Re: Deprecate geolocation and getUserMedia() for unauthenticated origins
On Sep 30, 2014, at 5:36 PM, Ehsan Akhgari ehsan.akhg...@gmail.com wrote: On 2014-09-30, 4:29 AM, Henri Sivonen wrote: More immediately we should make it impossible to make persistent grants for these features on unauthenticated origins. This I agree with when it comes to privacy-sensitive API: Granting a persistent permission to an http: origin amounts to granting a persistent permission to everyone who in the future has a chance of performing an active MITM attack on you. I also think that we should definitely stop persisting the geolocation permission grant for non-authenticated origins. I'm not really sure if web compat allows us to remove support for the API completely (although admittedly I don't have data on this.) Either way, we should collect some data before we take action. ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Deprecate geolocation and getUserMedia() for unauthenticated origins
On 2014-10-02, 2:34 PM, Richard Barnes wrote: On Sep 30, 2014, at 5:36 PM, Ehsan Akhgari ehsan.akhg...@gmail.com wrote: On 2014-09-30, 4:29 AM, Henri Sivonen wrote: More immediately we should make it impossible to make persistent grants for these features on unauthenticated origins. This I agree with when it comes to privacy-sensitive API: Granting a persistent permission to an http: origin amounts to granting a persistent permission to everyone who in the future has a chance of performing an active MITM attack on you. I also think that we should definitely stop persisting the geolocation permission grant for non-authenticated origins. I'm not really sure if web compat allows us to remove support for the API completely (although admittedly I don't have data on this.) Either way, we should collect some data before we take action. What data specifically? I'm fairly confident that we can make this change no matter how many websites use geolocation from non-authenticated origins. ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Deprecate geolocation and getUserMedia() for unauthenticated origins
On 02/10/14 11:58, Ehsan Akhgari wrote: What data specifically? I'm fairly confident that we can make this change no matter how many websites use geolocation from non-authenticated origins. I believe that usual practice before we remove something we don't like is to provide some warning. There are some big sites that use geolocation from http:// origins. That said, those sites will continue to work. It is most likely treated in the same way as the user rejecting the request for location. ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Deprecate geolocation and getUserMedia() for unauthenticated origins
On 10/2/14 1:07 PM, Martin Thomson wrote: On 02/10/14 11:58, Ehsan Akhgari wrote: What data specifically? I'm fairly confident that we can make this change no matter how many websites use geolocation from non-authenticated origins. I believe that usual practice before we remove something we don't like is to provide some warning. There are some big sites that use geolocation from http:// origins. That said, those sites will continue to work. It is most likely treated in the same way as the user rejecting the request for location. I think Ehsan is only talking about removing the persistent always allow permission. The effect of doing so would be that the user would simply be prompted each for each request, instead of it being automatically allowed. It's not treated as a permission denial, and is what happens by default in Firefox. I'd have no objection to this. We can monitor Input feedback on the prerelease channels, before/after the change, to see if it's a problem prior to shipping it on release. Justin ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Deprecate geolocation and getUserMedia() for unauthenticated origins
On 2014-10-02, 4:07 PM, Martin Thomson wrote: On 02/10/14 11:58, Ehsan Akhgari wrote: What data specifically? I'm fairly confident that we can make this change no matter how many websites use geolocation from non-authenticated origins. I believe that usual practice before we remove something we don't like is to provide some warning. There are some big sites that use geolocation from http:// origins. Oh, for sure. I don't have anything against warning people. :-) And in fact if we decide to do this we should publicize the idea on the hacks blog etc. That said, those sites will continue to work. It is most likely treated in the same way as the user rejecting the request for location. No, the way it would be handled is the user will get prompted again, and we wont show them the option of remembering their decision. The geolocation functionality itself would keep working fine. ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Deprecate geolocation and getUserMedia() for unauthenticated origins
