First, just to review: Now that the CAB Forum has published version 1.0 
of the EV guidelines:
http://www.cabforum.org/EV_Certificate_Guidelines.pdf
we've had a number of CAs asking for their root CAs to be marked as 
capable of issuing EV certificates, so that they can be accorded any 
special UI present in Firefox 3 and related products to display identity 
information in SSL certs. (See for example bugs 398944 and 383183.) 
However just as we have a formal policy for deciding whether to add a 
particular root CA certificate for "normal" use, we should also have a 
formal policy for marking a root CA's certificate as EV-capable (as 
noted by Eddy Nigg and others).
As noted in the bug, I think an EV-enabled root CA cert is simply a 
special case of root CA certs in general, so we don't need a whole new 
separate policy. At the same time I don't want to revise every section 
of the existing policy, and if possible I'd like to avoid changes that 
necessitate renumbering and reorganizing the current sections of the 
policy. I'm therefore leaning toward having an EV addendum to the 
policy, and putting all the EV-related stuff there. Then we could simply 
modify section 6 ("We require ...") to add an additional paragraph 
pointing to the addendum. This would result in a version 1.1 of the 
overall Mozilla CA cert policy.
In terms of the addendum itself, obviously we can reference the CAB 
Forum guidelines document (formally, "Guidelines for the Issuance and
Management of Extended Validation Certificates, Version 1.0") as the 
governing criteria. There's a broader question of whether in theory we 
could or should accord "EV-style" treatment to CAs that don't strictly 
speaking conform to the guidelines, but conform to other guidelines 
deemed to be equivalent. I'd like to declare that question out of scope 
for now, as there's no obvious candidates today for alternative 
guidelines (at least AFAIK).
In terms of audits associated with "EV-ness", I'm a little unclear on 
what other documents need to be referenced. Section J of the EV 
guidelines spells out the high-level audit requirements: basically 
either go through the WebTrust EV process or a process deemed as 
equivalent by the CAB Forum. There's a document "WebTrust for 
Certification Authorities - WebTrust Extended Validation Audit Criteria" 
on the CAB Forum web site; however it's marked as draft and the 
guidelines themselves don't mention it by name AFAICT. (The guidelines 
instead use the term "WebTrust EV Program".) My initial conclusion is 
that we don't need to reference the WebTrust draft document, but can 
confine ourselves to referencing the relevant section(s) of the guidelines.
Finally, I'm open to suggestions on other possible changes to the 
Mozilla CA certificate policy unrelated to EV certs. However I reserve 
the right to postpone such consideration of such changes to a future 
version of the policy (e.g., 1.2) if there's no immediate strong 
consensus on the need for any such change and the associated "patch" to 
the policy itself. My primary goal is to address the EV-related policy 
changes, and to do so as expeditiously as possible.
Anyway, if you have comments on this general topic please feel free to 
post them here. In the meantime I'll work to come up with an initial 
draft of proposed changes to the policy text, and will post that to the 
bug when done.
Frank
-- 
Frank Hecker
hec...@mozillafoundation.org
I believe that the two things are supposed to be equivalent; the 
different language is just sloppy drafting. As I understand it, this 
document is going to be finalised soon.
> My initial conclusion is 
> that we don't need to reference the WebTrust draft document, but can 
> confine ourselves to referencing the relevant section(s) of the guidelines.
Without an audit, how do we assure compliance?
One of the big advantages of EV is that we have a minimum standard for 
vetting that is actually enforced by audit - i.e. we don't have to 
assess the vetting practices of every CA (even if they would tell us 
what they were), because someone else has done it for us.
It seems to me that it makes sense to leverage that, by saying that our 
criteria for EV enablement is a passed WebTrust EV Audit.
Gerv
Gervase Markham wrote:
> One of the big advantages of EV is that we have a minimum standard for 
> vetting that is actually enforced by audit - i.e. we don't have to 
> assess the vetting practices of every CA (even if they would tell us 
> what they were), because someone else has done it for us.
>
> It seems to me that it makes sense to leverage that, by saying that our 
> criteria for EV enablement is a passed WebTrust EV Audit.
>   
I just would like to remind you about the promise made at the beginning 
of this year (at and after the conference call), that Mozilla will work 
at the CAB forum for alternative (and definition of equivalent) third 
party audit other than webtrust and/or implement its own alternative 
binding for Mozilla products only.
Since the guidelines for the EV audit are published in addition to the 
EV guidelines for CAs themselves, capable audit firms could perform this 
task without the CA having to use a webtrust accredited auditor. This 
should only have an affect on the choice of acceptable auditors and not 
reduce the value of EV and the audit itself. The EV policy extension 
could be the place for such a definition.
