On 11/29/05, Paul A Houle <ph18@xxxxxxxxxxx> wrote: > I totally understand why this would be a nightmare for the mirrors, but > the current status quo is a disaster in terms of user experience. If you can't convince the mirrors its worth mirroring this additional material.. its a non-starter. The argument you are making right now has been repeatedly made and I don't think you are going to make any head-way with theoretical cost-savings analysis. The rate of rawhide churn makes an excellent test case for you to gather statistics. Keep up a set of sequential patches for a couple of weeks of updates to rawhide on a private server.. long enough to see multiple updates for several bugfixes packages and see how much more space it takes to keep rawhide diffed. Build a toy yum plugin client to work with your collections and have users report back. Then compare against the approach Desbonnet has created which only requires one delta to be available on the mirrors, instead of a chain. -jef -- fedora-devel-list mailing list fedora-devel-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/fedora-devel-list