On 2014-10-02, 4:38 PM, Justin Dolske wrote: On 10/2/14 1:07 PM, Martin Thomson wrote: On 02/10/14 11:58, Ehsan Akhgari wrote: What data specifically? I'm fairly confident that we can make this change no matter how many websites use geolocation from non-authenticated origins. I believe that usual practice before we remove something we don't like is to provide some warning. There are some big sites that use geolocation from http:// origins. That said, those sites will continue to work. It is most likely treated in the same way as the user rejecting the request for location. I think Ehsan is only talking about removing the persistent always allow permission. The effect of doing so would be that the user would simply be prompted each for each request, instead of it being automatically allowed. It's not treated as a permission denial, and is what happens by default in Firefox. Indeed (I phrased it very poorly!) I'd have no objection to this. We can monitor Input feedback on the prerelease channels, before/after the change, to see if it's a problem prior to shipping it on release. Great idea. ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Deprecate geolocation and getUserMedia() for unauthenticated origins
On Fri, Sep 26, 2014 at 10:58 PM, Anne van Kesteren ann...@annevk.nl wrote: Exposing geolocation on unauthenticated origins was a mistake. Copying that for getUserMedia() is too. I suggest that to protect our users we make some noise about deprecating this practice. And that in that message we convey we plan to disable both on unauthenticated origins once 2015 is over. I should have followed up on the other thread by now, but following up here, since this is kind of a continuation of the other thread: I think EKR's point about a one-time (non-persistent) grant when you know what's in front of the camera isn't really much different from using a form to upload a file over http is persuasive. And like Richard said in his reply, a one-time, non-persistent geolocation grant isn't that different from typing in an address into a form and submitting it over http. For immediate user security, do we really need to block the use of there APIs on http if every API call prompts? By immediate I mean ignoring the long term effects on the rate of https usage. I want the rate of https usage to increase, but it's not clear that breaking existing http: sites' use of geolocation that aren't actually more dangerous that typing an address into an unauthenticated form is a good way to get there. (I think making HTTP/2 https:-only would be a much better way.) More immediately we should make it impossible to make persistent grants for these features on unauthenticated origins. This I agree with when it comes to privacy-sensitive API: Granting a persistent permission to an http: origin amounts to granting a persistent permission to everyone who in the future has a chance of performing an active MITM attack on you. Moreover, as long as the storage for the permissions doesn't store the full origin, we should make sure that when the APIs are used on http: origins, the code first realizes the call is on an unauthenticated origin before the code tries to look at the permission storage. -- Henri Sivonen hsivo...@hsivonen.fi https://hsivonen.fi/ ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Deprecate geolocation and getUserMedia() for unauthenticated origins
On 2014-09-30, 4:29 AM, Henri Sivonen wrote: More immediately we should make it impossible to make persistent grants for these features on unauthenticated origins. This I agree with when it comes to privacy-sensitive API: Granting a persistent permission to an http: origin amounts to granting a persistent permission to everyone who in the future has a chance of performing an active MITM attack on you. I also think that we should definitely stop persisting the geolocation permission grant for non-authenticated origins. I'm not really sure if web compat allows us to remove support for the API completely (although admittedly I don't have data on this.) ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Deprecate geolocation and getUserMedia() for unauthenticated origins
