Closed Bug 658707 Opened 14 years ago Closed 12 years ago

If I type an address without a protocol, assume I want https if I usually visit with https

Categories

(Firefox :: Address Bar, enhancement)

enhancement
Not set
normal

Tracking

()

RESOLVED WONTFIX

People

(Reporter: jruderman, Unassigned)

Details

(4 keywords, Whiteboard: [sg:want])

Steps to reproduce: 1. Launch Firefox with a profile that has visited https://www.bankofamerica.com/ in the past. 2. Type www.bankofamerica.com. 3. Press Enter. Result: Firefox guesses that I want http. Expected: Firefox should guess that I want https. You can think of this idea as HSTS-lite. It doesn't prevent me from visiting with http, so it's safe to infer. I want Firefox to guess the faster and more secure option when I don't specify a protocol. Trickier cases: * Typing 'bankofamerica.com', without the 'www'. * Sites that are frequently visited both with and without https.
Related: UI for HSTS is bug 572803. I like the idea of HSTS-lite. It's HNSSTS (HTTP-Not-So-Strict-Transport-Security). Perhaps we could implement a first phase of this by catching redirects from HTTP->HTTPS and marking that host name as "Default to HTTPS"?
>HSTS-lite If we base this on adaptive learning over time I think the normal naming convention would be awesome-HSTS :) This is a really fantastic idea.
We talked about this with bsterne just a few days ago. Experiment: https://www.amazon.com -> http://www.amazon.com http://amazon.com -> http://www.amazon.com https://amazon.com -> SSL error page: ssl_error_bad_cert_domain, because the cert is for https://www.amazon.com I believe there are many sites where https://<site>/<path> is an error page and/or different content than same http://<site>/<path> works fine. Accordingly, I think that we will still need an explicit indication for web sites to say that we can/should rewrite http:// links to https://. bsterne, dveditz, and I were talking about doing this as a way of bootstrapping mixed stronger content warnings. YouTube, for example, could explicitly tell us to rewrite http://youtube.com links to https://youtube.com links in the addressbar and/or only when the YouTube-hosted resource would result in a mixed content warning. We have not fully explored the idea though. See http://tools.ietf.org/html/draft-hoffman-server-has-tls-04 and http://www.chromium.org/spdy/spdy-protocol/spdy-protocol-draft2#TOC-Server-Advertisement-of-SPDY-throug
See also bug 635803, about improving the "https://<domain> uses cert for https://www.<domain>" experience.
See Also: → 635803
Brian, you'd only get this if you've successfully visited the https site before, so bug 635803 is not closely related.
See Also: 635803
I like this idea if we can do it in ways that sidestep comment 3, but I wonder what proportion of our users are getting this for free through the awesomebar. I suspect many users don't type "www.bankofamerica.com<enter>" they type "www.banko<down><enter>" which should select the https entry, if that's the thing they typically connect with (the assumption in comment 0). Again, no problem with this idea, just wish I knew what proportion of users we already help, albeit accidentally.
> I wonder what proportion of our users are getting this for free through the > awesomebar Probably fewer once we ship inline autocomplete! (Bug 566489)
Bug 720081 has been filed for fixing this for inline autocomplete.
fwiw I requested this in bug 577788
currently inline always tries to suggest https, if any page in a certain root is secure. Unfortunately, as you can imagine, some domains have secure pages but an unsecure root, though we don't have this information. I have filed bug 840027 to try figuring out a way to ask the user intentions when this happens, though if you have any insight that would be useful.
Not sure why this was given the perf keyword, as this sounds like a cool security feature and not performance related.
Keywords: perf
I gave it the perf keyword because it saves a redirect in the most common case (site redirects to https but does not use STS).
Keywords: perf
Auto-fill (inline autocomplete) now covers most of the cases (including the STR from comment 0) - it favors https: if there are such visits. When auto-fill is disabled, doing an extra round-trip to the places database may actually take longer than just going through the redirect, and would be complicated to implement, so I think this is otherwise WONTFIX.
Status: NEW → RESOLVED
Closed: 12 years ago
Resolution: --- → WONTFIX
You need to log in before you can comment on or make changes to this bug.