John Hardin said of system-wide spamtrapping:
| I would say the overriding concern in implementing sitewide mail filtering
| is to *be conservative*.
|
| Don't ever send anything quietly to /dev/null.
|
| If you want to trap spam then modify trapped spam messages to add a header
| or change the subject (e.g. by prepending [SPAM]) and show your users how
| to filter such messages automatically themselves using their mail clients.
Don't even be so cocky as to mark it as spam. Tag it "[possible spam]" or
"[spam?]". I'm quite serious: the last thing you need is an irate user on
your neck because an important message from a relative or a boss came through
confidently labeled "[SPAM]".
Another idea may be to add a header line rather than gumming up the subject,
depending on how sophisticated the user's filtering is. (Some depend
entirely on scanning subject lines by eye, others don't like having the real
subject pushed out of the way.) Again, use tentative language rather than
bold, sweeping statements;
X-Spamtrapped: possibly UCE in the judgment of our filters
is going to be a lot less offensive on a false positive than
X-Spam-Warning: Do not read nor respond.