Dependencies and Interoperability

• The package should work both with Python 2.7 and 3.x. A lot of people are still on Python 2.6 and potentially stuck there. Is that true for the AOS community?
• Only NumPy (and maybe SciPy) required for full core functionality of the package.
• Design the package so it can (eventually) handle cdms and Iris objects as input and return the same as output? Perhaps also with Pandas dataframes? Or just design so the routines can be used by cdms, Iris, and Pandas objects if they choose? Can these be coerced into array_like objects?
• Note that the recently defined “scipy stack” of “standard” scientific python has been defined to include pandas, and nose in addition to the traditional ipython, np, scipy. As a consequence, epd-free, anaconda-ce, python(x,y)… will all provide all these packages by default. I (JR) would suggest that we limit our dependency to that stack.