Re: [Doh] [Ext] Re: Use cases and URLs

[Patrick McManus <[e-mail]>]

On Wed, Mar 7, 2018 at 12:28 AM, Paul Hoffman <[e-mail]>
wrote:

> On Mar 6, 2018, at 20:04, Adam Roach <[e-mail]> wrote:
> >
> >
> >
> > Can you elaborate on this a bit? I fully expect that the typical
> user-self-provisioning case will involve copy/pasting server information,
> in which case the difference between a server and a URL is pretty minimal.
> Do you expect users to type server names in rather than copying them?
>
> Yes, exactly. If someone tells you that your bank runs this secure DNS
> server, that’s verbal, not copy and paste.
>
>
I don't agree with this.

Its logically an argument for not even using hostnames. The minimal
information in your example is your bank - so just assume www.BANK.com.
That's certainly easier to convey though it has some obvious shortcomings.

But even if you argue hostname is the goldilocks of granularity, the DNS
resolver market has already shown that its not rich enough. note that
quad-9 offers both 9.9.9.9 and 9.9.9.10 and that 10 is not simply a
secondary for .9 but its a different service with different policies and
different results.. but because the configuration scheme of traditional DNS
is so coarse they needed to burn IPv4 addresses just to convey
configuration information. Your suggestion uplevels that from 1 address to
1 hostname but the fundamental problem remains.

Lastly, we want this work to be consistent with the BCP56bis work which
explicitly talks about Initial URLs in 4.4.1. It says arbitrary URLs should
be used unless they are not practical. I would say if there were no
configuration at all we could talk about whether or not they were
impractical, but in the cases that are in scope we're definitely talking
about configuration.

-P