Gervase Markham wrote:
> My proposal is that we accept such CAs, but use this technical 
> capability to restrict them to signing certificates for domains under 
> the appropriate TLD. The logic is that citizens of those countries have 
> to trust their government anyway, but that citizens of other countries 
> should not be forced to.

I think there is one slightly weak point in that logic, what about 
citizens from other countries that access a web site under the given TLD 
? They will automatically trust it without warning under your scheme, 
but it's not their governement, will they be expecting that ?

That's why I'm not sure the option of restricting to one TLD makes so 
much sense, and think that it would be better to find a scheme where it 
can be clearly justified why they should be trusted without such 
restrictions.

> Note that both CAs have been accepted, unrestricted, into the Microsoft 
> Root Program, on the basis of "trust us, we did the audit" letters 
> written by the respective governments.

> A useful thought experiment might be to ask what would happen if a CA 
> from North Korea were to apply for inclusion under the same types of 
> condition.

After some good thinking about it, it think this kind of waiver 
provision makes sense and is a good way out of those situations, *if* 
the entity requesting the waiver has such strong background on 
cryptography matters that questioning if they should be qualified to do 
that sounds a bit ridiculous, and if you have a very precise way of 
defining that (as a result of course it will exclude North Korea and 
corruptible third world countries).
The very best would be if you had a way to define that where US, EU and 
other major industrial countries would all together officially say this 
entity is very trustworthy.

So how to do that ?

I think that given the Mozilla ca certificate policy, the most 
intellectually satisfying answer is :
        Consider it's the same people who are already considered as having 
enough authority to authorize someone to perform CA audits (let say it 
twice : not to do audits, but to authorize someone to do them).

Unfortunately, point 9 in 
http://hecker.org/mozilla/ca-certificate-policy is not extremely precise 
to this regard. And I'd like to have a very well defined list. I 
understand the purpose is to be open, but as it stands, I'm also not 
sure it handles the North Korea case extremely well.

But the documents the policy explicitly reference in point 8 (Webtrust, 
ETSI profiles) do provide very precise criteria in this regard.

If you want to do a WebTrust audit, you need to be entitled to by WebTrust.

And more interesting and even more adequate, the official use of the 
ETSI profile document can only be within the frame of Common Criteria. 
That means to do recognized audits based on then you need to receive 
entitlement from one of the Certificate Authorizing entities within the 
ccra members of Common Criteria :
http://www.commoncriteriaportal.org/public/expert/index.php?menu=6

If you check it, you'll see that the list of Certificate Authorizing 
entities is a very short list, this criteria very likely falls more in 
the too conservative side than the too open. An entity can appear in the 
list only if existing members of Common Criteria (EU, US, etc.) are 
ready to accept cryptographic products it certifies as fully trustable 
for cryptography in their own country.

So I'll resume the whole reasoning above in one sentence : If you don't 
trust WebTrust and Certificate Authorizing entities to issue such 
waiver, why do you trust them to authorize people to do audit that will 
be recognized under the Mozilla CA certificate policy ?

If this can be agreed upon, things get a lot easier for the first 
request because as luck would have it, DCSSI is the Certificate 
Authorizing entity for France ;-)
And probably the Koreans provide you with a waiver request coming from 
ITSCC ?
_______________________________________________
dev-tech-crypto mailing list
dev-tech-crypto@lists.mozilla.org
https://lists.mozilla.org/listinfo/dev-tech-crypto

Reply via email to