On 28 September 2014 17:38, Anne van Kesteren ann...@annevk.nl wrote: On Sun, Sep 28, 2014 at 3:08 PM, Karl Dubost kdub...@mozilla.com wrote: Imagine if I home developing my own little Web app on my computer, I need to get through the hops of deploying TLS. For testing purposes you can get by without TLS just fine. As far as I know the definition of authenticated origin includes localhost. What is the definition of 'authenticated origins', particularly when dealing with localhost, I am worried that mine and a lot of devs setup includes setting up a local /etc/hosts file and working in convenient environment in which it will be no longer be possible. Similiarly when setting up ad hoc local networks or creating a standalone intranet, something I do fairly regularly for demos at a conference, hackdays etc This has already been a major painpoint as the author of an IndexedDB library I am fairly constantly asked questions of my library doesnt work when server from file://index.html since the security model deals around the host, it makes things harder for all developers, new ones especially. There is a solution for transmitting private information over the network, and I believe the responsibility is on content authors to decide when that is appropriate and when it is not and not the standards bodies. I agree that any site in production on the live internet transferring users information should be using https, but every website doesnt follow that use case. Asking everyone to adopt TLS is a bit like asking everyone to switch to XML. Not really. XML requires redesigning your entire application from the ground up. Adding TLS is a little bit of configuration. Completely different ballpark. It doesn't visibly and directly improve the life of people. In the big scheme of things, it gives an additional layer of security on their communications, but not privacy. It gives privacy from passive and active network attackers, no? Even more so, telling to people that they have more privacy because the communication is secure end-to-end is deeply misleading. Secured geolocations end-to-end to an aggregator such as FourSquare, Google, Facebook, etc. doesn't change anything about your privacy. That's a question of trust, not one of privacy. -- https://annevankesteren.nl/ ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Deprecate geolocation and getUserMedia() for unauthenticated origins
On Mon, Sep 29, 2014 at 2:02 AM, Adam Roach a...@mozilla.com wrote: Yes, I saw that. Your proposal didn't see a lot of support in that venue. So far for geolocation there is nobody that is opposed. For getUserMedia() there are claims of extensive discussion that is not actually recorded in text. There was also a lot of pointing to geolocation which does not seem like a valid argument. I don't think they've made their case. -- https://annevankesteren.nl/ ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Deprecate geolocation and getUserMedia() for unauthenticated origins
On Mon, Sep 29, 2014 at 8:01 AM, Dale Harvey d...@arandomurl.com wrote: What is the definition of 'authenticated origins', particularly when dealing with localhost, https://w3c.github.io/webappsec/specs/mixedcontent/#authenticated-origin This has already been a major painpoint as the author of an IndexedDB library I am fairly constantly asked questions of my library doesnt work when server from file://index.html since the security model deals around the host, it makes things harder for all developers, new ones especially. Maybe browsers should have some kind of developer extension that allows you to temporarily read the file system through localhost. There's a host of problems when you're using file URLs. -- https://annevankesteren.nl/ ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Deprecate geolocation and getUserMedia() for unauthenticated origins
There's a host of problems when you're using file URLs. pun intended? :) But I agree, for a long time developing off file:/// is pretty much impossible and developers are now required to start a server in order to build or use their entirely offline completely unconnected application, is it really a surprise that noone is building offline capable web apps? 'have an option / extention' is a weak workaround that puts the onus on often the user(!) and the developer to be able to work around restrictions in the base platform that shouldnt exist in the first place. Similiarly with gUM, bluetooth, serial ports and tcp socket apis all enabling more peer to peer (and offline capable) it seems entirely the wrong approach to have the functionality of the base platform restricted to a single end transport mode. On 29 September 2014 10:06, Anne van Kesteren ann...@annevk.nl wrote: On Mon, Sep 29, 2014 at 8:01 AM, Dale Harvey d...@arandomurl.com wrote: What is the definition of 'authenticated origins', particularly when dealing with localhost, https://w3c.github.io/webappsec/specs/mixedcontent/#authenticated-origin This has already been a major painpoint as the author of an IndexedDB library I am fairly constantly asked questions of my library doesnt work when server from file://index.html since the security model deals around the host, it makes things harder for all developers, new ones especially. Maybe browsers should have some kind of developer extension that allows you to temporarily read the file system through localhost. There's a host of problems when you're using file URLs. -- https://annevankesteren.nl/ ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Deprecate geolocation and getUserMedia() for unauthenticated origins
On Mon, Sep 29, 2014 at 12:19 PM, Dale Harvey d...@arandomurl.com wrote: There's a host of problems when you're using file URLs. pun intended? :) Heh. (Note that file URLs apparently count as authenticated origins. Which makes sense.) But I agree, for a long time developing off file:/// is pretty much impossible and developers are now required to start a server in order to build or use their entirely offline completely unconnected application, is it really a surprise that noone is building offline capable web apps? I hope the reason for that is that our solution for offline has been bad thus far. 'have an option / extention' is a weak workaround that puts the onus on often the user(!) and the developer to be able to work around restrictions in the base platform that shouldnt exist in the first place. I'd like to think we have restrictions for a reason, although the reason is not always great (e.g. compatibility with deployed content is a rather annoying one). Similiarly with gUM, bluetooth, serial ports and tcp socket apis all enabling more peer to peer (and offline capable) it seems entirely the wrong approach to have the functionality of the base platform restricted to a single end transport mode. I'm not sure what you mean by this. The more personal applications are, the more reason there is to deliver them over TLS. Unless you talk about some alternate way of distributing them? -- https://annevankesteren.nl/ ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Deprecate geolocation and getUserMedia() for unauthenticated origins
On Mon, Sep 29, 2014 at 3:44 AM, Anne van Kesteren ann...@annevk.nl wrote: On Mon, Sep 29, 2014 at 12:19 PM, Dale Harvey d...@arandomurl.com wrote: There's a host of problems when you're using file URLs. pun intended? :) Heh. (Note that file URLs apparently count as authenticated origins. Which makes sense.) Chrome doesn't actually allow gUM from file:/// URLs, but Firefox does. -Ekr ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Deprecate geolocation and getUserMedia() for unauthenticated origins
On 9/29/14 03:02, Anne van Kesteren wrote: On Mon, Sep 29, 2014 at 2:02 AM, Adam Roach a...@mozilla.com wrote: Yes, I saw that. Your proposal didn't see a lot of support in that venue. So far for geolocation there is nobody that is opposed. I'm responding on the topic of gUM, but I'll point out that a response of this is nonsense as stated (Richard Barnes) does sound like an objection to me. I also read Karl Dubost's response as being less than favorable, and Chris Peterson (https://bugzilla.mozilla.org/show_bug.cgi?id=1072859#c2) seems to be worried about how your proposal breaks existing websites. Based on your statement, I do wonder what constitutes opposition in your mind. For clarity, I am opposed to your proposal for gUM. For getUserMedia() there are claims of extensive discussion that is not actually recorded in text. There was also a lot of pointing to geolocation which does not seem like a valid argument. I don't think they've made their case. Sure, but determination of consensus is the job of the chairs. What goes into a spec is based on direction of the working group, not on the criteria of Anne van Kesteren thinks the working group has made its case. Fundamentally, the problem is that you're coming to the conversation late, and you're not willing to do the work to catch up on the conversations that have already occurred. There's no WG historian whose job it is to research decision rationale for newcomers. The reason you're not seeing anyone responding with compelling reasons on the working group mailing list is that the issue has been discussed and closed. No one has much energy to relitigate old issues, especially when objections are brought forth in the rather weak form of I'm not sure this was a good idea. If you wish to challenge the status quo, then the burden of proof is on you to come up with arguments that are both compelling and lucid enough to change the minds of the working group participants, not on the working group to justify its past work to you. I note that one of the chairs has taken the exceptionally gracious step of inviting you to make your arguments in a more sensible, threat-analysis-based form. There's your opening -- take his suggestion, and make your case. -- Adam Roach Principal Platform Engineer a...@mozilla.com +1 650 903 0800 x863 ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Deprecate geolocation and getUserMedia() for unauthenticated origins
