Feature list for 0.2.0

I am considering the following features for a 0.2.0 release:

1. Multiple response columns
2. Fast MARS
3. Non-diagonal weight matrix
4. Multi-core computation (check multiple parent and variable candidates simultaneously)
5. PMML serialization (but not deserialization)
6. Smoothing and improved handling of derivatives (mostly done already)
7. AIC and cross-validation as alternatives to GCV
8. MARS for categorical variables
9. Computation on multiple machines (non-shared memory parallelism)
10. Improved numerical stability
11. Anti-derivatives?

I think my current top three are 6 (because it is mostly done already), 4, and 1. Both 4 and 1 require rewriting the forward pass. While I'm there, it makes sense to also think about 2, 3, 8, 9, and 10. Oh my. It would be good to get user input on which features matter.

Currently, I am thinking of leaving out 9, 3, 7, and 11 and keeping 5 as a nice-to-have. Then the core features of 0.2.0 will be 1, 2, 4, 6, 8, and 10. How to represent categorical variables?