- From: Harald Alvestrand <harald@alvestrand.no>
- Date: Thu, 31 Jan 2013 23:06:50 +0100
- To: Martin Thomson <martin.thomson@gmail.com>
- CC: "public-media-capture@w3.org" <public-media-capture@w3.org>
On 01/31/2013 11:00 PM, Martin Thomson wrote: > Actually, I got some really good feedback on this feature in a > discussion last night. The security concerns over giving access to a > screen capture are pretty serious. There is a very good reason that > applications are prevented from sampling any part of the page that are > not from the same origin. Screen capture would circumvent that. It > may be that a simple user confirmation/permission question is > insufficient to convince some people that capture is safe to permit > for this reason. > > It's actually very simple. I load an iframe to your bank, using your > login cookie, briefly display some highly sensitive resource, capture > the screen, ???, profit. > > I knew this was a problem, but I didn't realize the strength of the reaction. It's exactly the same problem as a remote control interface like PCAnywhere. Many people find those creepy (and with some justification). > > On 31 January 2013 21:11, Harald Alvestrand <harald@alvestrand.no> wrote: >>> Why then does one not exist yet? >> It's "just" a matter of defining the constraints, if the WG agrees to follow >> this convention. > :) I know you know that it's never that easy. > >> I don't know what API we should use to generate the identifiers for "things >> that can be used in a screen capture". The cameras and microphones can be >> "listed" with the VideoDeviceIds / AudioDeviceIds in Travis' "v6" proposal, >> but with "things that can be used in a screen capture", I don't know a) how >> to list them, or b) what the privacy restrictions should be on listing them. > Oh, maybe you don't need to list these along with other "devices". If > the feature exists, then you don't need to be that specific (unless > you want to do per-window capture and identity the same window over > time, which I think isn't really that useful). The identifiers that > the resulting objects are given could be allocated in a similar > fashion to devices, with a single "capture" device ID. You wouldn't > be able to learn that device ID a priori, hence the suggestion to have > another constraint for this purpose. (screen: true, or a generalized > videoDeviceType: capture/camera/file) inputdevicetype=capture, which pops up a "what do you want to capture" dialog? That sounds about right....
Received on Thursday, 31 January 2013 22:07:20 UTC