Skip to content
This repository was archived by the owner on Nov 6, 2023. It is now read-only.
This repository was archived by the owner on Nov 6, 2023. It is now read-only.

'Block all HTTP requests' should attempt HTTPS #2115

Closed

Description

@Manouchehri

Lately I've been using the 'Block all HTTP requests' feature in Chrome, and for quite a few smaller sites (e.g., personal blogs) I find myself manually changing the URL to https://. It works for a surprising amount of sites.

Wouldn't it make sense to attempt to switch a HTTP URL to HTTPS, even if there's no defined rule already? Since the site is already blocked at this point, I can't see any downside aside from a few extra milliseconds by trying HTTPS as a last-ditch effort.

Another benefit is it's possible that some SysAdmins will see a bunch of failed connections over 443/TCP in their firewall logs and hopefully get the message that their users want HTTPS.

My apologizes if this has already been brought up, I did glance at the issue tracker and didn't see it mentioned.

Activity

  1. jsha commented on Jul 10, 2015

    @jsha
    Member

    This is a decent idea. We would only want to do it for top-level site navigation, I think. It requires a little careful thought on how the UI should work. If HTTPS also fails, we would want the user to wind up still on the HTTP page, with a message saying it was blocked for being HTTP. We don't want this feature to simply wind up showing users cert errors more often.

    Also, there are some risks. Some sites (notably Forbes) serve different content on HTTPS vs HTTP. Also, I think many sites have an HTTPS version, but it is broken due to mixed content blocking. If we send users to the broken HTTPS version automatically, it makes things look like HTTPS Everywhere is broken rather than the site itself.

  2. Manouchehri commented on Jul 10, 2015

    @Manouchehri
    ContributorAuthor

    I definitely think that this should be implemented as a non-default option, because you're right, it will break several sites silently in the process. To help make it clear, maybe add an indicator to show that no proper rules were found to apply, and HTTPS is being blindly applied.

    I'm not sure if I'm following the reason behind only doing it for top-level URLs. What difference would it make?

    True, I noticed that Forbes is sending a broken certificate as well (ironically, when I went to an article about Akamai serving invalid certificates, I got an invalid certificate from Akamai...).

  3. chaccaronmacarondo commented on Jul 14, 2015

    @chaccaronmacarondo

    Yeah firefox was having that problem almost every website with any sort of Multimedia modules/trackers.

    For a while, the then-current Firefox wouldn't implement the feature you discuss without the aid of HTTPS everywhere.... Two new bespoke alpha builds later.... it was fixed..

    You could set the requirement for OCSP chains of trust, from their original issuing authority by default in Firefox... Now that was useful.... Thus, if the request for stapling the ocsp chain failed in Firefox, the HTTPS Everywhere is usually able to follow the chain-of-trust now.

    So, at one point, this watchlist for SSL observatory was working perfectly with a similar setting element enabled in 'about:config';
    I think it was something like extension.HTTPSEverywhere.orig.ocsp.GET (for the second layer of protection).

    It only evr has been giving me problems, when connecting to sites hosted by ,"akamaized-staging.net, ... , etc." CertificateAuthorites (CA's).

    Lastly, when it was working, HTTPS everywhere would subvert the warning in the Firefox taskbar, and, effectively send all the media requests through the website's server, of course with the ability to disable the https tracking rules for various services/adverts.

    tl;dr

    @jsha It WAS, ostensibly, and reassuringly, working with /more/ than just top-level authentication.

    So, in cases where ad-trackers' rules were enabled (at places like Youtube), these sorta sessions/connections got opened without any sort of 'hiccups', obviously., and would've still display the content, through the ad-trackers' rulesets by default....

    But now, the problem is back; and, on sites with "insecure images/content" the blocking is broken again... It's very disconcerting... All i wanted to do was have firefox do dns requests routed through ipv4, so I could leave the (NAT) Network Adresss Translation service alone; cuz my bro needs matchmaking/conferencing on his Xbox... I can never set the router wirelessly, either, (for security's sake).

    The fact that, where 'Block all HTTP requests' was disabled, the browser didn't have any problems, or warnings, (the "(!)" symbol in the address bar on Firefox) displaying that nice little lock icon in the address bar. Now.., it's broken again... VERY disconcerting.

    it should still be parsing the insecure elements in the webpage...

