Post ATxDS0jfQqPM6ap744 by seanfurey@mas.to
(DIR) More posts by seanfurey@mas.to
(DIR) Post #ATxDRthdZzcoIGotQe by mjg59@nondeterministic.computer
2023-03-24T17:13:11Z
0 likes, 0 repeats
Thoughts on SSH host certificates and how they could have made the Github situation less bad: https://mjg59.dreamwidth.org/65874.html
(DIR) Post #ATxDS0jfQqPM6ap744 by seanfurey@mas.to
2023-03-24T19:37:29Z
0 likes, 0 repeats
@mjg59 TLS is in a better position to start with, but is there any reason this approach wouldn't be useful there too? Or is it already done?
(DIR) Post #ATxDV4wKG1Qc0sCuAq by Doomed_Daniel@mastodon.gamedev.place
2023-03-24T17:47:08Z
0 likes, 0 repeats
@mjg59 Great article!I guess the link to cert_authority was not intended?
(DIR) Post #ATxDYHJugRwbHMlXQe by byterhymer@mastodon.social
2023-03-24T19:00:05Z
0 likes, 0 repeats
@mjg59 Those are certainly some thoughts!Humans do make mistakes!I can't help but think that OpenSSH does not facilitate that kind of mistake by default!GitHub needs to own it.While it is certainly possible to over-engineer an alternative methodology as you spent paragraphs outlining: KISS wins with SSH for *good* reasons! A lot of this reads as if it is unnecessary apologism for an M$ (worth $2037.96B) owned entity. M$ doesn't need ANYONE coming to their defense.
(DIR) Post #ATxDYLFo3f2HTmsJaC by mjg59@nondeterministic.computer
2023-03-24T19:20:21Z
0 likes, 0 repeats
@byterhymer OpenSSH absolutely facilitates this kind of mistake by default.
(DIR) Post #ATxDYLmm55PJ82IdyC by byterhymer@mastodon.social
2023-03-24T19:22:22Z
0 likes, 0 repeats
@mjg59 Wow, that is your reply?OK, we aren't going to concur.I have NEVER in DECADES made that error.I know many others, who also have not made that error.You're suggesting there is some OpenSSH function which automagically checks in private keys to public repositories?No.You think that its provisions for chmod on .ssh/ aren't already explicit enough?I doubt it.You think it should be over-engineered to prevent idiocy? Clearly.I don't think that is the ticket.
(DIR) Post #ATxDYMbp1IExgMBPXM by mjg59@nondeterministic.computer
2023-03-24T19:25:02Z
0 likes, 0 repeats
@byterhymer no, I'm saying that having shared private key material on disk on a large number of different servers facilitates fuckups
(DIR) Post #ATxDYNLCIaXJx5PeGO by byterhymer@mastodon.social
2023-03-24T19:27:56Z
0 likes, 0 repeats
@mjg59 OK, but that isn't OpenSSH's fault either.By default: OpenSSH will generate UNIQUE host keys.GitHub HAD TO DO OTHERWISE.Even if I am running load balancers (as I have administered for sites older than Google) which are doing TLS for their back end servers: every back end server still has UNIQUE SSH host keys (and the load balancers do too for that matter). The TLS "illusion" of presenting the same public key on the load balancers is a pinch point, and not something done en masse.
(DIR) Post #ATxDYQQynXwbXt8oC0 by mjg59@nondeterministic.computer
2023-03-24T19:38:30Z
0 likes, 0 repeats
@byterhymer As I explained in the post, if you have a different private key for every backend server, every time you get routed to a different backend server SSH will complain that the host identity has changed
(DIR) Post #ATxDk2E6cxSRwSeh5U by mjg59@nondeterministic.computer
2023-03-24T17:51:58Z
0 likes, 0 repeats
@Doomed_Daniel Urgh, thanks!
(DIR) Post #ATxDsVYTT6hFDfDsDw by mjg59@nondeterministic.computer
2023-03-24T19:42:05Z
0 likes, 0 repeats
@seanfurey I'm afraid I just don't know enough about the TLS ecosystem to know!
(DIR) Post #ATxKMWV57MyuZ0KbtQ by dickon@splodge.fluff.org
2023-03-24T19:05:57Z
0 likes, 0 repeats
@mjg59 I entirely approve of the last paragraph. We've all made mistakes -- you can tell: we're cursed to work with computers all day -- and this one seems to have been caught early enough to limit any damage.
(DIR) Post #ATxMD5VMJigK8m5cy8 by byterhymer@mastodon.social
2023-03-24T19:24:33Z
0 likes, 0 repeats
@mjg59 That, or maybe you and I have VASTLY different definitions of "facilitates".
(DIR) Post #ATxRi367odVzQIrybY by moshez@mastodon.social
2023-03-24T17:37:29Z
0 likes, 0 repeats
@mjg59 I've been slowly migrating most of my GitHub usage to use personal access tokens and "https". Luckily, "https" already has a reasonable story around certs...
(DIR) Post #ATy3AzcnVJ0BLzognQ by waider@octodon.social
2023-03-25T07:50:28Z
0 likes, 0 repeats
@mjg59 “Humans will make mistakes, and your systems should be resilient against that.”A million times this.
(DIR) Post #ATzbiW7vH4DJJSUaBM by LionsPhil@plush.city
2023-03-26T01:52:29Z
0 likes, 0 repeats
@mjg59 SSH's whack version of certs belongs with its "your session is tied 1:1 with a TCP connection" on why any serious remote shell implementation reinvets it with WebSockets instead, and why it belongs in the trashheap of legacy non-web protocols like FTP and POP3.(The me from a decade ago is screaming.)