On Tue, Mar 7, 2017 at 11:23 PM, Jakob Bohm via dev-security-policy < [email protected]> wrote:
> I saw nothing in Gervs posting suggesting that banning all kinds of > RA/DTP relationships was the intended effect. But would you also acknowledge that your originally stated "The point is NOT to prohibit RAs" is similarly not provided? I highlight this, because I want to make sure you're not prematurely dismissing policy suggestions because you disagree. > As to the rest of your message, you describe what have historically been >> called RAs/DTPs. You have not, however, answered my question - which is >> what impact, if any, do you believe would happen if Mozilla considers the >> policy I suggested: That is, to require that anything which 'historically' >> was considered an RA be treated as an externally operated sub-CA. >> > > For the kind of RA that is a combined reseller/control everything > related to issuing (the Symantec case), there would be no significant > change in burden for the RA. > > For the kind of RA that only does specific relevant parts of validation > (a "traditional" RA), the suggested policy as written would "simply" > require the CA to set up and maintain one (set of) subCAs for each of > their RAs, while your rephrasing as a ban on RA/DTP relationships would > impose the full cost of a formal WebTrust (etc.) audit on RAs that only > perform a specific limited function that could be audited much cheaper, > provided the CA systems were set up to have little dependency on > certificate specific activities and security at the RAs. > This misunderstands Policy 1 then, and is perhaps the substance of our unintentional disagreement. if a CA sets up and maintains one (set of) sub CAs for each RA, then each of those subCAs would need to be audited. This is no different than the existing requirements, within the Baseline Requirements, that each DTP be audited. I would highlight Section 8 of the Baseline Requirements for you, and ask that you clarify where, under the existing policies (e.g. ignoring any policy proposal) you believe there is any provision for allowing a DTP to be "audited much cheaper". If you believe such a thing is possible (I would argue incorrectly so, but hear me out), then what we're effectively saying is that a set of Principles and Criteria are not examined by the audit, because the operation and majority of the controls are managed by the "Issuing CA" - at least, within the world today of CA/RA relationships. If we believe such a thing is accepted (and again, not supported by the Baseline Requirements, and to the extent of my interactions with various auditor/practitioners, I do not believe currently supported by WebTrust), then I hope you can also see that we can use that self-same logic to conclude that a DTP operating as a externally operated sub-CA - but one in which the "Issuing CA" handles the operation and majority of controls for - is effectively the same thing. Do you agree with that logic? Can you clarify where and why you disagree with the analysis above? > For example, an RA whose sole involvement is to receive a daily list of > company name/idno/address/authorized signatory for pending > applications, go down to the state hall of records and report back > which ones match/do not match official company records (to support EV > certification for that state) would only need auditing of that activity > and the security of the system used to exchange that list and report > with the CAs central validation team. Please provide a citation to the Baseline Requirements or Mozilla policy to support this statement. I would suggest Section 8.4 provides counter-evidence to this claim, and as such, because the argument rests on this claim, needs to be addressed before we might make further progress. > They should not need to pay a > local WebTrust auditor to certify that the many activities done > centrally at the CA (in a different country altogether) meet any > particular requirements, as that should be in scope only for the actual > auditing of the central CA (which should include auditing of the > presence and compliance of the RA audit reports). > > For the "Enterprise customer" "DTP" case, I think subjecting each > Enterprise that has a "cheaply issue certs for subnames" account to any > kind of special audit would be as mindbogglingly onerous as requiring > every shop that accepts credit cards via a 3rd party such as Google to > go through and comply with full PCI certification for systems that > never see credit card numbers, only the shopping cart. For this case, the only thing that needs auditing is that the CA itself > has properly validated that the Enterprise customer has the full > authority to decide and state which of the acceptable subnames are who > they say they are (as per DNS zone delegations for DNSnames, per mail > domain ownership as per e-mail names etc.). In other words the only > thing the Enterprise customer is delegated to validate is those things > for which the Enterprise customer is the ultimate authority anyway. It sounds like you're describing how you think things work. Unfortunately, that's not consistent with how they're specified to work (again, Section 8) or how I've proposed for them to work (... again, Section 8). Worse, I chose the precise term of "Delegated Third Party" to avoid confusion with the explicitly-called out case within the Baseline Requirements regarding "Enterprise RA" I think what you're looking to do is try to base an argument using Section 1.3.2, in which you might have an opportunity to do so. However, what you've provided is unsupported and unsubstantiated claims, and I don't think those are useful for the discussion at all, because it's unclear their basis in reality. I realize that may sound unduly harsh, but I do believe for such a nuanced policy discussion, it's important to focus on what is _actually_ required by the current policies. I believe to do less than that is to do a disservice to the discussion and not to respect peoples' time. _______________________________________________ dev-security-policy mailing list [email protected] https://lists.mozilla.org/listinfo/dev-security-policy

