3 ms·
Completely agree, and arguably why stop at API servers? Depending on server-side HTTP -> HTTPS redirects for security reinforces/rewards bad practices (linking
by lxgr 2y ago
Completely agree, and arguably why stop at API servers?
Depending on server-side HTTP -> HTTPS redirects for security reinforces/rewards bad practices (linking to HTTP, users directly entering HTTP etc.), in a way that makes users vulnerable to one of the few remaining attack vectors of "scary public Wi-Fis".
- Aachen 2y agoWe are. Slowly, due to lots of legacy, but surely getting there. See the small steps over the years where it was first an add-on to force https-only mode (HttpsEverywhere, 2011), then browsers started showing insecure symbols for http connections (e.g. in 2019: https://blog.mozilla.org/security/2019/10/15/improved-security-and-privacy-indicators-in-firefox-70/ https://blog.mozilla.org/security/2019/10/15/improved-securi...), and more recently I think browsers are starting to try https before http when you don't specify the protocol. I've also seen a mention of strict https mode or something, not sure if that's a private navigation feature or something yet to come, but warning screens equivalent to insecure certificate pages are getting there
- Dylan16807 2y agoChrome's version of trying https first sure is annoying though. If a site is down entirely, when chrome can't connect to port 443 it confidently declares that "the connection is not secure because this site does not support https" and gives a "continue" button. Then when you click "continue" nothing happens for a while before it finally admits there's nothing responding at all. So it gives a misleading error and takes longer to figure out if a site is actually down.
- jraph 2y agoI'm highly surprised by this. It seems very dumb and I have never seen anything like this, though I never use Chrome and very rarely fire Chromium for testing something. Is there something to read about this, like a dev ticket?
- Dylan16807 2y agoThere might be but I'm not aware of any tickets. But if you open chrome and navigate to 192.168.20.20 you should see it. Or any domain that resolves to a non-responsive IP, if you have one in mind.
- jraph 2y agoJust tried on Chromium, I get ERR_ADDRESS_UNREACHABLE as I would expect.
- Dylan16807 2y agoYeah, someone else said it's not enabled for everyone. It might just be settings? Do you have "Always use secure connections" turned on? I don't remember if that was a manual toggle or not. However it's triggered, I hope they make the UI better before they deploy it more widely. And I'll quote the exact text for you: [warning triangle] The connection to 192.168.0.40 is not secure You are seeing this warning because this site does not support HTTPS. Learn more [Continue to site] [Go back] Then when I click continue, the only sign it's doing anything is that the reload arrow becomes a stop X. After 20 seconds it moves on to "This site can't be reached".
- jraph 2y agoYep, indeed I'm seeing this if I enable this option.
- o11c 2y agoFirefox has a similar bug, but for DNS rather than connection.
- plorkyeran 2y agoI have never seen this happen in Chrome. I just tried it to make sure I wasn't crazy and it did indeed go straight to telling my that the connection timed out without showing a security error first.
- hirsin 2y agoThis is only enabled for a small percent of people at random, otherwise just folks in the advanced protection program.
- nimih 2y ago> arguably why stop at API servers? I think this is pretty convincingly argued in TFA, honestly: modern browsers understand and respect HSTS headers, maintain enough local state that such headers are meaningful, and HSTS preloading is easy enough to set up that it should be achievable by most website operators. Furthermore, it is actually quite hard to concoct a scenario where a user clicking an HTTP link and getting immediately redirected constitutes a danger to their security: unlike with API endpoints, people clicking links (and in particular, links which were typed out by hand, which is how you get the HTTP protocol in the first place) are generally not making requests that contain sensitive information (with the exception of cookies, but I would argue that getting someone to have a sane configuration for their cookies and HSTS headers is a far easier ask than telling them to stop responding to all port 80 traffic).
- bigstrat2003 2y agoThe push for "TLS all the things" was already a massive overreach that actively made security worse overall, because it further ingrained the user tendency to click through scary browser warnings (all for the sake of encrypting things that were fine in plaintext). And you want to go even further? No thank you.
- tomsmeding 2y ago> because it further ingrained the user tendency to click through scary browser warnings (all for the sake of encrypting things that were fine in plaintext). Why should there be more scary warnings when more websites use TLS? Sure, you get more scary warnings if you set your browser to "warn if it's http", but then you're asking for it.
- xboxnolifes 2y ago> Sure, you get more scary warnings if you set your browser to "warn if it's http", but then you're asking for it. Defaults. They matter.
- lxgr 2y agoWhat scary browser warnings do you regularly get when visiting HTTPS sites? It’s also not just about avoiding transmitting HTML etc. in plaintext; somebody being able to inject arbitrary scripts into sites you otherwise trust is bad as well. But as I've said above, I think the HTTP -> HTTPS redirect should have never happened at the HTTP level. If we'd done it in DNS or at least as a new HTTP header ("optional TLS available" or whatnot), we could have avoided locking out legacy clients.
- zzo38computer 2y agoScripts other than those that the user had specifically allowed, can also be a problem, whether or not it is with TLS. With TLS, only the server operator can add malicious scripts; without TLS, spies can also do so; either way, it can do so. Specifying availability of TLS (and, perhaps, which ciphers are usable, in order to avoid the use of insecure ciphers) by DNS would do, if you can (at the client's option) acquire DNS records securely and can know that they have not been tampered with (including by removing parts of them). (This is independent of whether it is HTTP or other protocols that can optionally use TLS.) (Actually, I think that using DNS in this way, would also solve "Gopher with TLS"; the format of gopher menus makes it difficult to use TLS, but knowing if TLS is available by looking at DNS records would make it work. Gopher servers can still accept non-TLS requests without a problem, if none of the valid selector strings on that server begin with character code 0x16 (which is always the first byte of any TLS connection, and is very unlikely to be a part of any Gopher selector string).) It would also help to make cookies unable to cross between secure and insecure connections in either direction (always, rather than needing a "secure cookies" flag).