Re: wikipedia vandalism

[Roger Dingledine <arma@xxxxxxx>]

On Mon, Jan 24, 2005 at 06:04:52AM -0500, Roger Dingledine wrote:
> One approach that's in the middle ground would be to require logins,
> including captchas, and track edits by accounts like you do now. If

All in all, I think Wikipedia's two-tiered service is a plausible and
fine response to the situation, at least while they're not planning to do
any serious changes in their assumptions. Tor users can _read_ just fine,
which is what most visitors to the site do.

Brandon Wiley describes another option to this whole mess in his blog
at http://blanu.net/weblog.html#71
In situations where the service wants to operate with *only* IP
authentication, we can allow banning abusive IPs proportional to the
amount of abuse they produce. If it turns out that the trolls are not
particularly persistent with respect to trolling specifically through Tor,
then we can end up in a better situation than having the whole network
just preemptively banned. Here are a few snippets:



A better solution is temporary random exponential time bans. The key idea
here is that whenever a troll is found, the IP of the troll is banned
temporarily. Each time a ban occurs, the length of the next bans is
increased a random but approximately exponentially larger amount. After
an IP has gone without being banned for a while, the length of the next
future ban is decreased following the same curve until it reaches the
minimum amount.

This approach has a number of advantages over blacklisting all tor
nodes. First of all, it is not specific to tor. This method can be used
on all IPs. It is also automatic. No one needs to manually decide on
the interval that a ban should be. They only need to add IPs to the
ban list. They should also expire automatically. It only punishes
IPs which have actually been used for trolling, instead of punishing
innocent IPs which have committed no offense. Additionally, no IP is
ever beyond redemption. It will always be removed from the blacklist
eventually. In the best case, where no trolls use tor, everything is
eventually unblocked. In the worst case, where trolls use every tor node
to troll as much as they can, every tor node is blocked, which is no worse
than the existing policy of preemptively blocking all tor nodes and better
than that case because if all tor nodes are blocked and the trolls really
want to troll and not attack tor, they will move on to something else.