    Again, that's the culprit. Very annoying bc setting ipv4 IP protocol settings under the hood, now causes that nice toggle for HTTPS Everywhere to "staple" the original oscp's, to vanish, or like, 'glitch out'... Now, it's just 'iffy' ;_;

    P.S. [sorry, screenshot uploaded from linux is broken]

  4. Manouchehri commented on Jul 16, 2015

    @Manouchehri
    ContributorAuthor

    @chaccaronmacarondo I'm not sure if I'm able to follow what you're saying. Those sound like application issues, not plugin related. I don't understand how IPv4 relates to any of this; are you trying to talk about a SOCKS proxy possibly?

    The only HTTPS Everywhere related topic I can see is that you're saying that HTTP requests are still going through on mixed content pages when they should be blocked instead, correct?

    @jsha After discussing this a bit more with a friend, we're both not sure of the rational of forwarding users back to a blocked HTTP page. Creating an indicator (e.g., turning the icon to yellow) to show that HTTPS is being blindly applied should clarify what's going on to the user. (Blocking HTTP by default is something that's several years off, so this will definitely be an optional feature with a huge warning.)

  5. rugk commented on Jul 18, 2015

    @rugk
    Contributor

    @jsha
    You argue against your own FAQ? Why use a whitelist of sites that support HTTPS? Why can't you try to use HTTPS for every last site, and only fall back to HTTP if it isn't available?

    And because of the there mentioned reasons I also think this is not good to do this - at least not in normal 'mode'. But this issue is about the 'Block all HTTP requests' feature and there this can of course be attempted and in case a HTTPS request fails HTTPS Everywhere just blocks this request. So instead of a fallback to something unsecure (HTTP) this would be a try to load the content in a secure way and if it fails it will be blocked.

  6. Manouchehri commented on Jul 18, 2015

    @Manouchehri
    ContributorAuthor

    None of us are proposing for this to be on by default, only as a secondary option to 'Block all HTTP requests' mode. This addition makes no sense on it's own; it's logical when you're already blocking HTTP though, what do you have to lose?

    @rugk I don't believe @jsha was suggesting that we actually connect to the HTTP page, only forward back to it so the user gets the proper This webpage was blocked by an extension message instead of This webpage is not available. It's more of a UX point than a technical one.

  7. rugk commented on Jul 18, 2015

    @rugk
    Contributor

    @rugk I don't believe @jsha was suggesting that we actually connect to the HTTP page, only forward back to it so the user gets the proper This webpage was blocked by an extension message instead of This webpage is not available. It's more of a UX point than a technical one.

    Okay, in this way it's all right. 😃

  8. 1tischmann1 commented on Jul 19, 2015

    @1tischmann1

    @Manouchehri @1tischmann1 ;" bump* " ,please

    bump`` ,plz

    @jsha @rugk

  9. Manouchehri commented on Jul 19, 2015

    @Manouchehri
    ContributorAuthor

    @1tischmann1 I'm sorry, I honestly can't follow what you're saying. I don't think this has anything to do with the original topic though. There's no need to bump the thread, we all already got notifications about the original post (it's also the weekend in every timezone).

    HTTPS Everywhere already is able to parse everything and the efficiency seems fine to all of us. I'm not sure how else you could be changing HTTP to HTTPS to begin with without checking every page. Please read through the FAQ to understand what this project is.

    We can't force Akamai to start serving proper certificates. I'm still not sure how any of this is supposed to tie into DNS.

  10. 1tischmann1 commented on Jul 19, 2015

    @1tischmann1

    Thank you so much for replying to my post! I think you might do well to narrow the scope of your answer. Although, I get your point in the second paragraph of your answer.

    Nevertheless, regarding the idea about changing the apps behaviour for HTTP, as one op mentioned... that was one thing that I thought would be sensible to mention. Good ideas come in strange places, so why not make further mention that the feature request from @Manouchehri
    (oh wait, that was you!) either, might be "useful", or, post that just as a reminder that something like that might be too tedious, or inefficient to be implemented properly. I agree about breaking sites-- http links abound.

    My comment on the Akamai CAs was a bit of a jest. A bit of irony, too, since I've actually never-ever come across mention of this complaint about Akamai anywhere at all ,save for the complaints that I read on this page. I saw some irony in the fact that, browsers such as Firefox already have this "up-to-date" functionality, and they make use of some new elements in this standard (by "this", I mean Digital Certification and ocsp stapling, etc.).

