> Third and more generally, many anti-spam systems do things that
> violate the letter and/or intent of the email specs in the hope
> of efficiently disposing of more messages that might be spam.
> One of those things is sometimes to avoid sending out
> non-delivery messages at all. That is more or less the other
> end the process in which a system that receives such messages
> discards them, as in the above. All I can say about those
> approaches is that they involve tradeoffs and one should be
> careful what one wishes for.
As an example:
Although it isn't specifically part of our anti-spam system,
we sort of do this. Due to concerns about backscatter, if we accept a
message that we wind up believing is spam and delivery of that message
fails, we deliberately don't generate a bounce message[*]. Things not
scored as spam get normal bounce messages if they fail delivery.
Some people will say 'just don't accept such messages in the first
place'. That's certainly the ideal situation, but unfortunately this is
not completely possible in our environment for various reasons.
We're fully aware that this violates the letter and probably the
intent of the email specs. It's a tradeoff that we're willing to
live with.
- cks
[*: technically we generate a specially marked bounce message and
then have Exim immediately discard it via a custom router.
]