On Sat, Sep 27, 2014 at 10:10 PM, Richard Barnes rbar...@mozilla.com wrote: Are you making an argument more subtle than everything should be HTTPS, so we should make HTTP less functional? I'm not sure where you see me making that argument in this thread. I simply recommended we move to require TLS for privacy-sensitive APIs. And it still seems within the realm of possibility to do that for geolocation and getUserMedia(). I don't disagree that things should be HTTPS, but breaking the HTTP-schemed web is not the way to get there. That's like Verizon trying to get people to use their favorite video streaming site by slowing down all the others. No, it's not at all like that. It's about tightening the requirements for getting access to privacy-sensitive user data. All this is saying is that if you want this user data, you need to deploy TLS on your site. And that can be done for free although it is unfortunately still a bit of a hassle. But guides have been created, communities have collected pointers, and gradually it will spread through conferences how to get this done. -- https://annevankesteren.nl/ ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Deprecate geolocation and getUserMedia() for unauthenticated origins
On Sun, Sep 28, 2014 at 3:08 PM, Karl Dubost kdub...@mozilla.com wrote: Imagine if I home developing my own little Web app on my computer, I need to get through the hops of deploying TLS. For testing purposes you can get by without TLS just fine. As far as I know the definition of authenticated origin includes localhost. Asking everyone to adopt TLS is a bit like asking everyone to switch to XML. Not really. XML requires redesigning your entire application from the ground up. Adding TLS is a little bit of configuration. Completely different ballpark. It doesn't visibly and directly improve the life of people. In the big scheme of things, it gives an additional layer of security on their communications, but not privacy. It gives privacy from passive and active network attackers, no? Even more so, telling to people that they have more privacy because the communication is secure end-to-end is deeply misleading. Secured geolocations end-to-end to an aggregator such as FourSquare, Google, Facebook, etc. doesn't change anything about your privacy. That's a question of trust, not one of privacy. -- https://annevankesteren.nl/ ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Deprecate geolocation and getUserMedia() for unauthenticated origins
On Sep 28, 2014, at 6:26 AM, Anne van Kesteren ann...@annevk.nl wrote: On Sat, Sep 27, 2014 at 10:10 PM, Richard Barnes rbar...@mozilla.com wrote: Are you making an argument more subtle than everything should be HTTPS, so we should make HTTP less functional? I'm not sure where you see me making that argument in this thread. I simply recommended we move to require TLS for privacy-sensitive APIs. And it still seems within the realm of possibility to do that for geolocation and getUserMedia(). You say privacy-sensitive API as if it were a boolean variable. But there's a whole continuum of risk here. Arguably any API that can collect information from the user or transmit it on the wire is privacy-sensitive. Should onmouseover and XHR be limited to secure origins? After all, you can collect biometrics from keyboard and mouse events, and use them to identify users with fairly high fidelity [1]. If you continue down this line of thinking, you end up saying that HTTP-schemed sites can't have any meaningful user interaction. This is what I mean by breaking the HTTP web. Yes, we have to draw lines. The service workers and gUM examples show a good example of a line that can be drawn: Things that allow persistent, long-lived capabilities should only be granted for secure origins. There, especially in the case of service workers, you end up with a bell cannot be un-rung scenario. That could make sense for geolocation, in the sense that Henri raised. I haven't heard a similar argument for restricting geolocation altogether. I don't disagree that things should be HTTPS, but breaking the HTTP-schemed web is not the way to get there. That's like Verizon trying to get people to use their favorite video streaming site by slowing down all the others. No, it's not at all like that. It's about tightening the requirements for getting access to privacy-sensitive user data. All this is saying is that if you want this user data, you need to deploy TLS on your site. And that can be done for free although it is unfortunately still a bit of a hassle. But guides have been created, communities have collected pointers, and gradually it will spread through conferences how to get this done. It's pretty glib to call something a bit of a hassle that can trip up even such well-resourced organizations as Apple [2]. --Richard [1] http://www.darpa.mil/OpenCatalog/AA.html [2] http://www.macrumors.com/2014/05/25/apple-software-update-invalid/ ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Deprecate geolocation and getUserMedia() for unauthenticated origins
