Re: [Add] fixing coffee shop brokenness with DoH

[Eric Rescorla <[e-mail]>]

On Wed, Jul 24, 2019 at 7:33 AM Brian Dickson <[e-mail]>
wrote:

>
>
> On Wed, Jul 24, 2019 at 9:15 AM Eric Rescorla <[e-mail]> wrote:
>
>>
>>
>> On Wed, Jul 24, 2019 at 3:54 AM Ted Lemon <[e-mail]> wrote:
>>
>>> On Jul 24, 2019, at 6:40 AM, Eric Rescorla <[e-mail]> wrote:
>>> >> In many cases, whether you are being told the truth by your recursive
>>> resolver is a less interesting property than whether you can get the
>>> correct answer. For instance, if the recursive resolver replaces the A
>>> record you want with NXDOMAIN even if you know that it's done so, you're
>>> still blocked from going where you wanted to go.
>>>
>>> However, even if you are not blocked, without DNSSEC you do not know
>>> that the resolver gave you the right answer.  So ideally you want both: not
>>> to be blocked, and also to be able to validate the result you got.
>>>
>>
>> Well, this is true, but in the Web context, as we move towards 100%
>> HTTPS, the importance of getting the right IP starts to decrease quite a
>> bit: if you get the wrong IP address, then this turns into a connection
>> failure.
>>
>> -Ekr
>>
>
> This is an important assertion that requires refutation: this is incorrect.
>

I think this is a topic on which we are going to have to agree to disagree.

-Ekr


> The importance of the correct IP is essential, because of the relatively
> weak system under which TLS certificates are (or can be) issued.
> While individual CAs may be extremely trustworthy, there has been
> instances of other CAs (retrospectively discovered as being untrustworthy)
> issuing TLS certificates to parties other than the domain owner.
>
> This is why DNSSEC, and DANE, are an important mechanism available for
> strengthening certificate/browser security.
>
> DANE allows domain owners to securely link a specific CA or specific
> certificate to a domain name.
>
> Only with DNSSEC and DANE, is it true that the IP address being wrong
> would a connection failure be guaranteed to occur.
>
> That is, in fact, one of the underpinnings for why bad actors have an
> interest in altering DNS caches to provide forged IP addresses.
>
> Thus, I assert that not only is DANE not strictly orthogonal, it is an
> essential part of securing the web, along with DNSSEC.
>
> This is true even in a 100% HTTPS environment. The security of DNS as
> provided by DNSSEC is both orthogonal to, and more important than, the
> privacy of DNS recursive queries, when it to achieving that 100% HTTPS
> environment.
>
> Brian
>