We already had users confused by the installation process. I am still confused by the "config" documentation and would like to simplify it. How about we replace our install script with the standard jupyter kernelspec install? In other words, I don't see the point of customized kernels. A "vanilla" kernell, that can be customized later in a shell without a reinstall seems way cleaner. The advantages:

• The user wouldn't need to locate the right Python installation, as jupyter will be doing the install
• All the config options will only work as runtime variables reducing confusion as to the kernel behavior