Mark E. Mallett writes:
But really, a problem is that support for RCPT-TO (and other)
scripting probably needs more fleshing out anyway, since RCPT-TO
wants to see more states and more information returned from a script
than simply "accept" or "reject." To me, this suggests having a
separate scripts, or at least script fragments, for each mode.
Perhaps I'd better describe the specific idea that led me to this.
I want my code to send fewer bounces, as we all do, right? So I want to
reject using SMTP, not by sending a DSN/MDN.
It occured to me that if I provide a postffix content filter interface
to the sieve interpreter, then postfix can do SMTP rejections. The
sieve interpreter must run the user's regular active script, skip any
clauses that can't be evaluated (and don't contain stop), and see
whether a reject is executed. (In my code, this happens to be fairly
simple.) Done.
The only problem is that I cannot with good conscience execute a
possibly invalid action, can I?
Arnt