> -----Original Message-----
> From: Reijo Pitkanen
> Sent: Wednesday, May 28, 2003 12:07 AM
>
> A few ideas, below...
>
>
> I'm curious as to how many of these messages are inline-mime images, and how
> much this has to do with mucked-up mime bounds detection out of spamassassin
> (if it even does so, I'm not quite sure it does).   Granted, 2000 bytes is
> still pretty small when you come down to it, but it might we worth seeing
> how much time you save truncating encoded images -- mime-decode the message,
> truncate the weighty bits, re-assemble...

This would probably be the way to go. Until SA gets into the image analysis
business at least. <g> Of course, following this path, we only use the excerpt
to make the determination as to spam. Note, we're not using SA as a true filter
on this pass. Another pass would have to be made with SA in regular filter
mode,
if we're interested in ending up with the spam encapsulated in an attachment
with
the spam report added. This is true of the excerpt method that I was using as
well.

Slightly related factoid, I receive probably 250 spam messages a day, which SA
dutifully (and quite accurately) files away. I receive a lot of mail from
various
technical and discussion lists which I auto-file and never pass through SA. I
receive
at most 20 to 50 real live messages to me from friends and associates. I'd
place the
ratio of spam to ham at about 5:1. A sad commentary. In any event, with ratios
like
these, it's well worth the effrot and the CPU cycles to err in favor of
trapping
all the spam possible.

Back to main line, what does SA do at present with non-text (here, I'd say that
text and html are probably the only ones that qualify as text) attachments?
Does
it scan all of them (including gif's/jpeg's/zip files), or simply note their
existence
and exclude them from the rest of the scan? Maybe if it did that, it would
lower
CPU demand noticeably?

>
> Or I guess you could even just parse out the first X bytes of a mime-section
> and feed that into SA...

An interesting area of research. So many things to try, so little time. <g>




-------------------------------------------------------
This SF.net email is sponsored by: ObjectStore.
If flattening out C++ or Java code to make your application fit in a
relational database is painful, don't do it! Check out ObjectStore.
Now part of Progress Software. http://www.objectstore.net/sourceforge
_______________________________________________
Spamassassin-talk mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/spamassassin-talk

Reply via email to