I understand your perspective. And I agree: fighting spam is essentially always a losing battle, at least at a technical level. It doesn’t mean it’s not worth fighting, but it does argue for a rational upper bound on effort expended.
One point that Merk always stressed (and which, frankly, I discounted before I saw it firsthand) is that the spam here does not come from bots. We do have simple questions in place for registration and they do keep the bots out. All the spammers that get through are humans that are determined to post here. We had one extreme case where a guy pretended to be a novice I7 user and, after several incoherent posts, wrote about his “game idea” that involved buying UK flowers.
The typical spammer signs up for an account but doesn’t activate it right away. (Presumably they are registering on other forums during this window.) I delete these accounts in bulk every few hours; the total comes to about 50 per day that get scrubbed before they ever hit the members list. To give you a sense of the scale, in the last 30 days we added 50 legitimate new members.
Maybe once or twice a day, we get an overachieving spammer who signs up, activates his account within minutes, and begins flooding the board. These are the only ones that get an IP ban. They’ve demonstrated that they are a somewhat more determined adversary. Prior to IP bans, they would simply sign up for a new account and do the same thing immediately.
I disagree that Bayesian classification could work here; we simply don’t have a large enough corpus of posts. It is incredibly effective at scale, as Gmail’s filtering demonstrates, but we are many orders of magnitude below that volume.
Sending new users to a moderation queue is an option. My sense is that it would undermine the relative sense of open access here. Having someone be able to sign up, ask a question about their Inform 7 problem, and get a reply within a few minutes is a real strength. That happens every day and it’s a great thing.
David and I are both administrators now, with the important caveat that we do not have access to the server and cannot install or modify any of the software.
But the IP address filtering problem is a fair point. However stifling it is to new users to be held in a queue for moderation, it would be far worse for them to be greeted by a “permanent ban” message. I admit it was easier to justify the policy before it affected you. I admit also that my reflexive dismissal of Tor as a use case was misguided, offensive, and based largely on ignorance.
I will let David weigh in but I am inclined to handle IP bans somewhat differently. Perhaps we can continue to issue the bans, but expunge the list of banned IPs on a weekly basis. That should still provide some deterrent value without the risk of encroaching on legitimate users over time.