News for hackers. Stuff that matters.
Re:Mozilla shouldn't copy Chrome's permission prompt for ext... by amelius on Sunday July 03, 2016 @11:36AM (#12026339)
I think a permission model should have 3 choices for every question: allow, disallow, and pretend.

With the last option, app-makers cannot bribe the user into installing the app with unwanted permissions anyway.

For example, when the app wants file-system access, the "pretend" option would give the app a view of an empty or dummy (fake) file-system. When the app wants access to GPS coordinates, the "pretend" option would give the app dummy coordinates. Et cetera.

Re:Mozilla shouldn't copy Chrome's permission prompt for ext... by tetrep on Sunday July 03, 2016 @12:25PM (#12026542)
I don't see "pretend" as being productive. If you're dealing with an app that would otherwise refuse to run without a permission, even if it didn't truly need the permission to function, then it would simply escalate the arms race. It would be pretty easy to detect a fake filesystem (try to write a file, then read it back) and spoofed GPS (GeoIP, ping supposed nearby servers, etc). And, of course, these checks could be encapsulated in a convenient library for developers to easily use. Then you'd have the browser's response to this and we'd end up in the ad-blocker-blocker-blocker world in which we now live, only we're wasting the time of developers who would otherwise be doing productive things (more browswr bug fixes/features). At least with the current ad-blocking mess we're only wasting the time of adtech.
Re:Mozilla shouldn't copy Chrome's permission prompt for ext... by jimrandomh on Sunday July 03, 2016 @01:14PM (#12026757)
Yes, but at that point they've lost plausible deniability; they can no longer claim that they asked for permission because some obscure feature required it. People are much less likely to agree to "trade my filesystem for a game" than they are to agree to "play a game that uses the filesystem to save games".
Re:Mozilla shouldn't copy Chrome's permission prompt for ext... by drdaeman on Sunday July 03, 2016 @01:41PM (#12026876)
They tend to just say "we need this permission for the application/extension to work correctly". And stop at this.

Seriously, a certain (quite famous, 5M installs) app got an update that requested a permission to install packages (!). It was actually harmless, as granting this won't work - the system won't allow this to non-system apps, but nonetheless it had just looked wrong.

I've emailed their support, asking why is it there. And - guess what? The reply was, essentially, "we need it because we do, it's required for the app to work, sorry but can't disclose the exact details". Well, polite, but essentially like this.

Long story short, I've accidentally mentioned this on a friend's blog, and my comment was noticed by his friend who had connections to the app's company (world's small, hah!) - so the ticket got elevated and took a proper review. Result is, this had just accidentally slipped from some dependency's development builds. Or something like that. Got apologies for the dubious response and a discount coupon, and the issue was fixed.

The thing is, it seems that a) I highly suspect, most users don't give a second thought about about accepting this permission (would there be more reports on this, I guess, the tier-1 support would know about the issue and the response would be different) and b) the initial attitude is "we know better".

Re:Mozilla shouldn't copy Chrome's permission prompt for ext... by smeroth on Sunday July 03, 2016 @12:05PM (#12026452)
I'd like to add a fourth choice: ask every time. When the app wants to access your GPS coordinates, you can grant it temporarily. Much like Android's integrity tools (maybe it's just a Cyanogenmod feature) where you can choose to be asked every time an app wants to access some kind of data.
Re:Mozilla shouldn't copy Chrome's permission prompt for ext... by lifthrasiir on Sunday July 03, 2016 @12:17PM (#12026498)
The mix of the grandparent's and your options might give a very streamlined UX: the app is always given the full permission which is but only simulated by default, and any actual access the user is noted of that with a non-modal message box, where one can turn the simulation on or off. Some permissions (e.g. storage access) may not be hard to generalize like this, but it may work for other permissions like GPS and so on.
Re:Mozilla shouldn't copy Chrome's permission prompt for ext... by deno on Sunday July 03, 2016 @12:25PM (#12026548)
> Some permissions (e.g. storage access) may not be hard to generalize like this

In fact doing it this way would be very bad.

In sandbox model it’s usually the container that owns the file selection ui, perhaps accepting some plugins etc. from the client. The client is never granted permissions to see outside the sandbox. Pretend mode in this context would result in requesting unnecessarily broad access.

IMO for things like location the only way to do it both in user-friendly and privacy-friendly way requires a capabilities system. When an app is given privacy-sensitive data it shouldn’t be simultaneously granted any permissions that allow this data to be leaked.

Re:Mozilla shouldn't copy Chrome's permission prompt for ext... by drdaeman on Sunday July 03, 2016 @01:07PM (#12026731)
Won't happen. Hearing about this, some Developers get really pissed that they can't spy on users, they blow a hype how they'd completely go out of business, there won't be any great products anymore on the market and all that sort of stuff.

One relatively famous example is from some years ago when someone had proposed to spoof device IDs in Cyanogenmod and Swype had a spaz.

Re:Mozilla shouldn't copy Chrome's permission prompt for ext... by chr1 on Sunday July 03, 2016 @12:11PM (#12026479)
As a user i often wanted the pretend option, but it is definitely going to make life of tech support harder, as many users will come and complain about their files missing, or map showing them wrong location.

It probably needs to show notification when actual call to fake api is made.