R: R: R: [jdev] about spim techniques

[Sander Devrieze]

Op zaterdag 27 augustus 2005 13:15, schreef Roberto Della Pasqua:
> Hmmm...
>
> A whitelist of domains? Or users?
>
> So a central server has the list of ten millions domain, a sub server when
> a new domain come will ask the central server for the GOOD or BAD. If the
> central point will fail this can cause service break.
>
> Imho the white/blacklist need be full decentralized, perhaps based on trust
> ranking.

Just ranking will be also cool. But then ranking the authorities. E.g. if I am 
an authority that gives certificates to server owners, and it is known that 
servers with my certificate never send spim, my certificates will get a 
higher ranking. The reason for the high ranking can be that I only give 
certificates to people that I met in real-life, asked a copy of their 
identity card, asked their phone number, asked about their anti-spim 
policy,...

> Ear mine idea: user1 want contact user2. User2 don't has user1 in buddylist
> (server side), the xmpp query a central database (distributed in some
> points), if the user1 is spam, then don't forward.
>
> A new xmpp draft extension, permit to collect informations about user
> spammer. The list will be ordered and if a user reach a MAX level of spam
> it become signaled.
>
> So:
> 1) a company can contact another company without be blocked
> 2) the spammer is blocked

A 'mass spimmer' will probably set up his own server. On this server, he 
automatically creates hundreds of thousands accounts to send spam. In this 
way a user-based system will not work I guess. With the "certificates 
whitelist", we create an opt-in system that will increase the cost for 
professional 'mass spimmers' just enough to make it not profitable.

-- 
Mvg, Sander Devrieze.
xmpp:sander at devrieze.dyndns.org

ejabberd, the expandable Jabber daemon. --
http://ejabberd.jabber.ru/