On Mon, Oct 30, 2017 at 9:37 PM, Kathleen Moriarty <
[email protected]> wrote:

> Hi Richard,
>
> On Mon, Oct 30, 2017 at 6:27 PM, Richard Barnes <[email protected]> wrote:
> >
> >
> > On Mon, Oct 30, 2017 at 5:00 PM, Kathleen Moriarty
> > <[email protected]> wrote:
> >>
> >> Hi Richard,
> >>
> >> Thanks for posting the diff, I reviewed the changes and they look
> >> good, but I have a couple of questions.  FYI - I had started looking
> >> at the PR, but something came up, sorry for the delay and this was a
> >> good reminder.
> >>
> >> On Thu, Oct 19, 2017 at 12:45 PM, Richard Barnes <[email protected]> wrote:
> >> > Hi Kathleen,
> >> >
> >> > Thanks for the review.  Some responses inline.  I've started a PR to
> >> > respond
> >> > to these comments here:
> >> >
> >> > https://github.com/ietf-wg-acme/acme/pull/345
> >> >
> >> > I agree with Daniel that we should hold off on IETF LC until we've
> >> > addressed
> >> > the issues around proactive issuance / caching of CSRs.  That may
> >> > require a
> >> > re-WGLC, but once that's closed, it should be safe to send to IETF LC.
> >> >
> >> >
> >> > On Fri, Sep 22, 2017 at 4:14 PM, Kathleen Moriarty
> >> > <[email protected]> wrote:
> >> >>
> >> >> Hello,
> >> >>
> >> >> Thank you to the editors and WG for your efforts on
> >> >> draft-ietf-acme-acme, it's a well written and easy to understand
> >> >> draft.  I do have a few comments, that need to be address by the
> >> >> editors and SHEPHERD.
> >> >>
> >> >> Please review the idnits.  There are a few warnings that should be
> >> >> correctable and can be addressed, but more importantly, it calls out
> >> >> several downrefs that are not in the shepherd writeup.  These need to
> >> >> be in the writeup and then also mentioned in the IETF last call
> >> >> announcement (I'll make sure the latter happens).
> >> >>
> >> >> Here's my mostly editorial comments:
> >> >>
> >> >> Introduction:
> >> >> It reads very well, but it would be helpful to those unfamiliar with
> >> >> the work to explicitly state that ACME is about DV certificates.  The
> >> >> first sentence of the last paragraph seems like the best place to add
> >> >> this in.
> >> >> Current text:
> >> >>    This document describes an extensible framework for automating the
> >> >>    issuance and domain validation procedure, thereby allowing servers
> >> >>    and infrastructural software to obtain certificates without user
> >> >>    interaction.
> >> >> Proposed:
> >> >>    This document describes an extensible framework for automating the
> >> >>    issuance and domain validation procedure for DV certificates,
> >> >> thereby allowing servers
> >> >>    and infrastructural software to obtain certificates without user
> >> >>    interaction.
> >> >
> >> >
> >> > As others have pointed out, ACME is not just about DV.  PR has some
> >> > clarification on this point.
> >>
> >> I see the text on uses in other PKI contexts in the introduction, but
> >> don't see anything on the certificate types and out-of-band processes
> >> (as is the case for the STIR use case Mary mentioned).  Maybe I'm
> >> missing it, can you tell me where to look?
> >
> >
> > I don't think you're missing anything.  What do you mean by "certificate
> > types and out-of-band processes"?  How is that different from "other PKI
> > contexts"?
>
> I think by PKI-context, you mean the application of use, is that
> right?  I am separating out the type of certificate (DV, EV, etc.) as
> you have done in the second paragraph of the introduction.  When
> reading the rest of the document, there isn't text that says what type
> of certificate ACME produces without any additional processes
> (out-of-band authentication/validation).  If you think about
> administrators who operate servers (whatever the protocol) or even
> those who may wish to incorporate ACME into a new application, they
> will may have to figure out what level ACME provides as a default and
> also know that there can be additional processes added to obtain
> certificates types at a higher assurance level using ACME.  Since the
> text already does a nice job of explaining the certificate types, this
> should be an easy follow up that I think will be very helpful to
> people outside of the IETF or who look at this in the future.
>
> Maybe something like the following:
>
>    Strictly following the automation enabled by ACME, DV certificates
> are issued.
>    Additional processes can be provided by the CA to increase the level of
>    validation on an end entity to issue certificates at a higher
> assurance level,
>    for instance an EV certificate.  <Then maybe provide a reference to the
> STIR
>    example Mary mentioned in this thread.>
>
> Does that help?  Thinking about this in former roles and having been
> to some conferences lately, I think being explicit about this will
> help those trying to evaluate and deploy ACME.  They may look for
> providers who meet the assurance levels they need knowing there are
> options and understanding the baseline provided by this spec.
>


It's wrong to frame this as being about assurance levels -- you seem to be
presuming that manual processes can give higher assurance than automated
ones, a fact which I think has been repeatedly proven false in the web
PKI.

But I'm fine clarifying that certain things (like EV validation) have not
yet been audited:

+ACME can also be used to automate some aspects of certificate management
even
+where non-automated processes are still needed.  For example, the external
+account binding feature (see {{external-account-binding}}) can be used to
+associate authorizations with an account that were not validated through
the
+ACME authorization process.  This allows ACME to address issuance
scenarios that
+cannot yet be fully automated, such as the issuance of Extended Validation
+certificates.

