[Tutor] A NOVEL IDEA- no more spam!

[Allyn Weaks]

On 15/9/02, Kirk Bailey wrote:
>OK, recently I saw a site with an intresting idea. SPAM PROOFING.
>
>Idea is, it maintains a list of people who may send email to you. Anyone else
>sends in a email, they get one back, explaining it is in storage for 7 days,

It's not all that novel; there are several commercial services (sic)
that do this. In practice, it turns out to be a bad idea, except
perhaps as a programming exercise.

It will prevent you from getting a variety of real mail, such as
receipts and shipping notices when you buy things from web sites.  It
will stop you from being able to register at many web sites, or to buy
software on line when they email you the registration key.  And I'm far
from the only person who refuses to jump through hoops such as this.
Ah, you say, you can periodically check the rejected mail to make sure
you aren't missing anything good!  At which point, why bother with it?
Use a simple set of mail client filters or procmail and you're better
off--same number of spam subject lines to scan for false positives, and
you won't have confused or ticked off any real people.

This scheme is also guaranteed to keep you off of a lot of mailing
lists, and you may never know why, because no one can tell you without
jumping through hoops.  The list server won't be able to send you a
confirmation request.  If you do manage to zubscribe somehow, it's
downright rude to send such messages to the people who post to the
list, and just as bad to direct them to the listowner.  You've already
explicitly agreed to accept list mail by zubscribing.  As a listowner,
I'd never allow a member to punish contributers that way.  This topic
has come up on several lists for listowners in the last couple of
weeks, and the opinion has been unanimous against the technique.  It's
considered so annoying that there are discussions for the best way to
preemptively block anyone using such things before they manage to
harass anyone.  Not long ago someone on this list was using this
scheme, and maybe still is (I guess I'll find out when this message
goes out).  The 'jump through hoops' message I got back from a post I
sent was just the nudge I needed to auto-reject anyone using the
servers that specialize in this.

>Almost all spam is robotic, this process will never happen. The owner of such
>email account is spamproof.

Possibly, but the 'jump through hoops' message sent out to legit
correspondents is even more annoying than spam is.  Dealing with your
spam directly is a nuisance, but missing out on real mail is the pits.
Do you expect a prospective employer to jump through hoops to send you
a job offer?  What if your great-uncle gets confused about the process,
and you miss an invitation to a family reunion?  What if your wife
who's out of town has to send you an urgent message, but can only do it
from a borrowed account before she catches a plane?

The bayesian probability algorithm has a *much* better chance of being
successful at blocking spam with few or even no false positives.
There's been a lot of discussion about it all over the net in the last
month including on c.l.p.  There are active projects going on in nearly
all languages.  One in python (there may be more):

<http://sourceforge.net/projects/spambayes/>
-- 
Allyn Weaks    allyn@tardigrade.net   Seattle, WA  Sunset zone 5
Pacific NW Native Wildlife Gardening: http://www.tardigrade.org/natives/
"The benefit of even limited monopolies is too doubtful, to be opposed
to that of their general suppression."  Thomas Jefferson