What is MinistryOfPackages?
A simple Python package index server implementation, meant for internal use (at least for now). The expectation is that it runs behind a firewall, and also most likely a reverse proxy. It requires PyYAML and Tornado (http://www.tornadoweb.org), and it's tested with pip and Python 2.7 It is not tested with easy_install, and easy_install support is not a near-term goal or priority (Please use pip)
What Works Now?
Right now, MinistryOfPackages is usable for simple use cases:
It has DirectoryListing support, and that's completely generic. It's meant to be used by command line tools, but the browser presentation works -- it just doesn't have all the fancy icons :)
You can do, for example, 'python setup.py sdist upload -r http://localhost:8080/dist', and then leverage the directory listing capability to go to http://localhost/packages and browse to confirm your package made it into the repository.
You can 'pip install -i http://localhost:8080/packages <pkg>' on the package you uploaded in point 2 above.
If you want to check internal first and fall back to pypi.python.org, there's no global pip config file (yet), so you need to export two environment variables, either in your personal shell init, or in the global shell initialization files:
These features still need more testing and a little polish, but they generally work.
What's Up Next?
- Completion of a data model (using Redis) to support more of the CLI and browser UI features (like 'register').
- Fleshing out a proper browser interface.
What's After That?
- Proxying requests to PyPI for pip, so MinistryOfPackages can be your primary index server for everything.
- Package caching.
- PyPI Mirroring (this can technically be done now, but it's not a good solution as it stands).
Feature requests, new ideas, and pull requests are welcome.
Why are you doing this?
I'm after a few different things with this project:
- I want something that's blindingly easy to deploy. I don't want to muck with WSGI, CGI, FCGI, whatever. I want to write code and run it, and have something that works. Having used Tornado for numerous other projects (some 'web scale'), I can tell you it works :)
- I want some enterprise-y features like proxy requests and package caching. Those will come later, but I didn't want to start this using a huge multi-headed framework because I want to get to them sooner than later :)
- I want to understand distutils, Python package distribution, and all that stuff better.