Dear Debian people, Already possible or not, I would like to have a Debian system for which packages can be installed either by a specific user (root/sysadmin as usual) only or by any other (or a group of) users. But this would also depend on the class of the requested packages: 1. packages providing mainly commands or library facility (large majority packages?), 2. packages providing modifying system runtime like (root) services, new kernel modules, ... (very few packages?).
Any (or a specific group of) users could be able to install any package of the first class by their own without asking a sysadmin (or explicitly acquiring privilege of) user. With this goal (dream?) in mind, I tried to cluster all the packages between the one that wouldn't change the system runtime (and therefore even after a reboot) and whatever would be the sign of that.Those package installation should insure a sort of system default runtime reproducibility in fact. If needed in the future, any build process could help to class/tag a package regarding this purpose (as it could also read any debian/* packaging file). Finally the best would be for apt to handle such a scenario, allowing those users to run the apt command which may fail or not regarding each package class. To cluster packages, my first material is to look at the file contents of a package (having content related to sbin or a systemd service or init.d or ...) Opinions or ideas are welcome. Could this clustering by it-self be interesting fin any case (for Debian QA, metrics...)? Regards, Patrice
