What technique are you using to deduce what the content of the page is? A straight port of Arc90's open source code or some magic of your own? (I'm asking because I'm keen to improve their technique for a library I'm working on.)
I'd be interested in alternative methods as well. I wrote the algorithms within readability, which are far from perfect.
Can you expand on how it's different/better from readability? What were you hoping to accomplish with your fork?
I tried going to toadjaw.com but was given a login page.