- From: Adam Bergkvist <adam.bergkvist@ericsson.com>
- Date: Tue, 4 Dec 2012 07:26:19 +0100
- To: Martin Thomson <martin.thomson@gmail.com>
- CC: Stefan HÃ¥kansson LK <stefan.lk.hakansson@ericsson.com>, jonathan chetwynd <jay@peepo.com>, "public-media-capture@w3.org" <public-media-capture@w3.org>
On 2012-11-27 19:17, Martin Thomson wrote: > On 27 November 2012 09:49, Stefan HÃ¥kansson LK > <stefan.lk.hakansson@ericsson.com> wrote: >>> getUserMedia provides an error callback. There is, however, no error for "no device". That needs to be fixed. >> >> On surface it seems that this definitely needs fixing. But the original reason for no specific error code was AFAIK to avoid finger printing > > I considered that. You could decide that all errors are reported the > same way to avoid this sort of problem. You could also decide not to > provide a callback in case of any error. That doesn't work out very > well from a user experience perspective. > > I think that we have to resign ourselves to the fact that some > expansion of the browser fingerprinting surface is an inevitable > consequence of adding features. Personally, I'm OK with this much > information leaking in light of the advantages this feature provides. > Others could reach a different conclusion. > > On 27 November 2012 09:43, Harald Alvestrand <harald@alvestrand.no> wrote: >> In Lyon I think we had a few words about this, and agreement that we would >> have more error codes (strings), but that we needed a specific proposal. > > If we want to keep this separate, I propose that getUserMedia include > the following error values: > - nodevice === no device exists that meets the mandatory constraints > specified (even if this is just audio: true, or video: true) > - permissiondenied === the user rejected access to the requested device(s) > - deviceerror === the selected and approved device could not be > acquired successfully > > These should still be useful, regardless of what we choose with > respect to synchronous/asynchronous acquisition of sources. > Setting high constraints that makes gUM() return early with a "nodevice" error enables some fingerprinting surface, but I think the second thing Stefan mentioned is a bigger question mark - the possibility to select a file instead of a camera. One of the original ideas behind this was to not lock users out of services that required, for example, a camera. When would you get the nodevice error in that case - the user could, as long as the device is powerful enough, select a hd file to satisfy hd constraints. We need to nail down how the media file feature goes together with constraints at gUM-time. /Adam
Received on Tuesday, 4 December 2012 06:42:13 UTC