-- 
Regards 
 
Signer:  	Eddy Nigg, StartCom Ltd. <http://www.startcom.org>
Jabber:  	star...@startcom.org <xmpp:star...@startcom.org>
Blog:  	Join the Revolution! <http://blog.startcom.org>
Phone:  	+1.213.341.0390
 
I wasn't implying that we would require an audit; see below.
> It seems to me that it makes sense to leverage that, by saying that our 
> criteria for EV enablement is a passed WebTrust EV Audit.
But that's being more specific than what the guidelines actually say. 
(This is relevant to Eddy's point as well.)
Let's suppose we have language like the following (note that this is 
*not* proposed draft language at this point, it's just me thinking out 
loud):
   In order for its root CA certificate to be marked as EV-capable, the
   CA must comply with the "Guidelines for the Issuance and Management
   of Extended Validation Certificates, Version 1.0" as published by the
   CAB Forum.
Inherent in that compliance is compliance to section 4(a)(3) of the 
guidelines:
   (3) Comply with the requirements of (i) the then-current WebTrust
   Program for CAs, and (ii) the then-current WebTrust EV Program, or an
   equivalent for both (i) and (ii) as approved by the CA/Browser Forum;
as well as compliance to section 35 of the guidelines, which has similar 
language but goes into more detail. So I think the requirement for an 
audit is inherent in the requirement for compliance with the guidelines.
If we wanted to emphasize the audit requirement, we could say something like
   ... CA must comply with the "Guidelines for the Issuance and
   Management of Extended Validation Certificates, Version 1.0",
   including the requirements of section 35, "Audit Requirements".
The above is why I don't think we need to reference the WebTrust 
document specifically. Your thoughts?
Sorry, s/would/would not/
(Those double negatives will trip you up every time :-)
I'll consider myself reminded :-)
But how should that affect this current proposal to change the Mozilla 
CA policy? As I noted in my previous message, I think requiring 
compliance to the CAB Forum guidelines implies compliance with the audit 
requirements of those guidelines, and those guidelines already allow for 
the possibility of non-WebTrust audits. ("...or an equivalent for both 
(i) and (ii) as approved by the CA/Browser Forum", to quote the relevant 
clauses of section 35.)
We then have the following alternative options:
A. Change our policy to adopt language like I suggested in my last 
message, and then separately work through the CAB Forum to get some 
non-WebTrust audit programs approved by the CAB Forum as "equivalent" to 
the WebTrust programs.
or
B. Try to come up with an alternative audit program ourselves, and 
change our policy to allow that program to satisfy our own requirements 
for EV certs, independent of what the CAB Forum does or doesn't do.
My personal preference is to go with option A. First, I don't want to 
gate updating our policy on solving the problem of defining a 
WebTrust-equivalent EV audit regime. Second, I would prefer that we use 
the standard CAB Forum mechanisms to try and address this issue.
Mozilla has voted in favor of the guidelines and some time has past 
since. I haven't been updated if anything happened in that direction, 
but supposed that nothing could be advanced as of now, we should define 
when/where/what happens if there is no solution found to the 
"WebTrust-equivalent EV audit regime" problem (your words ;-) ). Your 
option B could be a very likely solution in such as case.
After thinking about it, I think it may be possible to do this just by 
adding a final paragraph to section 6 of the current policy:
   In addition, if a CA wishes its certificate to be marked to note that
   Extended Validation certificates may be issued under the associated CA
   hierarchy then we require that the CA comply with the "Guidelines for
   the Issuance and Management of Extended Validation Certificates,
   Version 1.0" (as modified by the erratum published by the CAB Forum),
   and have its compliance attested to in accordance with the
   requirements of Section J of that document.
I've attached a proposed patch to this effect to bug 399214.
Your comments are welcome.
I think that's fine in terms of stating the qualifications.  However,
I think there are also procedural questions/issues to be addressed.
During his term as administrator of the Root CA cert policy, Gerv created
a questionnaire that he routinely asked CAs to answer in their applications.
I think it would be good to publish that along with the formal policy.
To handle EV CAs, that questionnaire needs to be extended slightly.
It needs to ask, for each root cert being put forward, whether that cert
will be a root for EV certs, and if so, what is the EV policy OID that
will be used in all EV certs that chain up to that root?
-- 
Nelson B
Shouldn't be this be part of section 8 as this is a criteria for CA 
operations (and its associated audits).
