It's not an additional overhead. The calculation is exactly the same in both cases - a mul (position, matrix) (or mul (matrix, position) depending on your major-ness preferences) calculation. Position can be stored as two components (x and y) in your vertex buffer, and when it's moved into a shader register the extra "0, 1" will be automatically added to it.
Good to know, thanks (although in extreme cases I believe that automatic addition is the overhead).
Drawing 2D stuff on-screen using the same pipeline as for 3D stuff is something that's been done successfully and without any additional problems or overheads since at least Quake in 1996. Why do you think it would be a problem today?
Simple, because I don't know a thing about it.
Some kind of reference for this alleged junking of Direct2D would be nice as well......
Well, it's just floating around including some suggestions at these forums, so I imagine the overall picture.