On Mon, Nov 10, 2008 at 07:22:01PM -0500, Aaron Schaefer wrote:
It's primarily disk space and IO load issues.
I have questions, mostly meant to get people thinking about alternatives and ramifications of any solutions...
If the real issue is disk space and IO, what about the possibility of a hardware upgrade? What about moving the largest packages to unsupported (or something like arch-games) instead of basing it on votes? It looks like eliminating just the top 10 largest community packages would save 1.8 GB of space! See http://rafb.net/p/Xfw0gh39.html for package sizes. What about putting community on it's own server? What about fixing the AUR backend? What about adding a CVS commit hook in the mean time to fix permissions on upload instead of running a single cron job?
If we make these proposed changes, how will they actually impact the server and it's current problems? How will they effect Arch users? What is the price of convenience that the community repo provides to Arch users? Will there be a way to easily differentiate packages in unsupported that are actually maintained by TUs? How can we reliably tell what is popular? Download numbers, voting, pkgstats, etc. all have their own issues and biases...is there a better way? What makes the most sense in the long run when there are sure to be more TUs and packages in community eventually? Should we worry about things that are currently in community, or just new packages?
My main point is that there are many options, and any solution that gets acted upon needs to be based on hard evidence for improvement and account for all consequences of that change rather than just basing it on what sounds good. There has been a lot of rabble-rousing and not much investigation into the underlying problems and proposed solutions.
I've said this already in discussions but I'll say this again. Fixing the community back end, removing large packages, and removing unused packages are all possible solutions to the problem. If we implement all the solutions, then we get an incremental improvment. Each solution will build upon the others. We shouldn't only implement one measure. We should implement ALL measures within reason. I only raised the issue of unused or barely used packages in Community and pruning the repo. We should really be focusing on that before diverting the discussion and delving into other areas. http://wiki.archlinux.org/index.php/Community