Re: [Add] Mozilla's DoH resolver policy

[Peter Saint-Andre <[e-mail]>]

Hi Wes,

On 4/12/19 10:02 AM, Wes Hardaker wrote:
> Peter Saint-Andre <[e-mail]> writes:
> 
>> We (Mozilla) do not want to encourage centralization, which is why we
>> have published criteria for becoming a trusted recursive resolver within
>> Firefox.
> 
> There are problem with this:

There are always problems. :-)

> 1) Regardless of how many items you have listed, most of your users will
> take the default. 

We are well (perhaps painfully) aware of this.

> Have you considered randomly selecting defaults for
> each user?  If they all pass your policy test, then this should be
> acceptable?

We have considered this, as Ekr noted in his message on March 27th:

https://mailarchive.ietf.org/arch/msg/doh/po6GCAJ52BAKuyL-dZiU91v6hLw

> 2) Regardless of how many items you have listed, and which one a
> particular user may select, you're still providing centralization from
> their point of view.  It doesn't matter how many in the list there are
> if all a given users queries still arrive at the same place.  That's
> still centralization. 

Personally I don't know that most users have a "point of view" regarding
centralization. Definitions might vary, but in my experience many people
think centralization occurs when a given kind of service is provided by
only one or a few entities. If we can build an ecosystem wherein there
are dozens of DoH providers, is that centralization?

> Have you considered randomly rotating all queries
> through every item in the list, 

We've considered random rotation but "there are problems with this".

> or turning the list into a series of
> checkboxes instead of a single-choice radio dialog 

There are non-trivial UX challenges and we're studying how to present
the right options to users.

> to avoid centralizing
> my privacy (which suddenly becomes less private)?

The point of defining criteria and policies (with contractual
obligations in place) is to protect user privacy in comparison to the
current state of the world. It's not clear to me why you believe that
your definition of centralization results in degradation of user
privacy, but perhaps you could explain that more fully.

Peter