Right now Fetcher copies identical files from the configured caches to the destination path. When downloading metadata, Fetcher uses a tmp directory to download all metadata, and it gets identical files from the repos raw metadata directories. Normally, the raw metadata dir of the same repo we are downloading is the best candidate to have the identical files. Then the original repo dir is renamed to something tmp too, and the tmp dir is renamed to the repo raw metadata dir (this allows to keep always consistant raw metadata copies). I have experimented locally so when a identical file is found in a previous cache, I hardlink instead of copy, at the end I don't care which directory is deleted first, because hard links are "refcounted". My refresh is much faster now. But Fetcher is also used in other places appart of repository metadata download, and there, I have no idea if the cache dirs are in different filesystem than the destination directory. I tested and ln will fail. Can I do hardlink by default, and fallback to copy? Is there anything else I could be missing? If not, I will commit this :-) Duncan -- To unsubscribe, e-mail: zypp-devel+unsubscribe@opensuse.org For additional commands, e-mail: zypp-devel+help@opensuse.org