May I ask what content you are planning to display in closer zoom levels? 2d tiles mapped on a sphere or 3d models? Also, will you focus on “realistic” rendering or abstraction (e.g. with regard to road widths, map colors based on feature categories vs. surface texturing, …)?
Well, the globe isn’t a sphere. The tessellator accounts for the eccentricity of the earth. So, you’re looking at an ellipsoid. Having said that, the tiles should be whatever you want. I’m currently working on support for WMS servers so a political boundaries layer will work. The design also allows for an elevation model with however much detail you want. So, it should have the potential to render very detailed terrain with accurate surface normals.
I plan to support 3D models through an implementation of collada in a KML parser, but you shouldn’t expect to see that anytime soon. It’ll be some time before an implementation of KML even supports simple polylines.
In summary, I want this to be more of a framework rather than a program for end users. It should support abstract, cartoony maps, as well as photorealistic ones. Im aiming to be as flexible as possible. It should be rather simple for a developer to override the default behavior of almost anything.
Having said that, the default behavior will probably lean towards the photorealistic side of things.
I typed this post on my phone, so please ignore any of my bad spelling, grammar, or lack of organization.