Standardized bounce format (was Re: [Mimedefang] DNS Lookups in MD - Was RBL and DNS lookups)
David F. Skoll
dfs at roaringpenguin.com
Sat May 12 16:57:08 EDT 2007
Richard Laager wrote:
>> (In fact, thanks to DJB, we have code in our commercial CanIt product
>> *specifically* to deal with the brain-dead qmail bounces... thanks a
>> bunch.)
> Really, what sort of parsing are you doing?
We extract tokens for Bayesian analysis. The qmail bounce message
just includes the entire original MIME-encoded message as a big glob
of plain text. If you try to tokenize that, you get garbage because
you get MIME boundaries, fragments of Base64-encoding, etc.
So we have a chunk of code that looks at the first line of the message:
if( defined($line) && ($line =~ /^Hi\.\s+This\s+is\s+the/) ) {
canit_log("Not tokenizing qmail bounce message for Bayes training");
return 0;
}
Regards,
David.
More information about the MIMEDefang
mailing list