On Wed, Oct 15, 2003 at 07:09:53AM +0200, Thomas Roessler wrote: > That looks a little expensive -- you'll collect a rather long > whitelist over time. I'm attaching the scripts I'm using nowadays; > this is based on a Berkeley DB I put in my $HOME. As a secondary > use, I'm also using this whitelist for mutt's query feature. I also thought of DB_File, and perhaps Sendmail::AccessDB, http://search.cpan.org/dist/Sendmail-AccessDB/ but isn't it an overkill, really? How much mail do you need to send to different people to collect even a 1 MB whitelist -- and then again it's most likely to fit in memory on Linux anyway? Meanwhile, you're going to load the whole mastodon of perl with BDB every time. Just curious of tunings... For now, I modified my whitelisting shell script so that any patterns in ~/.whitelist.exclude are, um, excluded from ~/.whitelist. So, if you put a line with list-dns-notbulkmail in it, qsecretary's dns@xxxxxxxxxxxxx acks will not be seen in your whitelist. (Need to set up a procmail autoack to it somehow -- otherwise I have to explicitly procmail it into inbox, then wait and respond to [y]acks. Anyone has an autoacker?) I think more useful feature would be DCC integration, so known good mail is not checksummed. DCC has its own whitelist in a flat file... BTW, I hooked up DCC with bogofilter, add a little bit of mutt coloring for X-Bogosity, X-DCC, and my marks, and voila -- 5-fold decrease in spam review costs. (The corresponding procmail recipes went out to the bogofilter list.) Mutt -- it makes spam look different (TM)! -- Cheers, Alexy Khrabrov :: www.setup.org :: Age Quod Agis
Attachment:
whitelist.sendmail.sh
Description: Bourne shell script