> Valdis wrote: >There's an even bigger problem - you have to make the difficult choice between: > >1) Flag the DMZ mail server of every site that uses RFC1918 space, since the >"previous >hop" is in their 1918 space. This won't win you friends.... No need to flag it as being dishonest when it is not dishonest. > Valdis wrote: >2) Allow a "pass" for 1918 space, and just accept that spammers will use a >dummy >RFC1918 network (of possibly 1 node looped back to itself) to "look like" (1). But as I said in previous post I know of at least one proprietary way to detect this. > Dean wrote: >I do not think that _all_ anti-spam is useless. I asserted here _how_ all existing (that are in public domain and I am aware of) can be rendered useless (eventually): http://www1.ietf.org/mail-archive/ietf/Current/msg22190.html > Dean wrote: >I think that content analysis holds much promise. Agreed but not by ... [phrase withheld due to IP charter of this list] ... due to information theory. I can not say more until the patent is pending. > Only a few years ago, we >thought that speaker-independent voice recognition was science fiction. I personally did not think so since I got into studying filtering theory in 1986. Much of my early work was due to an interest I had in robotics and A.I.. >Along similar lines, I think it will be possible for intelligent agents to >decide whether email is likely to be interesting to us. Yes it already exists and will hopefully be demonstrated publicly soon. I guess (according to Vernon) that means I am a "Kook" until demonstrated :) Shelby Moore http://AntiViotic.com