Re: Realisticness of header rearrangement

On Apr 06 2004, Mark E. Mallett wrote:
> > 
> > The obvious issue with it is, as you point out, that filters must
> > either say "spam" or say nothing. I'd extend this a bit and say that
> > filters also aren't allowed to indicate degrees of spamminess.
> 
> I disagree there nor do I see a reason for it.  options are good;
> limiting options are bad.  Let's not limit filters to doing only
> the things you or I want them to do right now.  

Disagreement accepted :) Perhaps I can clear up the first quoted
statement, ie explain the reason.

Bill's suggestion is for filters to be either quiet or say something
bad. His system prevents tag spoofing because 

1) everyone can agree that (under this system) a tag saying something
good is fake

2) spammers aren't stupid, and (under this system) will not insert
fake tags whose purpose is to say something bad about their email.

I then suggested that if there are degrees of badness, spammers could
try to spoof tags such that they say the least bad thing, hoping that
would trigger failsafes designed to minimize classification errors.
So this implies one particular limit for that particular scheme.

All of this is academic if one's policy is to never trust other 
filters' tags, or one's own. The merits of such a policy are another 
discussion, which we should probably have on this list too. 

I take the view that discussing possible solutions of this trust problem 
can be independently valuable. You never know when they could come in
handy, and at that point it's important to understand the limits of
such schemes. For me at least, this discussion isn't about limiting
options for filters.


-- 
Laird Breyer.