I run a small forum that's been around for nearly 10 years now and I've found that over the years when we do get someone that's absolutely insane or trollish the best thing to do isn't to outright ban them, but to give them an area away from the rest of the group where they can spout off. This is similar to the advice in the article, I suppose.
The implementation of which was creating a 'Purgatory' forum where there was absolutely no censorship and if people crossed the line too often they were restricted to that forum (though anybody else who wanted to actually engage them could also post). It pissed them off, but in almost all cases, instead of registering new accounts and flooding the main forum with posts (as often happens with banned trolls), they just spout off for a couple days at most, and then either calm down or fuck off of their own accord.
Stacy Horn's book Cyberville has an interesting discussion about using that approach on ECHO in the late-80s/early-90s as well, though in that case it was also tied up with a very anti-banning ideology (the cyberspace-as-unfettered-discussion idea that was popular in WELL-influenced circles). They experimented with various versions of totally-unrestricted forums, giving people their own forums for weird topics, stronger and weaker on-topic enforcement, etc. In a few cases they seem to have eventually decided to ban people, despite the initial reluctance, because some of the diversion measures just didn't work. Not sure how well it generalizes, but it's an interesting read for the $0.01 price of a used copy.
>a very anti-banning ideology (the cyberspace-as-unfettered-discussion idea that was popular in WELL-influenced circles)
The WELL might have had an anti-banning ideology, but it also had an ideology that everything a person writes must be written "under eir real name", with the dual consequences of much-higher reputational consequences for bad behavior and everyone's realizing that if the administration ever did decide to ban a user, getting around the ban would be a lot more involved than just using Tor, a TCP relay or getting a new IP or email address.
(Unimportant details: The real names of most participants were probably verified when they paid the monthly fee by check or credit card. By "under eir real name" I mean that every post had a Unix user name attached, and there was a policy that every user's ~/.plan file had to contain eir real name.)
True, though one problem is there are a non-zero number of people quite willing to having their real name attached to their conspiracy-theory or racism-mongering stuff. They might even prefer it, because their rl occupation is selling their conspiracy-theory books or otherwise hawking their conspiracy-theory website/organization/etc. It gets rid of the kids trolling for laughs, but you still end up with the hardcore people who are 100% committed to their odd views as a lifestyle, and have to decide what (if anything) to do about them.
The BBSs of the 80s had the concept of the Twit Bit, which, if set for a user, they were the only ones who would see their posts. It was if everyone else were simply ignoring their posts.
Another approach was selective downtime: have the modem accidentally drop the connection when the problem user dials in, say, 50% of the time, in hopes that they'll get too frustrated to keep connecting, but without getting angry enough to plot revenge (because hopefully they'll think it's just an unreliable system).
Unfortunately, some of them are just too persitent, and come back. It's like whack a mole: different IPs all over a geographic region. It's hard banning library IPs, because you shut yourself off to a large group. They may keep coming back, but often they're very easy to spot :)
You could also display another version to the troll than to other users where the troll sees the forum as if his posts are successfully submitted while the rest doesn't see the postings.
They were later removed because they cause difficulties in performance optimization. They're hard to implement consistently and correctly (i.e, so things like post counts for each thread etc add up correctly for the hellbanned user) without blowing performance out of the water for the general case.
I never said there weren't outliers, but I have found them to be quite rare since using this strategy. If I do have to ban someone like that I ban them by email address, since I require email confirmation to sign up. It slows them down at least. Also, I can ban by regex on an email address so they can't use the gmail trick of adding a + after their username.
I find it just as effective as an IP ban and less likely to affect others.
This is somewhat offtopic but the plus is actually a defacto standard in most mailers. I believe but can't say for certain that sendmail was the first to ignore everything after the plus in the local-part of the address and most other mailers do the same.
You can also put dots at arbitrary places in your gmail username i.e. jhonsmith@gmail com, jhonsmith@gmail com and jhon.s.m.i.t.h@gmail.com is the same email address.
The implementation of which was creating a 'Purgatory' forum where there was absolutely no censorship and if people crossed the line too often they were restricted to that forum (though anybody else who wanted to actually engage them could also post). It pissed them off, but in almost all cases, instead of registering new accounts and flooding the main forum with posts (as often happens with banned trolls), they just spout off for a couple days at most, and then either calm down or fuck off of their own accord.