The Python edition is the most practically useful implementation of the ActivePapers concept, because it can build on the very well developed ecosystem for scientific computing in the Python language, which offers libraries for many scientific application domains. However, the Python platform also imposes a few limitations:
The restrictions on calclet execution cannot be strictly enforced. Circumventing the rules is possible, but unlikely to happen by accident. Users should inspect the code in a downloaded ActivePaper before executing it, or run ActivePapers in a virtual machine where malicious code cannot do any harm.
Reproducibility is limited, as the same Python code can produce different results with different versions of Python or one of the libraries on which ActivePapers depends (NumPy, HDF5, h5py and their respective dependencies). The Python platform has no formal specification and a history of regular minor incompatible changes. ActivePapers records the versions of all dependencies, but re-creating an identical environment a few years later may become a challenge.
Many popular libraries for scientific computing contain extension modules and can therefore not be packaged as ActivePapers. They can be used as external dependencies, but this increases the number of dependencies that every user must install before being able to use an ActivePaper, and it reduces reproducibility.
The development of the ActivePapers Python edition is hosted on Github.
ActivePapers supports both Python 2.7 and the Python 3.x series starting with 3.2. Since Python 2 and Python 3 are not fully compatible languages, a given ActivePaper may work with only one of them.
ActivePapers depends on NumPy (tested with 1.6 and 1.7) and h5py (at least 2.2, tested with 2.2). The current release also requires tempdir (tested with 0.6), but this is a non-essential dependency that may disappear in the future.