Comment 33 for bug 1033226

Mathhew: I still can't thank you enough for actually working on this, but I'm afraid I still disagree with some of the premises:

1) No update (except kernel if you aren't running ksplice) should require a restart for it's own sake. dpkg does provide for restarting daemons, in which case they will load the newly installed versions of any shared libraries. This is OS 101 (at least before MS DLL Hell)

2) If we can't convince maintainers of packages that would benefit from a restart *of the program* to do a micro-version release that restarts it (which we probably can in the vast majority of cases), display a list of daemons that would benefit from a restart.

3) Whereas I'll commend you for your own 2), 4), and 5), in the comment above, I still believe that:

A) Only kernel upgrades are actually valid reasons to reboot (if you require the new kernel). Ksplice seems to alleviate that somewhat. The premise for this assumption is that newly instantiated programs will pick the newest of the available libraries (unless their programmers came from MS land).

B) In re A), there still isn't a valid use case for replicating MS behaviour such as requiring reboots on a whim, when restarting the relevant programs will suffice.

C) What I'm describing here is the actual behaviour (and modus operandi) for most of the APT system, so I still have a hard time figuring out why anyone would want to, unnecessarily, replicate one of the most hated -bugs- / "features" of Windows...