[Mimedefang] Re: A place to skip tests (Questions on filter_recipient with multiple recip's)

[David F. Skoll]

Hi, Kelson.

> IIRC, filter_relay, filter_recipient, and filter_sender are all run before
> filter_begin, and filter_recipient runs as the relay names each recipient,
> not after they've all been listed, correct?  So filter_recipient has no
> idea how many recipients there are.

That's correct.

> So: If I get a message sent to two people, one of whom I want to skip
> filtering, I need to use stream_by_recipient.  However, that can't be
> called until *after* filter_recipient has run.  This means that the
> original message with both recipients will trip the logic in
> filter_recipient, and filtering will stop before it even gets to
> stream_by_recipient.

Right.

> Is there someplace else where something similar to
> ACCEPT_AND_NO_MORE_FILTERING can be used?

Not really.  All you can do is have your filter_begin, filter,
filter_multipart and filter_end routines do nothing if there is a
single recipient who is opted-out of scanning.

Returning the equivalent of ACCEPT_AND_NO_MORE_FILTERING from filter_begin()
would not save significant CPU load, because by then, the MIME message has
already been parsed and split into parts.

Regards,

David.