On Fri, Sep 26, 2014 at 12:58 PM, Anne van Kesteren ann...@annevk.nl wrote: Exposing geolocation on unauthenticated origins was a mistake. Copying that for getUserMedia() is too. I suggest that to protect our users we make some noise about deprecating this practice. And that in that message we convey we plan to disable both on unauthenticated origins once 2015 is over. More immediately we should make it impossible to make persistent grants for these features on unauthenticated origins. This is already the case for getUserMedia. I can reach out to Google (and Apple Microsoft I suppose, though I haven't seen much from them on the pro-TLS front) to see if they would be on board with this and help us spread the message. I already raised this issue with Google WRT gUM and they weren't interested in making the proposed change. -Ekr ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Deprecate geolocation and getUserMedia() for unauthenticated origins
Le 29 sept. 2014 à 00:38, Anne van Kesteren ann...@annevk.nl a écrit : It doesn't visibly and directly improve the life of people. In the big scheme of things, it gives an additional layer of security on their communications, but not privacy. It gives privacy from passive and active network attackers, no? This is not what privacy is and that's one of the issues. You are talking about security not privacy. Privacy is not technical, it's a social agency of behaviors and laws. If you would like an example, there is a bridge in Ottawa (Ontario) which links to Gatineau (Québec) in Canada. On one side of the bridge (Ontario), you can take a portrait of the person and publishes it. On the other side (Québec), you are forbidden to take that pictures and publishes without authorization. This is not based on trust. This is privacy. :) Secure communications are important and your proposal is important, but there are social costs and benefits in securing systems. It's not just a matter of saying let's do it. So I would love to see why you would like to do this and in which circumstances, what are the constraints around it, etc. Not only a Yeah bullet-proof vests are cheap these days, let's go out with them, instead of changing the guns law. -- Karl Dubost, Mozilla http://www.la-grange.net/karl/moz ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Deprecate geolocation and getUserMedia() for unauthenticated origins
On 9/27/14 02:24, Anne van Kesteren wrote: On Fri, Sep 26, 2014 at 11:11 PM, Adam Roach a...@mozilla.com wrote: This is a matter for the relevant specification, not some secret cabal. I was not proposing doing anything in secret. I also contacted the relevant standards lists. Yes, I saw that. Your proposal didn't see a lot of support in that venue. And that's why taking it to a Mozilla mailing list rather than continuing the discourse that you already started feels like an attempted end-run around the standards process. Surely you understand why that appears unseemly, right? -- Adam Roach Principal Platform Engineer a...@mozilla.com +1 650 903 0800 x863 ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Deprecate geolocation and getUserMedia() for unauthenticated origins
On Fri, Sep 26, 2014 at 11:06 PM, Richard Barnes rbar...@mozilla.com wrote: It is not our job to break the HTTP-schemed web to force everyone to HTTPS. It is for features where it matters for end users. Users and web sites have been using geolocation on unauthenticated origins for several years now without major implications. Citation needed. It's no more dangerous than me typing my address into a form. Those forms should also be behind TLS. But obviously forcing that is far less practical than taking steps with geolocation. -- https://annevankesteren.nl/ ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Deprecate geolocation and getUserMedia() for unauthenticated origins
On Fri, Sep 26, 2014 at 11:11 PM, Adam Roach a...@mozilla.com wrote: This is a matter for the relevant specification, not some secret cabal. I was not proposing doing anything in secret. I also contacted the relevant standards lists. -- https://annevankesteren.nl/ ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Deprecate geolocation and getUserMedia() for unauthenticated origins