After reading your proposal I would suggest alternatively the following:
_Under section 7_ "We consider verification of certificate signing 
requests to be acceptable if it meets or exceeds the following 
requirements:"
Add a fourth part with something like this:
"for a certificate to be used for and marked as Extended Validation the 
CA complies with the "Guidelines for the Issuance and Management of 
Extended Validation Certificates, Version 1.0" (as modified by the 
erratum published by the CAB Forum), and have its compliance attested to 
in accordance with the requirements of Section J of that document."
(As a by-note, I'm note sure the version number should be used and or 
include any future versions instead. Something like "Guidelines for the 
Issuance and Management of Extended Validation Certificates according to 
the most actual version as published by the CAB forum....")
_Under section 8_ "We consider the criteria for CA operations published 
in any of the following documents to be acceptable:"
Add the criteria and guidelines for the Issuance and Management of 
Extended Validation Certificates as published by the CAB forum.
(Obviously after performing an EV audit, this should allow to issue 
regular non-EV certificates as well.)
Additionally I would like to see somewhere a disclaimer which would 
allow Mozilla in the future to accept alternative auditors other than 
the ones approved by the CAB forum for EV without the need to change the 
Mozilla CA policy again in order to allow that. This doesn't mean that 
Mozilla _must_ accept, but would keep the option open in the same way as 
the policy has other such disclaimers concerning different aspects.
Sounds fine. :-) If, in some strange universe, the CABForum decided not 
to require audits for compliance any more, then we could just update our 
policy again.
Gerv
My apologies for the long delay in responding to your message. I am now 
ready to take up this issue again, and hope to conclude it soon.
> After reading your proposal I would suggest alternatively the following:
> 
> _Under section 7_ "We consider verification of certificate signing 
> requests to be acceptable if it meets or exceeds the following 
> requirements:"
> 
> Add a fourth part with something like this:
<snip>
I agree that section 7 is a good place for this (since EV practices 
primarily address the verification steps), and is perhaps better than 
putting it in section 6 as I previously proposed.
> "for a certificate to be used for and marked as Extended Validation the 
> CA complies with the "Guidelines for the Issuance and Management of 
> Extended Validation Certificates, Version 1.0" (as modified by the 
> erratum published by the CAB Forum), and have its compliance attested to 
> in accordance with the requirements of Section J of that document."
> 
> (As a by-note, I'm note sure the version number should be used and or 
> include any future versions instead. Something like "Guidelines for the 
> Issuance and Management of Extended Validation Certificates according to 
> the most actual version as published by the CAB forum....")
After thinking about it, my proposal is to use the text as you give it, 
except I would delete the "Version 1.0" reference. I'm not exactly sure 
how to word a provision specifying use of any version, or the current 
version as of the time of the CA's application, or whatever, so I'm not 
going to try that approach.
One issue here is that the CAB Forum didn't version the URLs for the 
guidelines and the erratum. I don't know if future versions of the 
guidelines will have a different URL or not, so I thought the best 
approach was to omit the version number and use the current URL for the 
guidelines.
> _Under section 8_ "We consider the criteria for CA operations published 
> in any of the following documents to be acceptable:"
> 
> Add the criteria and guidelines for the Issuance and Management of 
> Extended Validation Certificates as published by the CAB forum.
Your implication here is that the CAB Forum EV guidelines can be used as 
stand-alone guidelines comparable to the other criteria referenced in 
section 8 (WebTrust for CAs, ETSI TS 101 456 and 102 242, and ANSI 
X9.79). I'm not convinced that this is in fact the case. In this 
connection, note that the WebTrust for CAs EV Audit Criteria document is 
explicitly characterized as a supplement to the base WebTrust for CAs 
criteria, not as a replacement for or superset of the traditional 
WebTrust for CAs criteria. (See for example sections 7-9 of the WebTrust 
EV document.)
> (Obviously after performing an EV audit, this should allow to issue 
> regular non-EV certificates as well.)
Again, the current WebTrust approach is to audit to both the traditional 
WebTrust criteria and to the WebTrust EV criteria, either simultaneously 
or separately.
Because I'm not convinced that the CAB Forum EV guidelines does or can 
function as a standalone set of criteria for CAs, I'd like to omit 
mentioning EV stuff in section 8 of the Mozilla CA policy, at least in 
this version of the policy. We can revisit this question later once I 
and/or others have a chance to do a detailed comparison of the EV 
guidelines to the other criteria already referenced in the policy.
