Re: [Acme] kinds of proof

[Eric Mill <[e-mail]>]

On Sat, Nov 29, 2014 at 10:29 PM, Paul Hoffman <[e-mail]>
wrote:

> On Nov 29, 2014, at 2:11 PM, Viktor Dukhovni <[e-mail]>
> wrote:
> > Sure, and the domain owner can field servers on whatever port he/she
> > wishes after demonstrating control over the domain, which to me
> > means control over the DNS (be it direct, or indirect via whoever
> > administers the DNS).
>
> I think this will have to be an "agree to disagree" situation. The Let's
> Encrypt promotional material indicates that they want to get more HTTPS out
> in the world, so they need to deal with the huge number of folks who use
> hosting companies and thus have no DNS control. I believe that is a great
> goal.
>

I agree with your assessment of LE's goal, and with the goal itself, but
aren't hosting companies themselves a meaningful target audience for ACME
and expanding HTTPS?

Anything that's in charge of your domain's DNS, or to whom you've delegated
control, should also be able to automate the provisioning of a free
certificate for you. No one's disputing that. So even if ACME did not prove
domain ownership by checking port 80, there'd be reason to believe it would
expand the playing field, if hosting companies found offering free
certificates a competitive feature.

That all said, I just read through the ACME draft spec[1], and none of the
listed proof-of-ownership methods involved just checking a path on port 80.

The HTTP-based proof mechanisms function by fetching a well-known URI over
port 443, and ensuring that a *valid self-signed certificate* is used to
make the connection and display the correct response. That self-signed cert
needs to use the same keypair that the server is attempting to validate for
use in making the CA-signed certificate.

That's smart, and completely removes concerns over things like
user-generated content hijacking a URI for someone else's server. To prove
you own a server without owning its DNS, you need to be able to generate a
keypair on the box and tell a webserver to use it -- the exact same control
you'd need to use Let's Encrypt in the first place -- and nothing less.

-- Eric

[1]
https://github.com/letsencrypt/acme-spec/blob/master/draft-barnes-acme.txt


>
> --Paul Hoffman
> _______________________________________________
> Acme mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/acme
>



-- 
konklone.com | @konklone <https://twitter.com/konklone>