On 4/24/2015 10:14 PM, Ryan Sleevi wrote: > On Fri, April 24, 2015 7:52 pm, David E. Ross wrote: >> If a root has already been added to the NSS database, we must assume >> that it has undergone the Mozilla process for that inclusion. The >> process involves looking not only at the root but also at the >> certification authority; at least that is what appears in both the >> public discussion of the request to add the root and in the stream of >> comments in the bugzilla.mozilla.org bug report that initiates the >> request. >> >> If a root in the NSS database is transferred to a new owner and that new >> owner already has roots in the NSS database, I assert the new owner has >> already undergone sufficient scrutiny. I limit my assertion, however, >> to cases where the transferred root has characteristics (e.g., trust >> bits, EV status) in common with roots already owned by the new owner. >> That is for example, I would not accept EV status on a transferred root >> if none of the other roots of the new owner have EV status. >> >> We trusted the old owner of the root and we trust the new owner. Thus, >> we can trust the transferred root to the minimum level we trusted the >> two owners. In the environment of OpenPGP, this is analogous to the Web >> of Trust. > > I know this is how you want it to work, and how people naively think it > works, but it isn't how it works. > > I tried to make the logical reasoning as explicit as possible: > > - If a root is added to the Mozilla store, it is because it wants to issue > publicly trusted certs. > - The community cares about the entities issuing publicly trusted certs. > - However, not all organizations that can issue publicly trusted certs go > through community review. > - Therefore, there are organizations who can issue publicly trusted certs > that do not go through community review. > > Thus any argument that says "All organizations that issue publicly trusted > certs are community reviewed" is false, both demonstrably and in practice, > and arguments predicated on this can be shown to be based on a mistaken > assumption. > > Your ascription of trust to the organization, rather than the root, also > mistakenly understands what the community review process does. The *WHO* > of the organization is entirely irrelevant, according to the policy. It is > the *WHAT* they're doing that matters. If the WHAT does not change, then > it should not matter that the WHO changes. > > Again, I thought this was spelled out, but the community review is of > policies and practices, not people. We don't keep the brown-eyed CAs out > to the benefit of the blue-eyed CAs, we don't keep the non-American out > for the American, we don't take the brands we use every day versus the > brands we've never heard of. That's not the community review. It may be > how you review, but it just means that those arguments are ignored, > because that's not how the policy works. > > If you find this still unsettling or uncomfortable, then the exercise is > what I asked of you originally - demonstrate how this functionally differs > from a root that signs an intermediate with "CA:TRUE" capability. > > Under the current policy, the root is required to: > 1) Disclose the intermediate > 2) Disclose the CP/CPS > 3) Disclose the audit, which they must have > > That is the current standard for what "Capable of issuing trusted > certificates" must meet. How does this differ - purely on a technical > level? It doesn't. >
However, all certification authorities whose root certificates are in the NSS database have indeed undergone community review. My assertion about trusting the new owner of a transferred root that is already in the NSS database is limited to a new owner whose other roots are also in the NSS database. I already indicated that, if the new owner does not already have roots in the NSS database, the transferred root should not be trusted. I also indicated that, if the new owner already has roots in the NSS database but the transferred root has higher trust settings than any of the other roots of the new owner, then the transferred root can have only trust settings not greater than the new owner's other roots. -- David E. Ross I am sticking with SeaMonkey 2.26.1 until saved passwords can be used when autocomplete=off. See <https://bugzilla.mozilla.org/show_bug.cgi?id=433238>. _______________________________________________ dev-security-policy mailing list dev-security-policy@lists.mozilla.org https://lists.mozilla.org/listinfo/dev-security-policy