[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Certificate Policies (was Re: Trivial PKI Question)



Thanks to Santosh and Roger for correcting me on the
CP/CPS distinction. I apologize for my error. I don't
claim to be an expert on CP/CPS. For that, I defer to
Santosh and others.

In any case, we all agree that relying party software
cannot read and evaluate the text of the CP or CPS.
That's why the Certificate Policies extension uses OIDs
to identify policies in a machine-readable manner.

In his most recent email, Anders said:
> I'm sorry that I mixed CP and CPS but I would like to extend
> the scope of my previous "policy-rant" to include all legal
> information inserted in EE-certificates whatever it is called
> and whatever its purpose may be.

Since this is just a rant, I should ignore it. And I'm
no legal expert, so I can't say why lawyers want to
refer to the text of a CP or CPS in a certificate. But
as long as they don't use too many bits for it (by
including the actual text instead of a URL), I don't mind.

Anders also stated that "all commercial CAs of any
significance" use a separate CA for each policy. This
is probably true. So far, most relying party software
doesn't support the Certificate Policies extension. It
would be pointless or even dangerous for large commercial
CAs to ignore this. But that doesn't mean it must always
be so. If we refuse to work on anything that's not already
widely deployed, we'll have to stop all innovation.
Instead, we must continue to address real problems with
real solutions. If customers see value in these solutions,
they will be implemented by vendors.

Certificate policies have definite utility. Several
communities are piloting their use, such as the U.S.
Government. Without certificate policies, they would
need to maintain a separate network of CAs for every
policy and require identical policies throughout each
network. Using certificate policies allows them to
have a single network of CAs with various policies
in use, some within only a subnetwork.

Anders also said it is very hard to "standardize legal
stuff". I agree. That's one reason why PKI deployment
has been slow. In some contexts, it's not necessary to
agree on what a certificate means. But in most of the
contexts that really care about security, it *is*
necessary to agree. That's when CPs and CPSs become
important.

One emerging solution is to have a consortium develop
a common CP and run a bridge CA. When organizations ask
to cross-certify with the bridge, they must agree to the
CP and demonstrate that they have a CP and CPS that is
stringent enough for the bridge CA to accept it as
equivalent. This is a time when policy mapping becomes
especially important.

As I said above, I should probably just ignore statements
on this list that are misinformed, misguided, or misleading.
That's what most people on the list seem to do. But I
worry that leaving such statements unchallenged will lead
people to conclude they are correct. So I'll try my best
to answer these comments, when I can find the time. Please
correct me if my answers are wrong.

Thanks,

Steve