I've been toying with the idea of parsing my outbound mail server logs every few minutes and using that info to create autowhitelist entries.  (i.e joe@abc mailto:abc.comme@def.com - .com|me@def.com )

The thinking is that if one of my customers is sending email to someone, they expect to get a reply from that same email address. This should cut down on rejected emails dramatically.

My big fear is that the autowhitelist could grow to several meg and if its reloaded every minute, it would get too big to handle.

Anyone else doing this? Would doing this be impractical??

Thanks for your input. 

I like the idea of adding "user sent" mail to a whitelist. I do have a customer who's auto list is about 1.5M in size... I haven't noticed a problem or errors yet and I am assuming it's working ok.

An outbound log parser to build an autosentwhitelist would be awesome. Maybe a new text file could be added just for this purpose and be handled exactly the same way as the existing one.

What would even more awesome is if Spamfilter incorporated an outbound SMTP gateway as well as in inbound gateway and any mail sent out through that smtp gateway would be able to be added to something like a autosentwhitelisted file we are talking about. Once you had an outbound gateway it would probably be easier to implement things like spamassassin, etc.