Re: [Add] Mozilla's DoH resolver policy

[Wes Hardaker <[e-mail]>]

Ted Hardie <[e-mail]> writes:

> On Fri, Apr 12, 2019 at 1:38 PM Wes Hardaker <[e-mail]> wrote:
> 
>     > What this should do is associate the clients with the lowest latency
>     > resolver.
>    
>     I wish we (collectively) could get over the notion that latency was the
>     only, or the most important metric out there at all times.
>     Happy-eyeballs approaches to problems can go too far.
> 
> The world of browsers cares a lot about latency, since it effects page load time and
> thus revenue.  I don't think it is going away as *a* metric, in other
> words.

Nor should it!  It's an important one!  I never tried to say otherwise.

> What else do you want to see optimized here?

Like most "triangle-like" choices, there is likely no perfect balance of
metrics you could pick from and each of us would likely pick different
solution points.

The three big ones that are coming back time and time again are:

1. latency
2. privacy
3. authenticity

Right now, most web-browsers (as you point out) are optimizing for (have
always been in a race for) latency.  Only recently is the desire privacy
actually becoming important enough that there is a desire to use a
slightly slower protocol like DOH to achieve privacy away from
unencypted traffic to access point resolvers/ISPs, and from the access
point resolvers/ISPs.  This, itself, is probably a good trend.

But then you bring in the discussion of DNSSEC and authenticity, which
is the only current way to ensure end-to-end data integrity and combine
this with Martin's statement of

> We don't believe that DNSSEC is essential to our primary goals, which
> are improving privacy of browsing activity.

Which clearly shows that between the above three, they're willing to
sacrifice some level of authenticity to achieve better latency and
privacy.  Note, though, that this choice isn't wrong.  It's a choice.  I
may choose a different one, as may others.  And I'm more than sure my
list above is incomplete from an ideological stand point; it just
happens to be the three biggest I could think of.

It would be interesting to put a "pick two" dialog box in front of a
billion users and see what was selected.  Then watch them as the
usability changed and see if when presented the same choice a month
later if they'd stick with their original choice.

-- 
Wes Hardaker
USC/ISI