Most recently I have been programming a lot in Python. Everything I am developing is running on Linux servers, however I develop primarily on a Windows 7 Desktop machine. The developer tools, graphics drivers, and much more, are tons better than what Linux has to offer for a daily driver. I am not looking to switch from Windows to Linux as my primary desktop machine, anytime soon "again". Python is great, I love working with python until it comes to installing native extensions on Windows, and even sometimes on Linux. Native Extensions for Python needs to be fixed.
Required Compiler Toolchain
This is the first clue that the whole thing sucks and is in some serious need of attention. To pip install or easy_install python extensions which are completely or partially written in C, you have to compile them on your machine. This is ridiculous! This isn't even required for apt-get install on Ubuntu Linux. There are build servers which build the libraries and applications for every supported architecture. Python should take page from Linux' playbook here. Python, please provide precompiled binary files for libraries. I don't want to install and configure a compiler toolchain on my production servers.
Cross Platform Libraries
Python is normally Cross Platform. I do find more and more, as I am developing with Python, there are many libraries like Nameko, which are Linux only. This is not cool! I would like to be able to run a small Microservice setup on my development machine locally. Allowing me to develop against a running cluster of services, without running a virtual machine. At the moment to get around this ridiculousness, I have to run a Linux VM on my development machine. It does have some benefits, but the additional processing required to virtualize that server locally, while still running my windows host os, and all of my development tools is a big disadvantage.
All this can be fixed!
The solution to this might not be perfect, but I think it is a simple enough solution. PyPi should require that the packages include all required sources and dependencies, prebuild the packages, and host the binaries along with the source. If Linux can do it with a much larger pool of libraries and applications, surely we can do it with python.