Re: Cypherpunks Goals: Bad debate drives out good debate

[Derek Upham <upham@cs.ubc.ca>]

> Cypher version of Gresham's Law: bad posts drive out good posts.

> (The same is being seen in talk.politics.crypto, with the neverending
> Sternlight vs. Everybody Else dominating the traffic by a factor of
> 20-to-1. Detweiler recently reappeared (as tmp@netcom.com) and is back
> to debating _himself_ and answering his own delusional posts.)

Let's face it: Usenet is inherently broken.

There are two ways to filter for content: filter at the newsgroup
source via moderators, or filter at the newsgroup destinations via
killfiles.  Moderating does seem to get rid of most of the cruft, but
the moderators are required to read every post that comes through,
and, worse, make judgements with some degree of impartiality (which is
not always possible).  On the other side, kill files are useful, but
only to a certain degree.  Blocking sertain key words in subjects
doesn't help against topic drift; blocking Sternlight won't help you
in the slightest when umpteen other people respond to his posts.

The situation for most mailing lists is just as bad.  Anyone can send
a post to a mailing list, and there are usually no moderators to
enforce content.  A reader's only recourse is another kill file---and
most mail killing facilities are pretty lousy when compared with their
Usenet counterparts.  The only thing mailing lists have going for them
is that they tend to be less obvious than newsgroups.  Harder to find.

To fix them problem, then, we either have to either improve the kill
files or improve the moderation.  Personalized AI filters (see Moran's
"The Long Run") will be spiffy when they arrive, but they're not going
to be arriving any time soon.  This leaves moderation.  If we increase
the number of moderators, we can reduce the load on each and take into
account as many tastes as possible.  In the best case, the moderators
would consist of all the readers of the newsgroup.

How would this work?

Assume we're running a mailing list (Usenet v.2.0 will be just a
special case of a mailing list).  All posts are sent to the central
site.  The mailing list software picks one e-mail address from all of
the list receivers, and forwards the post to that e-mail address
(keeping the original post on file).  The forwarded post will have a
subject line something like:

  Subject: Post ACK, list cypherpunks, msg #435A77CF

with the post contained in the body.  The receiver reads the post,
then replies to it.  The reply subject line will be:

  Subject: Re: Post ACK, list cypherpunks, msg #435A77CF

and the body will contain "ACK" or "NAK" or "post" or "dump" or
whatever.  This goes back to the list maintainer, who can check that:

  * The message in question is in fact outstanding.
  * The person who sent back the evaluation was the one who was
    supposed to.
  * Other sordid details.

If the message was approved, it goes to everyone.  Otherwise, the
original poster is informed that the post did not meet standards.

We might even want to forward the evaluation body to the original
poster; this would allow the evalutors to send comments explaining why
the post was rejected.

The nice thing about this technique is that the more people a person
pisses off, the less likely it is that his or her posts will ever see
the light of day.  Even better, aware readers can nip MAKE.MONEY.FAST
and Green Card Lottery posts in the bud.

Something very similar to this exists now in the Internet Oracle, so
distributed moderation ought to be possible.  I suspect that the mail-
handling features could be incorporated entirely into "procmail" and
"SmartList" (a filter program and a mailing list program), although
the databases would need C maintenance programs for efficiency.

I'll go to bed and await comments.

Derek