> I wish someone could write a program good enough at taking apart headers
> and separating the forged ones from the real ones.  Unfortunately, 99% of
> the ones I do get either originate from hit-n-run dialups on national ISPs
> (impossible to block, not bloody likely to get a response from the ISP) or
> were shuttled through misconfigured mail servers in BFE with totally bogus
> header information and are truly impossible to trace.

It's super-simple to detect 90% of the SPAM messages.  Most contain no From:
or no To: line, or claim to originate from a non-existing account on your
own domain.   A scan for keywords with voting can strengthen or weaken
the detector.  Comparing the originator (if available) with previously
received or sent messages' addressees can also be used as source of information.

If someone of you is able to write a Netscape plugin with a "bool IsSpam(char *msg);"
call, I can fill that gap and complete the program.