I took the last 10,000 non-mailing-list E-mails sent to us, and 2.5% would be blocked by your regex. That's selecting *our* data, not selecting data to fit my position. If you have better data, please share it.
The false positive problem is this, where "positive" is a "reject" :
positives = true positives + false positives
Your "data" (non-rejected messages) is not even part of the equation. Your data is the "negatives", ie, the non-rejects.
Please stick to facts Len. :)
I don't use "new math." 100% of E-mail to @declude.com is delivered. The E-mail that is marked as spam is reviewed and if there are any false positives, they get moved in with the "good mail." So those 10,000 E-mails include E-mails no matter how many spam tests they may have failed.
Of those 10,000 E-mails that people attempted to send us, 2.5% would have failed your test. That's a 2.5% false positive ratio.
Please, Len, if you aren't sure how I am coming up with something, either ask or guess, don't assume. Rather than making an accurate statement ("I think your data..." or "Assuming that your data..."), you made a completely false statement. I don't like being harsh, but you keep avoiding answering questions (such as the list of well-known ISPs you refer to that have been blocking non-vanity reverse DNS entries), and keep making false statements. We're all wrong occasionally (me too!), but you're really pushing the limits.
I consider false positives to be percentage of total rejects, not a % of total legit. I don't care about legit mail. It's not the problem.
That's not how anti-spam companies calculate FPs. If you have 9,990 spams and 10 legitimate E-mails, and you block all 10,000 E-mails, you don't have a .1% FP ratio. You have a 100% FP ratio.
Come on, that's BS and you know it.
Len, that IS how it works. I've been working with anti-spam software since 1997. The false positive ratio is determined by dividing the number of legitimate E-mails that were treated as spam and dividing by the total number of legitimate E-mails. That's just how the formula works.
IMGate's method of rejecting E-mail is very useful, I do not doubt that. But using that method, it is nearly impossible to determine the false positive ratio (since you don't know the amount of legitimate E-mail that is rejected).
I block 10,000 messages as positive rejects (eg, using the "DSL subscriber PTR hostname" filter. Of those, 10 are legit, "false positives".
Rate of the FP in Total P is 10/10000 = 0.001 = 0.1 %
Yes. But that's not the false positive ratio. That says ".1% of the mail I reject is legitimate", not "I only reject .1% of legitimate E-mail." Both are useful numbers.
Imagine a hospital that runs cancer tests on patients. They test 1,000 patients, 1 of whom really has cancer. The tests says none of the 1,000 users has cancer. According to you, they have a .1% FP ratio. According to any statistician, they have a 100% FP ratio.
and [2] Everyone else should *only* use the reverse DNS entry filtering *if* it is used as part of an overall spam system (IE only blocking E-mail as a result of that *and* other failures).
Compound, weighted conditions are marginally more accurate...
Your method, if used on our server, would block at least 2.5% of our legitimate E-mail. That's 1 out of 25 E-mails. We use a weighted system here, that has somewhere around 1 FP out of 1,000 legitimate E-mails. The difference between 2.5% and .1% is not marginal.
-Scott
---
Declude JunkMail: The advanced anti-spam solution for IMail mailservers.
Declude Virus: Catches known viruses and is the leader in mailserver vulnerability detection.
Find out what you have been missing: Ask for a free 30-day evaluation.
--- [This E-mail was scanned for viruses by Declude Virus (http://www.declude.com)]
To Unsubscribe: http://www.ipswitch.com/support/mailing-lists.html List Archive: http://www.mail-archive.com/imail_forum%40list.ipswitch.com/ Knowledge Base/FAQ: http://www.ipswitch.com/support/IMail/
