On Mon, Sep 26, 2016 at 07:31:37AM +0000, Gianfranco Costamagna wrote: > >This "must" policy requirement has long gone past "pointless" deep into the > >"actively harmful" land. It is universally ignored, and I'd advise you to > >do so too. > > what about changing the policy? I admit I never cared too much about > such priorities, and I would like it to be relaxed/removed in policy whenever > possible.
#758234, which is somehow stalled for years, calls for removal of this requirement. I'd go further and require that packages should not[1] have an elevated/reduced priority only because of a reverse-dependency. Packages should be judged only based on functionality they actually bring to the user rather than on implementation details. Such an elevation may be transferred to a reverse dependency (a metapackage, or, say, from postgresql-9.6-client to postgresql-client despite the latter being empty) but not the other way. [1]. Not "must not" only because it's a bad idea to force everyone to update priorities at once. -- An imaginary friend squared is a real enemy.
