[cabfpub] Ballot 190

[Rob Stradling]

On 02/05/17 16:52, Ryan Sleevi wrote:
>
>
> On Tue, May 2, 2017 at 11:44 AM, Rob Stradling <rob.stradling at comodo.com
> <mailto:rob.stradling at comodo.com>> wrote:
>
>     And if, as today, the Leaf cert doesn't contain 2.23.140.x.y.z, then
>     the same is true: the leaf would never validate with the
>     2.23.140.x.y.z OID in the user-initial-policy-set.  Right?  If so,
>     I'm not really sure why you think this approach would be "crude", tbh.
>
>
> Because it's asserting a policy OID that's not valid within the policy
> tree. Obviously, not asserting a policy OID that's not valid in the
> policy tree is no big deal, as there's an infinite set of those that we
> don't assert ;)
>
> Again, I want to stress that it should technically work. But it also
> comes with the side-effect that say, if a CP/CPS says "We only assert
> policy OIDs that are valid in the policy tree" (and I have seen some
> CP/CPSes or regulations to that effect, IIRC), then it means asserting
> the anyPolicy in the intermediate, which may or may not be desirable.

Ah.  I hadn't seen any such CP/CPSes.

> It's assuming that CAs' software even lets them assert such policy OIDs.
> I can totally see an issuance pipeline that refuses to allow issuance of
> a leaf policy if the intermediate doesn't assert a compatible policy
> (via policyMappings - which I have zero interest to support and it can
> die in a fire, or via anyPolicy, or an explicit policy). So it's not
> necessarily even a 'better' solution, from a "How deployable is this".
> The purpose of policy OIDs is to constrain through the issuance tree the
> validity, and this is 'hijacking' that because it's expressing something
> about the leaf that is explicitly _not_ constrained by the issuing
> intermediate.

That's the purpose of policy OIDs in intermediate certs, certainly. 
However, I think it'd be reasonable to interpret this as a stand-alone 
sentence:
   "In an end entity certificate, these policy information terms indicate
    the policy under which the certificate has been issued and the
    purposes for which the certificate may be used."
(This kinda reminds me of the "gap" in the spec re: the use of EKU OIDs 
in intermediate certs, and the resulting arguments that have spanned the 
last decade or two ;-) )

But it's a moot point really.  "How deployable is this?" is indeed the 
right question to ask, and the side-effect you just mentioned would be 
undesirable.

> The extension avoids that to some extent (because it's "just" an
> intermediate extension), but comes w/o any ability to apply constraints.
> It's a different trade-off, but one I think at least captures the
> 'current' thinking. I was trying to avoid the over-engineering
> discussion (that's possible with either solution) that could come about
> if, say, we wanted to distinguish "This intermediate from this CA only
> uses methods 4, 7, and 10" - (for which policy OIDs is absolutely the
> correct thing, but which clients would need to supply it in the
> user-initial-policy-set), and instead focus on "facts about the leaf" -
> for which an extension is entirely sufficient and avoids the extra
> engineering (which I've now introduced and I'm sure at least some will
> latch onto as a means of derailing conversion :P)

-- 
Rob Stradling
Senior Research & Development Scientist
COMODO - Creating Trust Online