It seems to me that people legitimately working with Viagra, etc must know that there is an issue with spam and have already worked up solutions
I don't work on Viagra but I did miss my very important invitation to the XXX International Conference on Computer Science (i.e., 30th conference). My argument was the danger of content filtering (call it AI if you want), and that it can bite completely unexpectedly (so whitelists are of no use).
Someone sends in an email
Server farm attempts delivery from ip a.b.c.d
Hits greylist.
Server reattempts delivery from j.k.l.m
Hits greylist
Server reattempts delivery from w.x.y.z
The day when I see a server behaving in this way, greylisting is dead.
However, my strong assumption is that server farms allocate IPs mostly by subets, not completely random.
That's 181 netblocks or 1.328.916 individual IPv4 addresses
This confirms my assumption! You see: my technique reduces a million of IPs to two hundred. Worst case is 82, which means that typically their messages would pass after some 40 attempts. Not great, but better than half million attempts as it is now.
Ideally, if the subnet mask is made configurable, such as oMessage.GreylistingMask = 24; then I might want to try /16 and see what happens. Probably this is indeed a case when a big provider needs a special treatment:
if IP/16 is
in a predefined list of 82 known subnets
then oMessage.GreylistingMask = 16
else oMessage.GreylistingMask = 24. Still safe for less monstrous servers and
almost solves the problem with Outlook.
What a world. People really have too much time to waste.
(I can't agree more!)
And still please let me remind you that this thread is not about each of us's proven recipes of how best to fight spam (we will not solve this problem here). It's about making greylisting in HMS a bit more configurable, for whatever it's worth.