2 ms·
> Are they though? A lot of, well at least old software, lacked configuration files, their command line flags and unit scripts were for all intensive purposes,
by masklinn 6y ago
> Are they though? A lot of, well at least old software, lacked configuration files, their command line flags and unit scripts were for all intensive purposes, their config files.
Most of the article is dedicated to explaining how modern BSDs changed from these origin and how system owners are very much expected not to edit these scripts so… yes, they are?
> Whether this actually makes upgrade harder (the next point made), is kind of debatable no?
Not it’s not? The article explains exactly how: BSD upgrade process diffs and merged every config file, which includes core rc scripts which the system owner would or should not have edited. So said system owner finds themselves spammed with every upstream update to scripts they never touched, none of which would be relevant to them.
If you argue that this is a good thing, then surely they should also be spammed with upstream updates to every binary, utility or script these use?
> Arch does fine with pacsave and pacnew files.
The article is not about arch?
> Making changes or tracking changes to these files is going to be required
No.
> And for what its worth, I'd rather find out about a breaking update via pacsave/new than figure it out months later when my copied but now outdated and untracked unit starts failing.
Does pacsave/pacnew force you to review every built-in systemd unit?
- geofft 6y agoThe claim is that the Arch pacman mechanism (and the dpkg one, and the rpm one, and ...) only prompts you when there is an actual merge, and when it does that, it's actually fine and preferable for the package manager to leave old and new versions on disk and ask the sysadmin to resolve the merge by hand, and this works in practice. It does not ask you to review every change to /etc. This is why Arch is mentioned, as an example of a system where this situation works fine. (Debian and Fedora are also good examples.) If the BSDs' upgrade procedure cannot distinguish between changes to /etc where the local sysadmin has also modified the file and changes to /etc where the file is unmodified, and they prompt you to review every single change to /etc, that seems like an easily fixable missing feature. Do the BSDs really prompt you to accept every change to /etc on upgrade for files you didn't modify, such as sshd.conf or services?