On Sep 27, 2014, at 3:02 AM, Anne van Kesteren ann...@annevk.nl wrote: On Fri, Sep 26, 2014 at 11:06 PM, Richard Barnes rbar...@mozilla.com wrote: It is not our job to break the HTTP-schemed web to force everyone to HTTPS. It is for features where it matters for end users. Users and web sites have been using geolocation on unauthenticated origins for several years now without major implications. Citation needed. It's no more dangerous than me typing my address into a form. Those forms should also be behind TLS. But obviously forcing that is far less practical than taking steps with geolocation. Are you making an argument more subtle than everything should be HTTPS, so we should make HTTP less functional? I don't disagree that things should be HTTPS, but breaking the HTTP-schemed web is not the way to get there. That's like Verizon trying to get people to use their favorite video streaming site by slowing down all the others. Since the major barrier to HTTPS deployment is the difficulty of deploying and maintaining it, the better strategy is to make HTTPS simpler to deploy. Some of the hosting providers have been making good progress on this front. --Richard -- https://annevankesteren.nl/ ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Deprecate geolocation and getUserMedia() for unauthenticated origins
Exposing geolocation on unauthenticated origins was a mistake. Copying that for getUserMedia() is too. I suggest that to protect our users we make some noise about deprecating this practice. And that in that message we convey we plan to disable both on unauthenticated origins once 2015 is over. More immediately we should make it impossible to make persistent grants for these features on unauthenticated origins. I can reach out to Google (and Apple Microsoft I suppose, though I haven't seen much from them on the pro-TLS front) to see if they would be on board with this and help us spread the message. I filed https://bugzilla.mozilla.org/show_bug.cgi?id=1072859 for geolocation. -- https://annevankesteren.nl/ ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Deprecate geolocation and getUserMedia() for unauthenticated origins
Speaking as someone who (1) chaired the IETF working group on geolocation and privacy for several years, and (2) now manages PKI and crypto for Mozilla -- this is nonsense as stated. It is not our job to break the HTTP-schemed web to force everyone to HTTPS. Users and web sites have been using geolocation on unauthenticated origins for several years now without major implications. The most common uses involve one-shot access to location for things like content customization. It's no more dangerous than me typing my address into a form. I could agree with Henri's suggestion in the bug that we should limit persistent permissions to authentication origins, as we do with gUM. But in neither case have I heard any coherent rationale for disabling the features entirely, beyond Nobody should use HTTP anymore, which is clearly a non-starter. --Richard On Sep 26, 2014, at 3:58 PM, Anne van Kesteren ann...@annevk.nl wrote: Exposing geolocation on unauthenticated origins was a mistake. Copying that for getUserMedia() is too. I suggest that to protect our users we make some noise about deprecating this practice. And that in that message we convey we plan to disable both on unauthenticated origins once 2015 is over. More immediately we should make it impossible to make persistent grants for these features on unauthenticated origins. I can reach out to Google (and Apple Microsoft I suppose, though I haven't seen much from them on the pro-TLS front) to see if they would be on board with this and help us spread the message. I filed https://bugzilla.mozilla.org/show_bug.cgi?id=1072859 for geolocation. -- https://annevankesteren.nl/ ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform
Re: Deprecate geolocation and getUserMedia() for unauthenticated origins
On 9/26/14 14:58, Anne van Kesteren wrote: Exposing geolocation on unauthenticated origins was a mistake. Copying that for getUserMedia() is too. I suggest that to protect our users we make some noise about deprecating this practice. There have already been extensive discussions on this specific topic within the W3C, and the conclusion that has been reached does not match what you are proposing. I would be extremely loathe to propose that we implement outside the spec on a security issue that's already received adequate discussion in the relevant venue. More immediately we should make it impossible to make persistent grants for these features on unauthenticated origins. Our implementation of getUserMedia already does this, and the getUserMedia spec has RFC 2119 MUST strength language requiring such behavior. I can reach out to Google (and Apple Microsoft I suppose, though I haven't seen much from them on the pro-TLS front) to see if they would be on board with this and help us spread the message. The email address you're looking for is public-media-capt...@w3.org. This is a matter for the relevant specification, not some secret cabal. -- Adam Roach Principal Platform Engineer a...@mozilla.com +1 650 903 0800 x863 ___ dev-platform mailing list dev-platform@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-platform