This is getting ridiculous. The real culprit here is: Policy 2.5: # Packages must not depend on packages with lower priority values # (excluding build-time dependencies). In order to ensure this, the # priorities of one or more packages may need to be adjusted.
which came only as a technical quirk of how CDs were built at the dawn of time. It is actively harmful. Usually the fallout is merely tons of unnecessarily installed crap like gcc-4.712345-base, but sometime we're getting bugs like this. Another problem is making it hard to switch defaults, be that during a transition, distribution defaults or user choice: when replacing systemd with sysvinit, exim with postfix, etc, neither the user nor automated tools are told which dependencies are actually needed and which are cruft that at best is useless and might cause trouble. Thus, I believe the current (stalled forever) proposal of merely dropping the rule is not enough. We need to change "priority of rdepends MUST be raised" all the way to "priority of rdepends MUST NOT be raised, every package is to be evaluated only based on what it directly brings to the user (elevation possibly moved to a metapackage/etc but never copied the other way)". Meow! -- The bill declaring Jesus as the King of Poland fails to specify whether the addition is at the top or end of the list of kings. What should the historians do?
