* It would be nice to use and improve the independent Cython library [cypari](https://pypi.python.org/pypi/cypari) developed by M. Culler and N. Dunfield for [Snappy](http://www.math.uic.edu/t3m/SnapPy/). Have a look at the [cypari src]((https://pypi.python.org/pypi/cypari) stuck at version 2.5.5 of pari!!
* The file "gen.pyx" (Cython side) can be automatically generated from "pari.desc", see [Sage ticket #17631](http://trac.sagemath.org/ticket/17631)
* (done) Implement error handling using [cb_pari_err_handle], see [Sage ticket #14894](http://trac.sagemath.org/14894)
* Upgrade PARI! There is a huge difference between 2.7.X (stable release) and the current 2.8.X (git master). Hence effort put for versions 2.7.X will basically be useless for 2.8.X (see trac [Sage ticket #16997](http://trac.sagemath.org/16997))
* Talk about the patches
* Avoid copy from the pari stack. After each libpari function call in Sage, there is a copy of the result and cleaning of the pari stack. It should be possible to use pointer on the pari stack directly. It is not clear how much we will win as the results are often small (*needs some design discussion*)
* Join forces with other interface (i.e. gp, Julia)