The selected version in the "versions" tab is the candidate version that you will get if and when the package will be installed or upgraded. When you click on a package version manually that is different to the one that is currently installed, this automatically also changes the package status to "upgrade" (which may also be a downgrade). I think it's similar with packages that are not installed yet: If you click on a specific version, this will also set the package status to "install". But that's all just convenience. The rationale is that if you explicitly select a specific version, you probably want that version to be installed, so the package selector also does the other action: Change the package status accordingly. For retracted package versions, things are different, though: We cannot automatically decide that it's best for the user to install a different version. In the typical use case this means a version downgrade which in turn may trigger other package downgrades because of package version dependencies. This may easily lead to dependency problem reports that may be hard to resolve. Having retracted package versions in the first place means that doomsday is near; something went terribly wrong. Having retracted package versions installed means you are in the middle of doomsday, and meteorites are impacting all around you. When the concept of retracting patches was brought up (and thus retracted package versions became a possibility as a consequence), it was decided not to automatically downgrade anything. Solutions how to handle the case that a user was quicker installing patches than our maintenance team could realize that they are broken and need to be retracted included anything from "the user needs to reinstall that system" (ouch!) to "tell the user to sit tight and wait until we release a fixed version". A user may decide to try to downgrade individual packages manually. This may or may not work as desired; it may lead to the aforementioned dependency problems. But the desired behavior was not to make the package selector do anything that would initiate such a downgrade half-automatically which is the behavior you observed here. The only consolation is that all this is a very unlikely scenario which we are doing our best to avoid. Once users are in that scenario, we need advice on a case-by-case basis and publish that advice (as a security advisory?) in a high-visibility public place.