> The article has massive selection bias: all people surveyed are actively > using Python in their work. Many PETSc users don't write Python at all, > PETSc is just using it for the build. If latter population was as > up-to-date as these astro folks, we'd be able to drop compatibility for > 10-year old versions of Python and upgrade a lot of cruddy code.
I was actually thinking of the Python interface for PETSc rather than the build system. For the build system we should support any 2.x version. And, even worse, I expect that at some point both 2.x and 3.x need to be supported concurrently... Best regards, Karli