    It interested me to note that they've got some issues with compliance, there-- Compliance is a bit of a rough idea, in that regard; but, standards tend to change, as new tech is being developed. I need not mention: how LONG these broken Cert IDs are gonna be around, considering how many people probably ignore some aspects of browsing security and add an exception, anyway, causing the host "akamai...blabla" to be exempt from blocking over secure connections. A travesty, I'm sure. Since I know the major focus must be "security".

    I still visit Weather.com encoded in unencrypted HTML, because I don't want to be collecting those darn Akamai certificates which are lower down the chain of trust

  11. rugk commented on Jul 26, 2015

    @rugk
    Contributor

    @1tischmann1
    Your 'TL;DR' is longer than everything else so I doubt that many people will read this text.
    Can't you just summarize this a bit? And put the things which does not have to do with the original issue in a new/other issues?

  12. 1tischmann1 commented on Jul 26, 2015

    @1tischmann1

    well, I only relate this issue here because I first noticed the thing Manouchehri mentioned, about the behaviour of block all http requests mode. The point is, Firefox, sometimes doesn't send requests for unsecure elements on a webpage. HTTPS everywhere adds a layer of protection by enforcing rule-sets (among other things, operating securely, in the entire breadth of this useful app).

    Specifically, now, which is what I’ve said in every single paragraph of that longer blurb, the behaviour of HTTPS Everywhere, that is of interest to me, is that, by chance, Firefox would automatically display the exclamation ("!") symbol about unsecure elements on the page.

    Here's the issue:

    In effect, what Manoucheri has already mentioned here--- in order to properly connect to the websites that are broken in this manner, using 'block http requests' mode, the user sometimes must type in the url again.

    Notwithstanding the way the browser normally behaves, even with HTTPS everywhere enabled, this behaviour occurred at a "safe" website (YouTube), regardless of whichever mode was selected in HTTPS Everywhere, Also, regardless of whichever rule-sets were enabled for various elements on the webpage.

    Now, one peculiarity, is that the warning would normally periodically reappear while browsing youtube pages, unless I closed the session/connection to YouTube, by refreshing the page. After changing the default ocsp handling in Firefox about:config, somehow, the warning (which was usually present while browsing this website, even with the app) was no longer showing up.
    .___. . . ___ .
    DONE. (I hope this is of interest, here vv, Read On!)

    Recently, I reset 'about:config to defaults, as well as deleting some of the current SSL tokens, Sig.'s, and some irrelevant Software Security Devices; this was in order to clear up some clutter. so I was able to re-authenticate my connections at certain websites.

    (This notion, hopefully, won't start you Open Source-er's hesds spinning round, because it IS rather tricky to describe, in its entirety.)

    It seems like, somewhere, the way that HTTPS Everywhere handles the request/response pipeline for SSL connections, doesn't evaluate, in rather important situations, whether-or-not Firefox is set to use incompatible handling/behaviour.

    It was interesting, because somehow after refreshing firefox, and cleaning out outdated tokens, for sites I needn't visit anymore, another strange thing occured... HTTPS Everywhere, for some odd reason, started showing in its menu, duplicate rulesets (i.e. a ruleset, on one page, appearing to have loaded twice, from just one host).

    ^^That last behaviour is where the issue becomes a bit more than a simple feature-request, as humbly submitted by OP, and more of a request for more strictures regarding SSL chain-stapling, as requested; but, the peculiarity of this issue, sort of 'smacked' in a very similar manner, when I read this discussion..
    @rugk @jsha @Manouchehri

    P.S. @jsha Regarding this :: *'This is a decent idea. We would only want to do it for top-level site navigation, I think.'

    On websites with so many different APIs and ad trackers, it's a wonder that it's going awry in the first place 😁 It's odd that Youtube should be causing any trouble... top-level might be a bit of an understatement

  13. rugk commented on Jul 26, 2015

    @rugk
    Contributor

    The thing you talk about in the first part is the Mixed content blocking problem of HTTPSE.
    So I just summarized this in one sentence. 😃

    The second thing seems to be a different issue. AFAIK HTTPSE has no influence on the validation of TLS connections. Firefox or Chrome validate the connections like they would always do.
    However the thing with duplicate rulesets displayed in the tray menu has nothing to do with this issue as far as I see it so if you want to report this, please open a new issue.

  14. Manouchehri commented on Dec 10, 2016

    @Manouchehri
    ContributorAuthor

    Thanks!

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions