I have started a new project, by the name of Pungi. This project aims to be a tool to compose Fedora releases. The goals include simplicity of both code and interface. I hope it will be a candidate for the Fedora Project's official tool to create user specific composes of Fedora comprising of both Core and Extras packages (or just Fedora packages once things merge) as well as the tool to compose the official Fedora Project spin of Fedora. It should prove useful to anybody creating a Fedora based distribution as well. The project is GPLv2 licensed.

How it (I hope) works: A package list is fed into Pungi, either by comps or some other means. A set of yum repos to find packages in are also fed in. Pungi will search for matching packages in the set of repos and add the package to the download list. The download lost is depclosed (somewhat different than depsolved, no local rpmdb involved), all the deps that are pulled in are depclosed, etc... The list of packages is then downloaded into a configured cache dir and hardlinked into an arch specific dir within a configured destination dir. Then anaconda provided tools such as buildinstall, pkgorder, and splittree are ran on the directory of packages turning it into an installable tree and splitting packages into CD iso sized sets. Mkisofs would be used to create teh CD isos and DVD iso.

These are the basic steps. The tool could further be extended to run some simple tree sanity to ensure the compose completed correctly, or other post processing type things.

How to help: The code for this project lives in a public mercurial repository: hg clone http://linux.duke.edu/projects/pungi pungi

Write access is via ssh and can be given upon (validated) request. Discussion of the project will make use of the fedora-buildsys-list@redhat.com mailing list. There are no web pages (other than the hg-web interface at the above URL) currently. The source includes information about design and some files for testing functionality.

Since this is my first 'from scratch' python project, I welcome input not only on code content but code design, project layout, etc..

Also, even though I work for Red Hat, this is not so much a Red Hat software project. I'm developing this in the community space, a lot on my own time.