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.

I am in favor of textcaptcha, esp. with questions geared toward this site. If you want to cheat, you can use this for now:
http://textcaptcha.com/

Still if its not warranted, we can do the ephemeral key thing, with maybe a user and password if we do not want to record IP addresses. The user, password, and key pair will only be used for allowing posts, and they will be destroyed every month (week?). The user can chose to recreate it. This way you can remove post on the fly per user. And when they have calmed down, they can recreate the user. The key point is that the user creation needs to be human validated, maybe with textcaptcha. That way they do not automate user creating to spam.

The weakness of just the textcaptcha is that they can post it elsewhere so that someone else solves for them (e.g. file hosting sites).