I'm working on getting FieldTrip to build. What approach would be most likely to be merged upstream? I'll concentrate there. Option 1: build against old OpenGL I have ugly patches to do this, mostly by specifying upper version limits in the cabal file. Ugly in that I introduce at least one exported function with GLfloat in the type, and I'm not sure how hard it will be to make that go away. Option 2: build against new HOpenGL This means accepting that all the rendering code needs GLfloat or GLdouble. It seems reasonable to keep all that in one module, and only export polymorphic functions which accept normal haskell types. There are many more design decisions to be made than for Option 1. For instance, where does the abstraction barrier belong? My current thinking is that tesselation shouldn't depend on OpenGL, so I'll need a data structure to hold triangles. STL export would use the tesselation code, but not the rendering code. Downsides include not taking advantage of OpenGL's compact formats for triangles, or else contaminating the triangle representation with knowledge of how OpenGL does things. I haven't thought yet about the equivalent abstraction for color information. Thanks, Daniel