Reverend Jim has written a pretty critical review of B2 entitled cafelog: a look at bad code. You get the feel for the article from the title. I think he could have done a little more research to be a little fairer, which would have answered a couple of his points. Whilst in essence I agree with a lot of his comments about the code design and quality, I think he neglects to mention the fact that the thing works, and works well (performance notwithstanding) for a lot of people.
Edit: 02:22 Jim has posted a second article about Cafelog/B2 cafelog: a few more thoughts. There is also a continuing comment dialog in the original story.