antivirus race, or the previous generation spam race where people
input specific strings to watch for.

-- 
Laird Breyer.