dropping 2.5 support and single code base (not yet)

[josef...@gmail.com]

scipy and scikit-learn are discussing a single code base for python
>=2.6 including >=3.2 (or 3.x ?) which avoids running 2to3.

We haven't dropped support for python 2.5 yet, and so far I have seen
only a very few problems with 2to3 in statsmodels.

Since we have a working solution now, I don't see any need to worry
about this now (and we can let the others work out the details of a
new solution), but it's something we might want to keep in mind for
later this year.

Josef

[Ralf Gommers]

With the current setup you can't very well work on statsmodels with python 3.x (no in-place builds). So once scipy and scikit-learn have succesfully moved it would be good to follow them.

Ralf