On Wed, 2006-09-06 at 16:25 -0700, Leslie Satenstein wrote: > This solution would take a little change on the client side, and > perhaps on the host side as well. > > On the client side, we could have a "first free number" facility, such > that each yum update would be sequentially numbered, with of course, > a log file. Associated with the number would be the download > timestamp and a llst of items that were downloaded. A CVS type of > repository for replaced items would be necessary. > > If one had to recover, then by looking at the log, one could determine > how far to backtrack. > > Backtracking would undo all updates to that version. I could then, > although somewhat tedious, one could redo the yum updates for the > files really wanted. > > A second feature could be the limitation on the number of versions of > backups before wrap around occurs.. I would think that 45 days would > be sufficient. I chose 45 days, because many businesses have financial > month ends, and we would want a recoverable stable system for at least > that amount of time. > > That is my two pennies worth of thoughts. New development? I am not > into the internals of yum and rpm, so I cannot tell you. and all of the above is meta-meta-data about a packages and repositories. Therefore it should stay safely outside of yum in a plugin or a yum-util of some kind. -sv
