One of the examples I put off until I had to do it was CATiledLayer. Basically the docs were not very complete so I was hoping that by the time I really needed them I'd have better docs. Well the docs never materialized but I had 2 or 3 errata about the lack of an example so I figured I'd get off my back side and get something done. I did an example that will make it into the book. It is on drawing png's with multiple versions of the image that are drawn depending on the scale factor, its cool you will like it :).
In the mean time there were still a few things that seemed like I did not grok. So, I went to the CA lab time while out here at WWDC (besides networking that is the other 'coolest thing' about being here). After spending some quality time with the folks in the lab I think I finally have the tiled layer nailed.
I've uploaded the example here.
The example shows a delegate doing all the drawing (which is drop dead simple) and the view setting up the layers. The thing I had wrong in my head was related to levelsOfDetail and levelsOfDetailBias. Levels of detail determines the # of levels of detail (seems obvious doesn't it...) and the bias determines how many are 'reserved' for zooming in. So if you set the levels of detail (LOD) to 4 and the bias to 1, you'd get 2x zooming in (2^1) 'native zoom' (2^0), 2x zoomed out (2^-1) and 4x zoomed out (2^-2). If you set the LOD to 8 and the bias to 5 you'd get 32x zoomed in (2^5) to 4x zoomed out (2^-2). Hope that makes sense, if not feel free to ping me.