How could a failing/corrupted external drive cause the entire system to halt?

@jfleach I didn't read the whole discussion, but that sounds like a case for tunneling SSH over SSH so that you connect to the right-hand side device directly from the left-hand side system (without the middle one touching that SSH connection). Port forwarding with -L and ProxyCommand.

or, in general, if you have a host with all the smarts you initially connect from, do the smarts there, not on the limited jump box

@terdon I started thinking network switches actually :D it's not uncommon that they can do outgoing SSH connections too, but I'm not sure if they support keys with that. Haven't looked, though, and you probably wouldn't use a switch as a jumphost...

@ilkkachu I thought (but I know next to nothing about this) that ssh supports keys by definition.

@jesse_b sadly, you're relying on non-POSIX behaviour there: "if an operand resolves to the root directory, rm shall write a diagnostic message to standard error and do nothing more with such operands". GNU rm and --no-preserve-root?

@AndrasDeak oh, you dirty little... :P

@terdon well, you usually also don't get a general-use shell when logging in to a switch either, so it's not impossible for some things to be just left out of the UI. I really expect people would SSH in to a switch far more often than out of one.

@ilkkachu Into or out of, you still need key support, right? I mean, both sides need to be able to support keys for keys to work and I would guess that keys are an integral part of the SSH code.

I'm not sure I can answer that, because I don't know the protocol spec that closely, and anyway, there's a risk of falling into a discussion about what changes can be done and still being able to call it SSH

but AFAIK, in the usual case of a client authenticating with a password, there's the server's host key, but no key on the client side. You can nuke all the id_* files in .ssh and still make outgoing connections. And, if you only have a limited UI that doesn't allow creating or uploading private keys for the client to use, then, well, you don't use the keys.

@ilkkachu Sure, this is just idle curiosity on my part anyway. I guess I was thinking just that: if it doesn't support keys maybe it shouldn't be called ssh anymore.

@ilkkachu probably for the best it doesn't work

Well, I'd say SSH is quite useful even with just password authentication :) And stuff like sshpass and expect help in automating anyway.

But argh, I forgot how painful it is to be on the side of making bug reports. All the clarifying questions you get, even though you tried to explicitly explain everything the first time. :P

(Worst thing, I only have myself to blame, really. I found what seems to be a bug in $DEVICE, tested it, reported it, and then messed the environment to test other stuff with those devices. Idiot.)

does anyone know of a way to allow you to execute files without the execute bit being set?

One of the repos we work with has a bunch of bash scripts and none of them are set to executable. One of my coworkers is able to execute them with just ./file.sh on his WSL ubuntu instance and I'm pretty sure he isn't manually setting them to executable and then removing it each time he pulls/pushes to the repo so I'm wondering if maybe he has some sort of hack on his machine that ignores it?

Might be the W part of WSL... superuser.com/questions/1260528/…

(╯°□°)╯︵ ┻━┻