One possible approach would be to explicitly mention the WebTrust EV 
criteria in section 8 as a supplement to the referenced WebTrust 
criteria. We could do the same thing for the ETSI criteria if/when an EV 
supplement document is created for those criteria. (IIRC someone is 
looking to create something like this.)
However at least for now I'd prefer not to mention the WebTrust EV 
criteria (or any other EV audit criteria) explicitly, for the reason 
I've discussed before: The WebTrust EV audit criteria and any other EV 
audit criteria are in effect incorporated by reference into section J of 
the EV guidelines, and so don't really need to be separately referenced.
> Additionally I would like to see somewhere a disclaimer which would 
> allow Mozilla in the future to accept alternative auditors other than 
> the ones approved by the CAB forum for EV without the need to change the 
> Mozilla CA policy again in order to allow that.
I don't think adding such a disclaimer really saves any work on our 
part. If we included such a disclaimer then if we ever wanted to accept 
alternative EV auditors we'd still have to have public discussion and 
consensus on which auditors we'd accept -- most likely the exact same 
discussion we'd have if we had to change the actual policy.
We have the freedom to set our own policy in any case, and all other 
things being equal I'd prefer not to include such a disclaimer at this 
time. I'd prefer to let the CAB Forum alternative auditor discussions go 
forward and see where they lead. If the discussions don't go anywhere 
then we can revisit the issue.
Frank
P.S. I've attached a new proposed patch to the CA policy document to bug 
399214.
-- 
Frank Hecker
hec...@mozillafoundation.org
Frank Hecker wrote:
>
> My apologies for the long delay in responding to your message. I am now 
> ready to take up this issue again, and hope to conclude it soon.
>   
Relax, I'm kinda used to it from you ;-)
> I agree that section 7 is a good place for this (since EV practices 
> primarily address the verification steps), and is perhaps better than 
> putting it in section 6 as I previously proposed.
>   
+1
>> _Under section 8_ "We consider the criteria for CA operations published 
>> in any of the following documents to be acceptable:"
>>
>> Add the criteria and guidelines for the Issuance and Management of 
>> Extended Validation Certificates as published by the CAB forum.
>>     
>
> Your implication here is that the CAB Forum EV guidelines can be used as 
> stand-alone guidelines comparable to the other criteria referenced in 
> section 8 (WebTrust for CAs, ETSI TS 101 456 and 102 242, and ANSI 
> X9.79). I'm not convinced that this is in fact the case. In this 
> connection, note that the WebTrust for CAs EV Audit Criteria document is 
> explicitly characterized as a supplement to the base WebTrust for CAs 
> criteria, not as a replacement for or superset of the traditional 
> WebTrust for CAs criteria. (See for example sections 7-9 of the WebTrust 
> EV document.)
>   
But doesn't it imply - in this case - an AICPA audit. To all of my 
knowledge it does, hence my suggestion to add it as a valid criteria. 
Also in your own words you say that it's a supplement and not a 
replacement. In my opinion an EV audit *extends* the traditional 
criteria and  *implements* EV.
> Again, the current WebTrust approach is to audit to both the traditional 
> WebTrust criteria and to the WebTrust EV criteria, either simultaneously 
> or separately.
>
> Because I'm not convinced that the CAB Forum EV guidelines does or can 
> function as a standalone set of criteria for CAs, I'd like to omit 
> mentioning EV stuff in section 8 of the Mozilla CA policy, at least in 
> this version of the policy. We can revisit this question later once I 
> and/or others have a chance to do a detailed comparison of the EV 
> guidelines to the other criteria already referenced in the policy.
>   
OK
> One possible approach would be to explicitly mention the WebTrust EV 
> criteria in section 8 as a supplement to the referenced WebTrust 
> criteria. 
Possibly an option, so this wasn't my point...
> We could do the same thing for the ETSI criteria if/when an EV 
> supplement document is created for those criteria.
OK
>
> However at least for now I'd prefer not to mention the WebTrust EV 
> criteria (or any other EV audit criteria) explicitly, for the reason 
> I've discussed before: The WebTrust EV audit criteria and any other EV 
> audit criteria are in effect incorporated by reference into section J of 
> the EV guidelines, and so don't really need to be separately referenced.
>   
Correct.
> I don't think adding such a disclaimer really saves any work on our 
> part. If we included such a disclaimer then if we ever wanted to accept 
> alternative EV auditors we'd still have to have public discussion and 
> consensus on which auditors we'd accept -- most likely the exact same 
> discussion we'd have if we had to change the actual policy.
