I thought I fixed it, but apparently not. It is driving me crazy.
Having /etc/ssh/sshd_config not do what it appears to do is inexcusable, however that comes about.
Even if the fix is no more than structuring the file a little differently so there are parts both before and after the .d loading, and comments that say what overrides what, and all the actual config files and default behavior are still the same.
Downstream distros sometimes want to override the default settings. That's what .d is for. It doesn't violate the dogma "principle of least surprise", especially if you run `man sshd_config` and learn the first thing about sshd's config files.
"That's just not how it works" Incorrect.
"dogma "principle of least surprise"" That is not dogma, it is a principle. It's right there in the name.
Dogma might be the understanding that the principle of least suprise is a valid principle.
But this suggests you don't subscribe to this silly "dogma" eh?
I also don't know anyone else who would assume that others don't know "how this has always worked" except someone who only learned themselves last week and think they know a lot now.
I didn't say anything about adding comments as an end user. My point about comments was that least suprise could be resolved trivially without even changing the existing desired behavior, by nothing more than the OS shipping a default sshd_config that was structured so that it clearly had a section that will be overridden, and a section that will override, which requires nothing more than a few comments.
I'm looking at a couple sshd_config files on my servers, and the Include directive for the overrides is the first non-comment line in the file.
Having that Include be first means that changing any settings in the sshd_config file will override anything set by sshd_config.d/* files.
What in the heck are these people doing to break their config? There really is no legitimate issue here.
You shouldn't need to edit an undocumented file that's force enabling an insecure authentication method in the first place.
"You're using it wrong" doesn't change the fact that the defaults are poor and likely to ultimately cause more devices to be improperly configured than if it was left as sysadmins and end-users expect.
I've been looking for info about whether it's force-enabled with a vanilla install where the user doesn't actually choose for it to be enabled. I'm happy to be educated here.
And if you downloaded and installed the authentic Debian 12 image from debian.org, you don't get it either. Must be a Ubuntu thingy.