Hi,

I think there is a bit of confusion here. I meant to ask feedback
regarding exposing the type, not exposing the bandwidth and the metered
status. Or maybe exposing the type would be less of a problem than
exposing the bandwidth and the metered status? I was mostly worried
about privacy issues other than fingerprinting. Fingerprinting is an
issue but it is something well understood and we can easily reduce the
impact.

Some more comments inline.

On 12/08/13 23:00, Sid Stamm wrote:
> On 08/12/2013 12:55 PM, Stefan Arentz wrote:
>> My only concern is that it might add tiny little bit of data to a
>> browser fingerprinting algorithm.
> 
> Tiny?  Looks like a double, boolean and a callback on change
> (potentially more).
> 
> What's the expected granularity of these values?  Are bandwidth values
> coming from a set of, say, five choices, or is it precise and completely
> variable?

Obviously, metered is false/true. Regarding bandwidth, I believe there
could be solutions like we found for the battery API where we can expose
a more or less precise value. Saying that the battery level is 0.9 is
not much fingerprinting data. Saying it is .95 is a little more but
saying it is 0.9543573819 is quite too much for regular content [1].
Basically, the precision can be increased or reduced depending on some
factors.

Regarding the event, we can easily send it with a delay. This is a
mechanism that reduces the impact of fingerprinting across opened tabs:
if one tab get the event at T, another at T+x and another at T+y, it
might be less obvious that those three events are coming from the same user.

> I also wonder what kind of UI we'd make available for users' control.
> Ideas?

Examples:
"Is your data connection metered? [Yes/no]"
"Is your current connection metered? [yes/No]"

This said, the UI is a bit OT here I believe.

> What are the alternatives to get the same outcome?  What are the main
> use cases?

Using Resource Timing API can be a less precise replacement of
.bandwidth. For .metered, the less precise replacement would be to
expose the connection type and let the content do some wrong assumptions.

[1] I know we had this idea for the Battery API but I wonder if we ever
implemented it...

Cheers,
--
Mounir
_______________________________________________
dev-privacy mailing list
[email protected]
https://lists.mozilla.org/listinfo/dev-privacy

Reply via email to