Hi all, So I looked at apt's package pinning in more detail tonight and I think I understand what bothers me about it. it combines two concepts and makes the config format REALLY hard. concept one: determining which repository is considered 'better' for any one package that both repositories have. For example: two repositories: repo1, repo2 repo1 packages: foo bar baz repo2 packages: foo quux doe in the default mode pkgpolicy=newest. yum compares repo1.foo vs repo2.foo to see which of them is newer in terms of their version. if pkgpolicy=last then yum sorts the repositories and the last one in the sort order is used regardless of its version. so for our example repo1.foo vs repo2.foo - repo2.foo would always be used b/c repo1 sorts before repo2. So if we're trying to give a general score to a repository we just give a score based on any number of arbitrary criteria. Then the repository with the highest score is used for any duplicated packages. concept two: We have a bunch of packages and a bunch of repositories and we want to say which packages should be drawn from which repositories. In this concept there would need to be some way of listing packages and specifying which repositories should be used for those packages. So concept one is best fulfilled in the existent yum.conf file per-repository by providing a scoring mechanism for the repository. concept two is probably best suited to an additional config file that specifies package identifiers versus the repository they should be drawn from. So if the key:value pair is pkg:repo then the config should in this other config file. but if the key:value pair is repo:pkg then the config should be in the current yum.conf. the file format in concept two if kinda tricky b/c on one hand it is could just be a simple format like yum.conf but on the other hand the package identifiers could be many and varied so a flexible syntax would be useful. suggestions? -sv
