Neil Harris wrote:

Regarding Tor, does anyone have, or has anyone considered, an auto-discovery robot to find Tor proxies?

This would be a Tor client which would connect to Tor at regular intervals and hit a special URL with a magic authenticating token in it, that would automatically ban the IP in question.

Sooner or later, it would work its way through all, or almost all, of the proxies in the Tor cloud.

It's not necessary, Tor have a public exit node list. See for example http://serifos.eecs.harvard.edu:8000/cgi-bin/exit.pl . The Tor developers are actually very sympathetic to our situation... or at least they became sympathetic after a series of conversations between our developer Domas Mituzas and Tor developer Roger Dingledine, starting at the CCC last December.

My question to Roger at his CCC lecture was "are you going to provide us with a client library for automated blocking of Tor exit nodes?" to which his answer was no, but several months later we received this:

http://tor.eff.org/cvs/tor/contrib/exitlist

and the Tor developers even made plans to integrate it into MediaWiki for us. That hasn't eventuated, but I appreciate the gesture.

Roger's preferred solution in MediaWiki is to enable admins to make short-duration blocks (say 15 minutes) of all Tor exit nodes simultaneously. My preferred solution is to delay edits:

http://article.gmane.org/gmane.science.linguistics.wikipedia.technical/18932

...although that is quite a bit more complicated and thus less likely to get done. At least my proposal serves to highlight our differences in viewpoint. Tor supporters like to justify their existence from the moral high ground of protection against government persecution or industrial espionage. But what the bulk of Tor users are really interested in is obscuring their identity server administrators, and that carries with it a different set of ethical implications.

Administrators of wikis, forums, webmail and IRC all use IP blacklists as a means to enforce a code of behaviour. Roger counters that server administrators should move from IP-based access control to more secure identification methods such as PKA coupled with credit card authentication. But would that really be a step forward for privacy?

-- Tim Starling

On 9/27/05, Jimmy Wales jwales@wikia.com wrote:

What I recommend is that Tor resolve this problem in this way:

user -> tor cloud -> tor authentication -> tor trusted cloud -> website

If a website complains about a particular ip at a particular time, in the trusted cloud, then tor retains enough information to track it back to the authentication server account. They still have no clue who the original user is, but they can then use whatever methods they want to keep jerks off the trusted cloud -- and then we could treat the trusted cloud like any other dynamic ip range.

I have my doubts that Tor is going to want to get into the business of policing its user base for "appropriate conduct", or even get into the business of credentialling its users at all. Also, an abusive user could easily create a new identity whenever his old one gets banned; there'd be no way for Tor link the new identity to the old, banned one.

No matter how you slice it, if the default trust level allows edits, then we'll get vandalism; if the default trust level does not allow edits, we'll never learn whether any particular identity should be trusted. There's simply no way with Tor in the picture that we can reliably link an identity at the server end with a specific end user without issuing credentials via some outside channel that the user has to validate after they connect. As long as trustable credentials are available anonymously and automatically, vandals will continue to acquire and use them.

Kelly