I've did some tree-walking on the document in order to render text rudimentary on the screen. It works mostly for simple layout sites like hackernews. As soon as positioning comes into play, the renderer fails (obviously). I also made the walker to accept a visitor, which allows you to do an "enter_element" and "leave_element" call. I use this to render text found in a 'a' element blue, which makes it look like an almost real rendering of the site :)

I'm currently working on the CSS style computations. I've found some very good reading material about how to do this so it makes sense to spent some time in getting this up and running. It's the next step in getting the whole pipeline ready for rendering. Hopefully, we can soon iterate the node tree and discard not only elements that are not renderable, but also nodes that are hidden through css..