In the A2K playground: Play the Synbot game today!

Aah, yes, see, thats what we did! We looked up our respective IP addresses on the net - I dunno if it was whatismyip.com, probably, or some site like that.

OK, so then it all makes sense. Kinda.

I feel like such a Luddite right now.. :-)

Actually it's usually much worse than that.

In reality many times we'd need to ban the whole ISP to get rid of one user. For example AOL users often share IPs and for any one AOL IP there are often over 20 legitimate members here who use it. So any attempt to IP ban an AOL user would block many other people around the world who had nothing to do with it.

What makes it a bit pointless is that the user can just use a proxy and circumvent the ban in seconds while the rest of the good users are locked out.

IP banning is a tool we use rarely for this reason and it's usually reserved for computer security (i.e. dealing with robots who are trying to bring down the site).

On small sites you can get away with it, because you don't have enough traffic to run into the false positives you banned, but on larger sites this is problematic.

I edit wikipedia very frequently, and often find that I need to log in and use an account because the IP I'm on was banned for a previous user's spam or vandalism. Their bans are "soft" and you can just create an account to circumvent it because of this limitation.

What it all comes down to is this: the internet is based on fundamentally stateless protocols. It was not designed to even allow a users to log into anything and cookies were just added later to provide user state.

All methods of deriving state are optional to the user, so any attempt to get picky about identity requires guesswork and can be circumvented just as easily as it was put in place.

We'll still use it on occasion, but to think that it's much more than a temporary inconvenience on the banned user is to put way too much faith in it.

By the way nimh, I was wrong and after more investigation I now think your list is correct even though his posts are coming from many different IPs and locations.

Banning all of AOL might not be such a bad idea....

Interesting.

I could do that, if I wanted to misuse my access to our clients' networks.

Another possibility is that he/she/it is using zombie computers infected with a trojan.

Another, I suppose, is that this individual travels widely.

The first two scenarios would indicate a certain sophistication and/or dedication.

Derf. Forgot the most obvious scenario, which is using a proxy or anonymizer.

Just thinking out loud…

Perhaps you could add a line to your algorithm that compares the current IP to past troll IPs and in the case of a match; alters the criteria for collapsing posts? For instance: In the case of a match, multiply neg by 2.5... so with the threshold set at 5, that would result in the auto collapse taking place at -2 for suspected trolls.

If that’s possible; it probably wouldn't be too terribly difficult to except AOL and other known dynamic IP ranges from the process, if you deemed it necessary. This would result in expediting undesirable post collapses.

I think it might also be a good idea to weight users input on a sliding scale that reduces the effect of voting by those with lower reputations. This could dramatically reduce the effect of trolls acts of retribution and their ability to prop up each other. Ultimately; he who earns a good reputation would have greater say over he who earns a bad one.

Personally, I think it would be kinda fun to build the arguments to get you closer to the "Holy Grail of algorithms”. I don't speak your language, but I imagine the arguments could be tested and tinkered with in Excel until the formula returned the desired results. I'm guessing the favorite or friend feature will probably be required as a final step to eliminate false positives.

Once you have a high degree of confidence in the algorithm; the effect of banning could be accomplished with a system wide collapse that eliminates the troll’s posts from everyone’s view… except the troll. This would make it difficult for the troll to distinguish between automated administrative action, and people simply choosing to ignore him.

Perhaps another line could be added that enforces a waiting period for suspicious IP addresses. Or a probationary period that further restricts new user activity, strengthened by negs and suspicious IPs.

Just talking out loud in brainstorm fashion on the off chance it might be useful... I’m not trying to play 20 questions with your security or give the trolls a head start by getting you to disclose it.

Cool. An instructive illustration about the limited use of IP bans, plus vindication of my list

All interesting stuff, btw.

RG - I would have felt much more comforted had you simply said this to begin with. I have more than ample amounts of tolerance for working with a limited tool set. I just wanted to know that the Mods care to address the problem at all.

Thanks
K
O