Updated PR: https://github.com/ietf-wg-acme/acme/pull/348

--Richard




>
> Thanks,
> Kathleen
>
>
> >
> >
> >>
> >> >
> >> >
> >> >>
> >> >> I do like how the introduction is framed as it’s clear the level of
> >> >> security provided by the certificates issued via ACME.  Thanks for
> >> >> that.
> >> >>
> >> >> The introduction should also make it clear that ACME can be used for
> >> >> other services using DV certificates, not just HTTP.  This is
> >> >> mentioned in the Terminology section, but I think a clear sentence
> >> >> upfront would be helpful.
> >> >>
> >> >> Section 2: Nit/suggestion: Change the tense to read as a published
> RFC
> >> >> in the last paragraph, last sentence (take it or leave it).
> >> >> Current:
> >> >>    Such close integration of ACME with HTTPS
> >> >>    servers would allow the immediate and automated deployment of
> >> >>    certificates as they are issued, sparing the human administrator
> >> >> from
> >> >>    much of the time-consuming work described in the previous section.
> >> >> Proposed:
> >> >>    Such close integration of ACME with HTTPS
> >> >>    servers allows the immediate and automated deployment of
> >> >>    certificates as they are issued, sparing the human administrator
> >> >> from
> >> >>    much of the time-consuming work described in the previous section.
> >> >
> >> >
> >> > Done.
> >> >
> >> >
> >> >>
> >> >> IANA Section:
> >> >> Everything looks good except that RFC5226 has been obsoleted, so the
> >> >> new reference is RFC8126 and “specification required”
> >> >> https://tools.ietf.org/html/rfc8126#section-4.6 still means the same
> >> >> thing, so using that is fine.
> >> >
> >> >
> >> > Updated reference.
> >> >
> >> >
> >> >>
> >> >> Security considerations:
> >> >> I think it would be good to mention here that they are DV certs so
> the
> >> >> reader understands from he introduction and this section the level of
> >> >> cert issued through ACME and doesn’t assume a higher level of
> >> >> assurance.
> >> >>
> >> >> s/ACME is a protocol for managing certificates/ACME is a protocol for
> >> >> managing DV certificates/
> >> >
> >> >
> >> > Given the discussion about DV, I don't think this change is
> appropriate.
> >> >
> >> >
> >> >>
> >> >> 10.1 - I know this is obvious to the people in the WG, but a
> reference
> >> >> to RFC7525 to properly configure TLS 1.2 should be included to
> protect
> >> >> against the mentioned attacks.  If you want to also reference TLS 1.3
> >> >> and specify no 0-RTT that would be good too.  It’d be great to see
> >> >> this get widely deployed, so there may be a number of newcomers
> >> >> reading this to make their lives easier with cert
> >> >> issuance/maintenance.
> >> >
> >> >
> >> > I don't see how TLS configuration plays into the security
> considerations
> >> > for
> >> > ACME.  How you use the certificates you get from ACME is your
> business,
> >> > and
> >> > indeed, several of the envisioned applications have nothing to do with
> >> > TLS
> >> > (e.g., the email and STIR use cases).
> >> >
> >> > That said, I have added some language to suggest that ACME servers
> >> > should
> >> > follow the advice in RFC 7525.  0xRTT should be safe here, because
> there
> >> > are
> >> > anti-replay protections at the application layer.
> >>
> >> Can you add a pointer in your text about the provided anti-replay
> >> protections in the application layer?
> >
> >
> > Draft PR here:
> >
> > https://github.com/ietf-wg-acme/acme/pull/348
> >
> > --Richard
> >
> >
> >>
> >>
> >> Thanks,
> >> Kathleen
> >>
> >> >
> >> >
> >> >>
> >> >> 10.2 - what if the server that the account verifier has an account on
> >> >> (client for ACME) is compromised and is used to request new
> >> >> certificates or perform other actions?  I think this is one of the
> >> >> larger risks, so mentioning this is possible and hardening measures
> >> >> should be taken to prevent compromise would be prudent.  Hardening
> >> >> measures or appropriate security controls should be broadly
> understood
> >> >> terms (I think) so that you wouldn’t need to list out things like
> >> >> turing off unnecessary services.
> >> >
> >> >
> >> > Added a brief note.
> >> >
> >> >
> >> >>
> >> >> 10.5 - I would think you’d see requests for new phishing domains
> >> >> rather than known ones through this process.  What would you look for
> >> >> there as that’s a complex problem - it would be hard to know all
> legit
> >> >> names to know if one was a play off of the original.  This happened
> >> >> with equifax - they provided a site name with a typo and a ‘white
> hat’
> >> >> attacker had set up a site that looked just like theirs at that site.
> >> >> But it was announced by equifax.  Tough problem as automating a check
> >> >> on the DNS registration may not detect something unusual, but maybe
> >> >> recommending this check could help?  RDP lessens what needs to be
> >> >> shared from whois though, although I don’t think that’s widely
> >> >> deployed.
> >> >
> >> >
> >> > This is a policy question, not one for the issuance protocol.
> >> >
> >> > --Richard
> >> >
> >> >
> >> >>
> >> >>
> >> >> --
> >> >>
> >> >> Best regards,
> >> >> Kathleen
> >> >>
> >> >> _______________________________________________
> >> >> Acme mailing list
> >> >> [email protected]
> >> >> https://www.ietf.org/mailman/listinfo/acme
> >> >
> >> >
> >>
> >>
> >>
> >> --
> >>
> >> Best regards,
> >> Kathleen
> >
> >
>
>
>
> --
>
> Best regards,
> Kathleen
>
_______________________________________________
Acme mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/acme

Reply via email to