08-01-07 12:19 AM
> Thanks Sandy, but in this case it isn't the spam-filter.
Are you really not going to confess the spam filter's name? That isn't
too helpful.
A spam filter implemented as an IIS SMTP event sink can make changes to
message metadata that result in different outbound message properties.
You say you have copies of the messages after they pass through an event
sink-based spam filter but before they enter the outbound message stream.
Are you getting these from the \Drop folder? Those are the only messages
that can be said to *have* exited the event sink chain, but *not* have
entered an outbound queue. How do these compare to messages purposely
left in \Queue?
Or are you grabbing message bodies from the \Queue folder? If so, have
you compared with messages sent to a local domain (on the IIS SMTP server
itself) to confirm that messages in \Drop are not corrupt?
If you are getting your message copies from any other step in the process
(such as being archived by the spam filter itself), this is not a
trustworthy source. Just because software writes a clean copy to disk
doesn't mean that's the same copy that is handed back to IIS as a message
object. I'm not saying that your spam filter is the problem... but you
must eliminate it more scientifically.
If you can confirm that, from a single origin message sent to one local
and one remote address, the body in Queue is different than the one in
Drop, that would point to a bug outside of your spam filter and within IIS
SMTP (though not a known one). If the body in Queue and Drop are
identical, but with a packet sniffer you can see that the transmitted data
is different, that would point more to a a NIC, cabling, or other network
issue.
> We have sent several million messages this year as well without a problem
> until Windows 2003 SP2.
Yes, I'm talking about SP2 machines passing millions per *week*.
--Sandy
[ Post a follow-up to this message ]
|