- From: Harald Alvestrand <harald@alvestrand.no>
- Date: Fri, 4 Dec 2015 16:43:52 +0100
- To: public-media-capture@w3.org
Den 04. des. 2015 16:40, skrev Harald Alvestrand: > Den 04. des. 2015 08:02, skrev Stefan Håkansson LK: >> On 04/12/15 06:04, Martin Thomson wrote: >>> The options would seem to be: >>> >>> 1. do nothing >>> 2. add an allow-usermedia label to the sandbox attribute, which would >>> block gUM calls if sandboxing was enabled, but leave it enabled >>> otherwise >>> 3. add a disallow-usermedia label to the sandbox attribute, which >>> would block gUM calls only if the attribute and label were present >>> 4. disable gUM by default and require the use of a new allow-usermedia >>> attribute to enable it >>> >>> Note that 3 is quite irregular in that the sandbox attribute only has >>> "allow-x" labels currently. >>> >>> I think that 2 is simplest. It's least disruptive to existing uses, >>> while giving sites a way to prevent misuse. However, 4 is the most >>> privacy-preserving and I can see a fairly good argument for it. >> >> This is where I end up too. It would be good to know many apps would >> break by going with 4. If very few, 4 seems most compelling to me, but >> if not 2 seems like the logical path. >> >>> >>> Of course, choosing option 2 is easier if we choose option 4 for issue >>> #267 (i.e., we key permissions on both top-level and iframe origin). >> >> I agree. > > Would choosing double-keyed permission mean that if an iframe appears in > 7 different contexts, the permissions will have to be stored 7 different > times? > > Note that permission keying affects whether we're able to unify our > permissions model with the Permissions API model - that model doesn't > appear to consider iframes separately at all at the moment. Correction - the permissions API *does* consider that, but in a rather non-normative manner: "The permission storage identifier must contain the website's origin and MAY contain other information like the embedding status or the embedder's origin." (that's an RFC 2119 MAY, so you can't read it as "MUST contain if it's embedded"). > >> >>> >>> On 3 December 2015 at 22:26, Adam Bergkvist <adam.bergkvist@ericsson.com> wrote: >>>> Hi >>>> >>>> To make the discussion is this issue [1] more visible we move it to the >>>> list. >>>> >>>> [1] https://github.com/w3c/mediacapture-main/issues/268 >>>> >>> >>> >> >> >
Received on Friday, 4 December 2015 15:44:24 UTC