jdow wrote:
Color me smartassed but I want numbers not accusations. Can the
rhetoric and in bland neutral terms describe what you see in terms of
numbers, possible business relations, however loose, and so forth.
Here's some numbers to play with:
~500K messages delivered daily (as in, passed on to from Postfix to the
program that actually writes the message to the customer's mailbox tree
somewhere)
~16K of ~48K accounts have spam filtering enabled
Since Jan 1 2009, hits on HABEAS* rules have resulted in an average of:
rulename | spamperday | hamperday
-----------------------+------------------------+-----------------------
HABEAS_ACCREDITED_COI | 0.04154302670623145401 | 161.4124629080118694
HABEAS_ACCREDITED_SOI | 6.4124629080118694 | 3887.0326409495548961
(I run a daily script to stuff yesterday's SA log data into a database;
so far I haven't gotten around to doing anything with the data.)
I can't attest to the accuracy of any of the hits because this is an ISP
mail system. But even considering only a third of the accounts have
filtering enabled, that's still somewhere in the neighbourhood of 1% of
all mail hitting HABEAS_ACCREDITED_*.
Checking the spam reporting account shows no actual spams reported with
HABEAS hits, and one legitimate book fair travel ad from a publishing
company hitting _SOI; about 8500 messages have been reported and
confirmed. A further ~350 have been reported, but considered legit.
Admittedly, I have to consider a broader range of mail to be
"legitimate"... but I really haven't had to strain very hard in making
that distinction in hand-confirming messages reported as spam.
Checking my own personal account on my own server shows a newsletter for
a rewards program with my bank, occasional messages from eBay, and a
message from Adobe. All legitimate. I don't keep spam around all that
long, but what's still sticking around doesn't show any HABEAS* hits.
-kgd
