I'm looking forward to MT-Blacklist 1.5, not wanting to bother with 1.0 with such a release impending, but for the moment I'm implementing some of Yoz Grahame's seven tips for halting comment spam. I much prefer the idea of making MT blogs less cookie-cutter, and so more difficult to post programmatically to, but it's just another arms race.
I had already implemented Yoz's tip #2, don't link to the comments script on the front page, on aesthetic grounds. However a big chunk of my spam was to this post on Typepad, the comments for which I closed for expedience. You easily get around a lack of comments form on the front page by spamming the results of searching "typepad comments" (I'm somewhere around 560-570; I moved while I was writing this entry). All of Yoz's tips can be gotten around easily, actually, which is why I'm looking forward to MT-Blacklist 1.5 still.
Here's another fanciful Club solution that's sure to annoy: don't allow <a href> in comments, but rather <j href> or something else made up. Clearly mark for human viewers the translation, of course. (To James' credit he also offers a Lojack solution, Bayesian filtering. I'm also curious how many complaints about the inaccessibility of his captchas were from people who actually know what they're talking about, though it's pretty obvious they're a bad solution for the blind.)
I don't suppose I'll use that, as it's too specific and nonstandard. However, the idea of mixing up input is sound: you could disallow HTML in comments, but allow Textile or Wiki markup. Anything easy, nonproprietary, and not HTML will help fragment the "market" of comment markup enough to make spamming harder.
But--and you knew this was coming, right?--I was thinking of comment spam in the context of authenticated comments. You can't simply allow any signed comment free reign, as spammers will make keys and, when you ban those keys, spam keyservers with temporary keys, and that simply won't do. So authenticated commenting as a spam blocking tool would have to rely on the web of trust: both anonymous and untrusted posters are given lesser privileges to post (comments could be screened, or posted with no links), but when you trust (or transitively trust under some system undescribed herein) someone's key, e can post comments freely with whatever URLs e cares to include. Editing might be similarly limited to keep people from posting an innocuous screened comment, then editing it into spam, though that's working pretty hard for little gain.
I'll write a DAC program someday! Honest!!
