Hi,
On Thu, 11 Mar 2004 18:15:05 -0800 "jdow" <jdow@earthlink.net> wrote:
> From: "Bob Apthorpe" <apthorpe+sa@cynistar.net>
> > "Your system, your rules" but IMO anyone who scores HABEAS_SWE as positive
> > is a moron and anyone who advocates that stupidity to others shouldn't be
> > trusted to operate a mailserver. Set the score to zero and get on with
> > your lives; I'm tired of hearing about it.
>
> I guess I am a moron, Bob. I go with what I see. I do not see any evidence
> outside of their website that they exist, win lawsuits, deter spammers, and
> so forth.
Here's a technical argument for you then.
The Habeas mark is not widely used but up until, what, mid-January? The
vast majority of this list's readership had only seen the Habeas mark on
ham, if they had seen it at all. So at the time SA 2.60 came out, ham
very occasionally had the Habeas mark and spam virtually never did. That
is, up until that point the HABEAS_SWE test was a pretty good one for
detecting ham.
Then around 1/10/2004 some ballsy dirtbag(s) started spamming using
Habeas marks to pick up some negative points in SA. There was no mad
rush among legitimate mailers to remove the Habeas mark from their mail
since a) many of them paid for it, and b) up to that point there was no
reason to expect it on spam based on observed spam. So from 1/10
onwards, the Habeas mark was no longer as good a sign of ham since it
was appearing in both ham and spam.
Q: What does the GA[1] do to tests that match equal amounts of ham and
spam? Hint: Look at the score for BAYES_50.
If the HABEAS_SWE test is hitting both spam and ham, it's no longer a
good indicator and should be zeroed out. It gives you no useful
information in telling ham from spam and setting it to anything but zero
will degrade SA's accuracy.
That's the technical argument for setting HABEAS_SWE's score to zero, if
you feel the need to change it at all.
> If I see objective evidence that they have won some lawsuits thoroughly
> enough that spammers were seriously financially injured by the win, $1e6
> or more, then I might start believing them. It'd take something that big
> to deter the bastards.
I take it you've never spent time around lawyers. I shared an office
with one for a few years and I learned a little about how they operate.
They're normal people, which is to say they're lazy. Their
clients/employers are normal too, which is to say they're cheap. If what
you want is for a spammer to stop diluting the value of your trademark
or to stop illegaly copying your copyrighted poem, you find them and
tell them to stop in increasingly more formal and expensive ways until
either they stop or you decide that it's no longer worth your money to
pursue the issue.
Usually this starts with a simple phone call, maybe a few letters
through registered mail, and (rarely) with the filing of a civil suit.
Usually once the suit has been filed, the defendent figures out that
you're serious and settles. You as a member of the unwashed general
public will not see that, but each time a would-be defendent settles,
the plaintiff's lawyers 'win.' The got what they wanted without spending
a lot of money or time trying to find parking near the courthouse. Both
sides will often declare victory after a settlement, and in a weird way,
they're often both right, but not for the reasons they state in public.
Pursuing a suit to conclusion is often a losing battle for the plaintiff
due to a lovely two word phrase: "judgment proof." This means that the
although the plaintiff spent $150k winning a case where the judge
awarded them $1M in damages (plus court costs!), it turns out the
defendent's assets consist of an El Camino on blocks, a can of Skoal,
and a half-eaten 9-pc. bucket of The Colonel's finest (the double-wide
is a rental.) It doesn't matter what you win, it matters what you can
collect, and if you can't collect more than you've spent (often the
case) then you effectively lose. And since it takes a lot of time and
money to pursue a case to the end, and there may be no way to tell how a
judge or jury will react, it's often in both parties' interest to settle
(judges and juries are lazy too.)
For that reason it's unlikely you'll ever see any big rewards. And from
an IANAL legal perspective, Habeas is on strong legal footing so it's
doubtful that you'll see much court action beyond filing suits and
having spammers settle. The weaknesses are that if enough spammers start
abusing the mark, it takes a prohibitive amount of work to track them
all down and file suits, especially if they're outside the US.
Conveniently, most spammers are in the US (see
http://www.spamhaus.org/rokso/index.lasso) and there are only a few of
them abusing the Habeas mark, otherwise we'd see it on a lot more spam.
Regardless, if HABEAS_SWE is causing you problems, the best thing to do
is set the test's score to zero, forward the spam to Habeas as evidence
of abuse, and get back to the PS2. Complaining about Habeas, criticizing
their business model, giving technically unsound advice on scores, etc.
does nobody here any good because nobody here can fix any of that. That
is the reason I'm tired of this topic and why I'm wasting so much time
trying to kill it.
Sorry for the length; most people are blissfully unaware of the workings
of the US legal system (myself included) - corrections and elaborations
by real lawyers & paralegals are appreciated and will not be taken as
legal advice, yadda, yadda...
-- Bob
[1] Genetic algorithm - that which automagically sets most of SA's rule
scores. Since replaced by a perceptron-based system that's as accurate
but much much faster.