Spam Prevention Techniques (ideas)

So I've been thinking of various spam prevention techniques, from stats & NN to NLP, but, how would you all feel about textual challenge questions?

Either questions I can make up in my spare time about programming (i.e., "Lisp is a ____-level language."), or textual captchas? So you could still post through links or whatever, and nothing would ever leave my server, but it would make it a lot harder to automate posting.

>>4
For the question bank I can easily turn one question into about 4 - 6 questions automatically, and delivering one form at random, therefore making it harder to build a reliable bank. Though, yes, that is the main problem, it would take a lot of work to build a big bank.

You wouldn't even be banned, maybe after 5 errors or so the time between attempts goes up or something.

>>3
I can easily implement text captcha myself that isn't annoying and isn't a bunch of questions like the one in the link. The whole user and password thing puts me off because it's too close to an account, and I'd rather not do that or force anyone to do that. Though maybe after the IP has solved a captcha, I will whitelist that IP for a short time, so if someone keeps posting from the same IP, they'll not see any captchas, and if their node changes they'll have to enter it again, and the "whitelist" for the last node wears off after they don't use it anymore. That seems like it would work.