What framework must I use to develop mobile native application?

I would like to develop a 2D CAD application using .NET Framework WinForms or WPF, can you give me some architectural advices, how to structure the application or recommend some books?

• I'm not looking to design something extremely complicated, I want to do this just for practice, to get myself familiarized with this kind of application.

• Answer:

  I did something not too dissimilar. It was an application for a graphics modeler and engine, with the graphics done in Open GL and Direct 3D. You're sounds similar, because at least you are going to have to draw some lines for CAD, right? Here's what I found. The .NET part was totally easy to do, with the user interface involving panels and tools and stuff. I didn't even need to look at books to do it. Just open Visual Studio and drag and drop. It's all obvious. The serious graphics part: no way in hell, and I mean that. I had to drop back to C++. Which wouldn't have been so bad, except... The internal interface between C++ and .NET was like arguing with a spastic colon. Getting them to work together took at least 80% of my time. Plus, it was so difficult to do unit tests that it made every other part of development take longer. ETA, because I was asked. Model-view-controller is a fine way to structure, and I think was essential for what I was trying to do. The controller had to be in .NET. The view (well, at least the 3-D part) had to be in C++ or C. Actually, there were multiple views, and some were in .NET, but these were coarse representations of geometry as objects. Some of the controllers had to be in C++, because of certain technicalities. There was a pattern that C# was almost but not quite good enough to implement in an elegant manner, but I could do it in C++ with templates. Anyway, I digress... The problem wasn't the architecture, though it would have been worse under other architectures. The problem was getting meaningful communication between the .NET part and the C++ part. Another question I've recently seen asked whether .NET was a failure. My view is yes, partly because of this awkwardness. If you have a C++ code, it's truly a pain to refactor it. There is no obvious transition mechanism; it's pretty much all or nothing through a narrow channel. So the scaffolding hacks that make transitions possible are pretty much impossible. This was a new code, so it wasn't inconceivable, but it was still way hairy. It's not, for instance, like Objective C, where you can use C functions and even use C++ with .m files. So you can put bits together roughly to see if they work and then refactor to use your nice, well designed glue layer.