Frank Scholz wrote:
First, there is no overall solution, that I've learned. ;-)
The question is to find the right balance between
effort and result - as always.
One way is to raise the hurdle. Only allow a closed
user group to comment or trackback. Authorize users
only on personal knowledge or certificate/pgp signature.
For the blog world imho an impracticable approach.
At least as long the UN citizen certificate key is
unavailable. :-(
Next thing is verifying the posting via captcha or
acknowledgment via email reply. Or tar-pitting
and greylisting. All this raises the effort @ the
bad guys side, but this will only work as long as
there are easier, more attractive targets.
So the only way I can think of is a way
similar how email spam is handled:
 - lexical analysis of the content
 - cookie based whitelisting
But the problems with the pattern list are,
that we need blog specific ones. There are
blogs which spin around viagra or poker.
If we don't do this, the false positive
rate will for some blogs way to high.
Therefore each blog has to be trained,
comments/trackbacks classified a spam
need to be quarantined and under certain
conditions reclassified.
For email spam prevention I use now quite some
time DSPAM,http://dspam.nuclearelephant.com/
with great success and very, very few false positives.
Even in very heterogeneous environments with
German/English mixed and normal users and
high traffic mailinglists.
There is(/was?) even a python wrapper around
libdspam, that I think I have somewhere in
my closet and can go looking for it.
  
I think you have made a very nice summary of the problem, Frank. But as to the solution you propose (libdspam, speciffic training of each blog, quarantining and reclassifying), it is too complicated for me to develop (I don't know as much Python as that), and besides I have no instance of Zope to change anything in the sources at the moment (my blog is hosted in a public hosting service). But maybe the developper's of Coreblog will do something...
As to me, I think I'll see if I can do something easier, captcha images or acknowledgement via email reply...
Btw.: Has anybody tried my implementation
of the spam cleaner within COREBlog? Any
feedback on this?
  
I can't, as I have no access to the Zope sources. I am using Luistxo's script, as I am in the same blog community as he is.

------------------------------------------------------
Igor Leturia Azkarate
Elhuyar I+G+B
Zelai Haundi kalea, 3
Osinalde industrialdea
20170 Usurbil
(+34) 943 36 30 40
[EMAIL PROTECTED] / www.elhuyar.org



_______________________________________________
COREblog-en mailing list
[email protected]
http://postaria.com/cgi-bin/mailman/listinfo/coreblog-en
Unsubscription writing to [EMAIL PROTECTED]

Reply via email to