Ned Slider wrote: > Some implementations, postgrey for example, keep a whitelist of > servers known to retry, so once a small number of mails (3 by > default iirc) have been successfully delivered from a given server > (or servers in the same /24 subnet), it is auto-whitelisted on the > basis that there is little point continually greylisting servers > that you *know* will retry anyway. > > This approach works extremely well and after a few weeks normal > usage very few legitimate mails are delayed by greylisting.
My grey time is 35 days, which is effectively the same thing. By greylisting only Windows desktops, I can ensure emails sent during a conference call are received without delay, during a support call, etc. (though the support address is configured to bypass greylisting during the work day). A server's first connection matters! That said, it would be nice to revoke the grey time for a server that didn't come back. That might even get me to make the grey time bigger, though the larger the database, the longer it takes to search (which must be a larger problem with postgrey with all that non-expiring data...).
