On Thu, Sep 09, 2004 at 02:56:32PM -0400, Bill Nottingham wrote: > Christof Damian (christof@xxxxxxxxxx) said: > > > At the moment, the Fedora updates tree is somewhat out of hand > > > (look, wow - 27G). > > > > > > Does anyone have any specific and concrete objections to removing > > > older, superceded, updates? > > > > I guess this is just meant for cleaning up the directory and not to > > save space? I don't mind the cleanup, but would like the old updates > > archived somewhere. > > It's for both, actually. > > Frankly, I see no need for archiving of: > > a) old test updates > b) test updates that have been superceded by final updates Your mirror and rsync folks will like you for reducing the size of the site. Test updates old and superseded by a final package should have a short life. > Others, there could be some use for, I suppose. For others a longer life makes sense. Some new updates trigger troubles for some users and keeping the older bits handy has value. I should comment, IMO src rpms should have a longer life than the binary. With a src rpm bits can be rebuilt and differences inspected for debugging things. Rsync sites have control over the --delete function so those that have the space can elect to keep old cruft and remove it with a local policy. Start with a simple straw man policy that has some potential of being automated and being done in stages. Something like: I. Test a) old test updates b) test updates that have been superseded by final updates will be moved to 'old-test' 7 days after a newer package is available. After 14 days they will be removed from 'old-test'. II. Released c) Superseded binary packages will be move to "Superseded Packages" 14 days after the general release of a newer version and removed after 30 days. Src packages will be retained an additional 30 days. -- T o m M i t c h e l l Just say no to 74LS73 in 2004