>   
No problem! Lets have the changes to the Mozilla CA policy done first 
and continue with this later on...
> We have the freedom to set our own policy in any case, and all other 
> things being equal I'd prefer not to include such a disclaimer at this 
> time. I'd prefer to let the CAB Forum alternative auditor discussions go 
> forward and see where they lead. If the discussions don't go anywhere 
> then we can revisit the issue.
Agreed!
I think I understand what you mean here: Are you saying that the 
WebTrust EV criteria in effect incorporate the traditional WebTrust 
criteria by reference, since a WebTrust EV audit has as a prerequisite a 
regular WebTrust for CAs audit?
This is a subtle point, and one that could be argued both ways IMO. The 
relevant statement in the WebTrust EV document reads: "WT EV Audit 
Guidelines are to be used only in conjunction with the Principles and 
Criteria in the WebTrust Program for Certification Authorities. CAs that 
wish to issue EV Certificates must first go through a WT audit and then 
a WT EV audit." I don't read that language as implying "incorporation by 
reference", but it is indeed clear that you can't have a WebTrust EV 
audit without a WebTrust for CAs audit. (For example, it wouldn't make 
sense to have an audit against ETSI TS 101 456 and then try to do a 
WebTrust EV audit as a supplement to that; presumably no 
WebTrust-authorized auditor would agree to do this.)
> Also in your own words you say that it's a supplement and not a 
> replacement. In my opinion an EV audit *extends* the traditional 
> criteria and  *implements* EV.
I'd agree with that statement.
After thinking about it, if we want to reference the WebTrust EV 
criteria in section 8, I think the best way to do this would be to add a 
final item to the list in section 8 as follows:
   We consider the criteria for CA operations published in any of the
   following documents to be acceptable:
...
   * "WebTrust for Certification Authorities - Extended Validation Audit
     Criteria" in <a href="...">WebTrust for Certification Authorities -
     Extended Validation Audit Criteria</a> (in conjunction with
    "WebTrust Principles and Criteria for Certification Authorities")
(Note that the item in quotes is the relevant section within the 
linked-to document of the same title. Compare this usage to other items 
in the section 8 list.)
I think the above language clarifies that the WebTrust EV audit criteria 
  don't stand on their own but imply use of the standard WebTrust 
criteria as well.
Your thoughts?
OK, I think we're on the same page now.
>> I think the above language clarifies that the WebTrust EV audit 
>> criteria   don't stand on their own but imply use of the standard 
>> WebTrust criteria as well.
>>
>> Your thoughts?
>>   
> +1
I've gone ahead and attached a new proposed patch to bug 399214. Check 
out the URL
http://hecker.org/private/index-ev3
to see what the resulting 1.1 policy would like.
Maybe I'm being optimistic, but I think we're getting closer to a final 
1.1 version. If anyone has any further comments on the current draft 
please post them.
My apologies, I forgot about this. See my latest patch attached to bug 
399214. Basically I added a new item in section 14 as follows:
14. ... The request should include the following:
...
     * for each CA certificate requested for inclusion, whether the CA
       issues Extended Validation certificates within the CA hierarchy
       associated with the CA certificate and, if so, the EV policy OID
       associated with the CA certificate;
The proposed patch also contains some other new changes, but this is the 
only substantive one.
Incidentally, just to be clear on this, my understanding is there is no 
EV certificate extension per se. The EV policy OID is just added as 
separate metadata to be associated with the pre-loaded root CA 
certificate; there is nothing in the CA certificate itself that is 
specifically EV-related.
OK, absent any objections my plan is to go ahead and commit this change 
later today or tomorrow. I submitted bug 399214 on October 9, so we've 
had a month to discuss this issue and the proposed policy changes; based 
on the results of those discussions I think we now have consensus on a 
1.1 policy revision, and I want to finish this up so we can proceed to 
invite CAs to submit requests for EV treatment for their CA certs.
Done. The 1.1 release of the Mozilla CA certificate policy is now live 
at the standard URL:
http://www.mozilla.org/projects/security/certs/policy/
Thanks to Eddy, Gerv, and Nelson for providing comments on this 
revision, as well as to all the people who participated in the creation 
of the original 1.0 policy.
Frank
P.S. to Nelson: I haven't forgotten about your suggestion to create a 
companion document to the policy to provide more details about how to 
submit requests, and answer any other questions that might be 
unaddressed by the policy itself. Unfortunately right now I don't have 
time to create such a document; hopefully I'll be able to come back to 
this later in the month.
-- 
Frank Hecker
hec...@mozillafoundation.org