Kee Hinckley <nazgul(_at_)somewhere(_dot_)com> wrote:
If an ISP is not filtering for spam, they can't even answer the
question, "how much spam do you get".
Which is why I proposed asking the question "how many emails did you
mark as spam." My original post did not make say that, but I tried to
clarify my position in the discussion which followed.
If they are filtering for spam
and 20% gets through, they their numbers will be 20% off. How are
you--looking at the aggregate information--going to figure that out?
You don't. You can't. You have no way of knowing how accurate
their marking system is. But you DO know that they've marked some
number of messages.
I did not say do nothing. I simply pointed out that you should
realize that there is going to be a lot of error in some of those
numbers.
Not exactly. The questions which are asked should be such that any
answers will contain little measurement error. They may contain
system errors, but since we have no way of quantifying them, there's
little we can do to deal with them.
However, over large populations, we can assume that the systematic
errors follow some standard distribution (e.g. gaussian). We can then
correct for their effects. The important thing is that we can reduce
unknown systematic errors from 1000's of domains to a *model* of
systematic errors with a few well understood parameters (distribution,
mean, variance)
Alan DeKok.
_______________________________________________
Asrg mailing list
Asrg(_at_)ietf(_dot_)org
https://www1.ietf.org/mailman/listinfo/asrg