----------------------
* Collaboration with the Pytroll project and the creation of the new SatPy
library will result in large internal changes to Polar2Grid.
* SatPy is a replacement for the internals of Polar2Grid and a replacement
for the Pytroll mpop package.
* Migrate Polar2Grid frontends to SatPy readers.
* Use SatPy for resampling and output writers, not just some readers.
* Rename frontends to readers, backends to writers, products to datasets.
* The remaining roadmap bullets are subject to change based on the Pytroll/Polar2Grid merger.
* Create new grid file format for more flexibility
* Allow resampling parameters to be configured based on dataset identifiers (satellite, instrument, etc)
* Further fornav updates
* Move all module logic to C++ and remove cython dependency for this module (simple one function cython wrapper should be easy to remove)
* Try rewriting in either opencl or use openmp for multiprocess work, but we're getting to the point that fornav is not the slowest part of fornav (intermediate disk use)
* Consider linking directly to PROJ.4 C library for ll2cr (removing pyproj dependency for ll2cr) to make it faster
* Update rescaling with cython wrapper (test performance before committing to this)
* Needs change of clipping and masking logic so that its a decorator and can be easily excluded from cython code (which would use internal logic for those steps)
* Add proper handling for product data being kept in memory (should speed up quite a few things)
* For better handling of in-memory data, should either let the user choose or determine it based on available memory
* Frontend's could choose logical default (VIIRS should probably write to disk, DR-RTV should stay in memory)
* Glue script can use memory analysis to come up with default but can be forced by command line argument
* Python 3 Compatibility