Paul Smith <paul(_at_)pscs(_dot_)co(_dot_)uk> wrote:*
If a user set 'discard if x,y,z', and the sender doesn't support PRDR,
we have to either:
- discard the message silently (ugh)
- send a bounce message back (backscatter)
- quarantine the message (support issues, as above)
- fake PRDR by temp rejecting the message, then on the next retry only
accept one recipient at a time (double-ugh)
At the moment we just don't give the user an option to 'discard'.
Their bad messages get quarantined - but that's what they know will
happen, so at least it's deterministic, even if not exactly what the
user wants. Having a 'sometimes discard, sometimes quarantine' option
would be nasty from a user's PoV.
Paul, I want to be sure I understand your point.
In your implementation, there is no possibility of a disposition
conflict because the configuration options exposed to your users do not
permit it. To provide meaningful support for per-recipient data
responses, you would need to add a configuration option whose behavior
would be non-deterministic. Right?
My problem is that the disposition conflict already exists (arguably due
to Bad Choices by the author of the code ten years ago). Having the
ability to return per-recipient status to the sender-SMTP after the dot
would make the behavior of my filter less surprising.
This is distinct from the question of whether the current PRDR draft is
the correct solution, or whether something more LMTP-like would be
ietf-smtp mailing list