Hello, Todd, The short answer to this question is I have "per domain" "hold weights" ranging from as low as 5 and as high as 13.
The long answer is as follows... We have around 100 E-Mail Sites on our IMail Server. We are using "per domain" filtering and have currently set up DJM Pro for 5 sites. The hold weight for each site is mostly dependent upon 2 factors. These 2 factors are "importance of any delays in receiving e-mail" (if we have to release something that was blocked as a false positive) Vs. "importance of eliminating as much spam as possible". If "delays" are more important to you the "hold weight" will probably be higher. If elimnating as much spam as possible is more important then the "hold weight" will be lower. For our in-house domains it is very important that we receive every legitimate e-mail so our current "hold weight" is 13. This lets in more spam than I would necessarily like but I've never seen a piece of legit e-mail over that threshold. 13 is currently the highest "hold weight" we have in play on any of our domains. On the other hand we have one customer who is being inundated with spam and has asked us to drop the "hold weight" as low as I think I can reasonably get it. This client has a "hold weight" of 5. There have been a few pieces of legit e-mail which came through higher than that but I'll usually catch those within a day and release them and will also adjust the filters to make sure e-mail from that domain name has enough subtracted from the weight to make sure it passes through the next time. 5 is currently the lowest "hold weight" we have in play on any of our domains. We are using all of the default weighting which comes with DJM when you first download it. My current plan with any new domains I'll be setting up is to do a COPYTO for 2 days on that domain to a temp e-mail address that I have set up. For each domain that I am currently analyzing, every e-mail I receive at that address I route in to one of 3 folders, "Not Spam", "Maybe (Not) Spam)" and "Spam". Each folder represent a striation of the weighting scale with no overlap. The weight range for each folder might look something like "Not Spam" -> -21 to 05, "Maybe (Not) Spam" -> 06 to 09 and "Spam" -> 10 and Up. I analyze all of the e-mail sent to the temp address and see if I can route it to these folders. Messages which are definitely not spam go into the "Not Spam" folder, except if it's weight is higher than my current highest weight for the "Spam" folder, in this example 5. If higher then it goes into "Maybe (Not) Spam". Messages which are definitely spam go into the "Spam" folder, except if it's weight is lower than my lowest current weight for the "Not Spam" folder, in this example 10. If lower then it goes into "Maybe (Not) Spam". If I'm lucky and I've guessed right then I'll have very few e-mail messages in the "Maybe (Not) Spam" folder. If I'm not lucky then there'll be lots of messages in the "Maybe (Not) Spam" folder and I'll have to analyze them closely and try to adjust the lowest weight of my "Spam" folder down. By analyzing the "Maybe (Not) Spam" folder I can try and lower my lowest "Spam" weight and eventually come upon a good "hold weight". Anything left in the "Maybe (Not) Spam" folder has the option to have an exception made for it somewhere in the DJM settings. I realize the way I'm doing it intially is very convuluted but I think there might be a seed of a good algorithm in there if I could just express it elegantly. After I've established an initial good "hold weight" for a domain, then I continue to analyze the types of e-mail that are making it under the "hold weight". I then have the option of lowering the "hold weight" with further analysis (and make exceptions) for any False Positives which might be caught by the new lowered "hold weight" or by leaving the "hold weight" as is and with further analysis to try and make exceptions and add weight on to any real Spam which are making it under the "hold weight". Generally speaking there might be too many False Positives created by lowering the "hold weight" further or there might be too many exceptions for real Spam which need to be made by leaving the "hold weight" as is but I can usually find one or the other which requires less maintenance on my part. That was a long answer to a very short question but I though I'd float some of my methods out there for further feedback. Take Care, Dan ----- Original Message ----- From: <[EMAIL PROTECTED]> To: <[EMAIL PROTECTED]> Sent: Monday, July 07, 2003 5:07 PM Subject: [Declude.JunkMail] HOLD Weight > I was wondering what everyone that HOLDs by Weight is using for a value to > hold at. > > We only HOLD by Weight, not buy individual test, currently we are set to > HOLD at 12. > > The config files are mostly default with a couple minor changes, I think I > bumped the value for Badheaders just bit. > > Todd > > > > > > --- > [This E-mail was scanned for viruses by Declude Virus (http://www.declude.com)] > > --- > This E-mail came from the Declude.JunkMail mailing list. To > unsubscribe, just send an E-mail to [EMAIL PROTECTED], and > type "unsubscribe Declude.JunkMail". The archives can be found > at http://www.mail-archive.com. > ==================================================================== > This E-mail is scanned and free from viruses. www.nexustechgroup.com > > ==================================================================== This E-mail is scanned and free from viruses. www.nexustechgroup.com --- [This E-mail was scanned for viruses by Declude Virus (http://www.declude.com)] --- This E-mail came from the Declude.JunkMail mailing list. To unsubscribe, just send an E-mail to [EMAIL PROTECTED], and type "unsubscribe Declude.JunkMail". The archives can be found at http://www.mail-archive.com.
