> -----Original Message-----
> From: [EMAIL PROTECTED]
> [mailto:[EMAIL PROTECTED] Behalf Of
> Abigail Marshall
> Sent: Friday, August 22, 2003 4:35 PM
> To: Paul Adams
> Subject: Re: [SAtalk] big attachments taking too long to process
>
>
>
>
> Hello Paul,
>
> Wednesday, August 20, 2003, 8:37:30 PM, you wrote:
>
> PA> This has happened a few times: when a message is retrieved from the
> PA> upstream POP server with a large attachment (100K), my local copy of SA
> PA> takes 50 seconds or so to process it. I am using Procmail to filter
> PA> deliver the messages locally -- not passing them on to another MDA.
>
> 1. Limit the size of the files going to SA, via the procmail
> call:
>
>

The following thread may be of interest:
http://sourceforge.net/mailarchive/message.php?msg_id=5079393

In that test, I found that excerpting the first and last 7500 bytes of the
message,
for messages whose length was >= 15000 yielded an accurate determination of
spam
with no additional false negatives (ie, no additional spam tagged as ham).

Excerpting text from both the beginning and end of the message was superior to
truncating the message at the same total length.

The procmail rule might look like this:

# Filter small messages the regular way
:0fw:spamassassin.lock
* ! > 14999
| spamassassin

# Otherwise, just test an excerpt, and deliver spam
# directly into big-spam.mbox.
:0E:
? (head -c 7500; echo ""; tail -c 7500) | spamassassin -e
big-spam.mbox

Note above, that in the case that message is large, we run SA with the '-e'
switch, to simply test the excerpted message, and if it tests as spam, then
it is directly deposited in the big-spam.mbox. In this case, no spam report
is produced, and similarly all non-spam (ie, ham) messages of 15000 bytes or
greater length will not have an X-Spam-Status header, or any other SA headers.
That's because we're operating only on an excerpt of the entire message, and
we don't want that excerpt to become the result of a filter -- otherwise we
eliminate message content, which is needed in both the spam and ham cases.
With some more work, we can change the recipe to save the header info. that
resulted
when running the message excerpt through SA, and we can even add the spam
report back but it is rather complicated.




-------------------------------------------------------
This SF.net email is sponsored by: VM Ware
With VMware you can run multiple operating systems on a single machine.
WITHOUT REBOOTING! Mix Linux / Windows / Novell virtual machines
at the same time. Free trial click here:http://www.vmware.com/wl/offer/358/0
_______________________________________________
Spamassassin-talk mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/spamassassin-talk

Reply via email to