[cabfpub] Definition of "Browser" in the Bylaws
haavardm at opera.com
Fri Oct 11 08:29:58 UTC 2013
On 10/10/2013 10:40 PM, Tom Albertson wrote:
> Maybe we should start from the definition of "Trust Anchor Managers"
> in the NIST IR, or the equally good TAM definition from ETSI which I
> don't have handy at the moment. TAM is a pretty universal definition
> for someone who operates a trust root repository, and I believe it
> should extend to new members and categories of members we can expect
> such as Oracle. It may be a little wordy and detailed (the first
> sentence seems to say it all), but better to converge on agreed
> terminology rather than define new ones. Can anyone read this
> definition not recognize current members, or potential future members?
> /Trust Anchor Managers (TAMs): /Authorities who manage a repository of
> trusted Root CA Certificates. They act on behalf of relying parties,
> basing their decisions on which CAs to trust on the results of
> compliance audits. A TAM sets requirements for inclusion of a CA's
> root public key in their store. These requirements are based on both
> security and business needs. The TAM has a duty to enforce compliance
> with these requirements, for example, requirements around the supply
> of compliance audit results, on initial acceptance of a root, and on
> an ongoing basis. TAMs will follow their normal practice of requiring
> CAs to submit an annual compliance audit report. It is our intention
> that the requirements in this document will be included in those
> compliance audit schemes. As specified in Section 5.7, the TAM will
> require the CA to provide notification of a compromise, and in
> response, the TAM will take appropriate action.
> The one problem here is the relatively recent situation with Opera
> adopting the Google trusted root store, and I feel obligated to ask:
> will Opera in fact maintain independent decision-making with respect
> to the root store for Opera users? I know Opera will maintain their
> legacy root program for legacy version of their browser, but I haven't
> heard (can't recall from Turkey) the status of the program for current
> and future versions. Sig, can you confirm?
Opera has not adopted Google's trust store. We, just like chromium, use
the rootstore on the OS platform. I'm not aware of google running any
rootstore (perhaps on Android?), so it seems Google and Opera are in the
same boat here.
Also, even though we don't run a trust store, we still take relevant
security decisions in the ssl stack and the UI level.
> Disclaimer: No WAY am I trying to toss out Opera or any other CABF
> membership via a definition change. Convergence though has its issues
> and we should sort them out. We allocate one vote per CA regardless
> of how many CA brands they represent; I don't think we want to
> allocate voting among browsers if the members behind those votes don't
> exercise independent beneficial control of their root store(s).
> Apologies in advance to Opera folks everywhere, I've tried to raise
> the issue as tactfully and sensitively as I can. Personally I hope
> they come back and confirm they can meet the guideline options under
> discussion. This is about the time when someone should haul out the
> picture of me wearing the Opera t-shirt during a past CABF face to
> face meeting.... J
> All the best, Tom
> *From:*public-bounces at cabforum.org
> [mailto:public-bounces at cabforum.org] *On Behalf Of *Ryan Sleevi
> *Sent:* Thursday, October 10, 2013 12:46 PM
> *To:* ben at digicert.com
> *Cc:* CABFPub
> *Subject:* Re: [cabfpub] Definition of "Browser" in the Bylaws
> Regrets that I was not able to make the call this morning. I have a
> few questions inline.
> On Thu, Oct 10, 2013 at 12:28 PM, Ben Wilson <ben at digicert.com
> <mailto:ben at digicert.com>> wrote:
> During today's call we discussed the definition of "browser" in
> the bylaws. This is relevant due to Oracle's potential interest
> in joining the Forum. Back in 2009 we were looking at revising the
> browser category of membership. I have pasted excerpts from that
> discussion below. Today we had a similar discussion, which ended
> in a suggestion that we start this thread on the list.
> The current bylaws define a "browser" member as "[an organization
> that] produces a software product intended for use by the general
> public for browsing the Web securely."
> One suggestion is that we amend the definition in two ways -- (1)
> adding the concept of "trust agency" and (2) broadening the
> definition beyond just "browsing the web".
> I'm reasonably concerned about the second definition.
> We've obviously seen that there are a set of security decisions being
> made for products that operate on the Internet, but are not meant for
> browsing the web. These products and decisions can have significant
> negative impact on overall user security. For example, note the
> challenges and objections raised by several members with regards to
> deprecating internal server names or 1024-bit RSA keys.
> While I recognize the need for CAs to be able to engage their customer
> base, I would like to see our group be able to focus on the broader
> public's needs through web browsers, and let those flow into other
> product lines as appropriate.
> That is, if Payment Company 1 makes terminals that only support
> 1024-bit RSA keys (or MD5), I would not want to see them able to hold
> the Internet's security hostage.
> For example, the definition could be revised:
> Browser/User Trust Agent ("Browser"): The member organization
> produces a software product widely used by the general public that
> uses a root store for making trust decisions for relying parties."
> During the conversation, we noted that we're not talking about
> root store operators, but systems, operating platforms, and browsers.
> Also, to what degree should we consider the size of the applicant,
> if there were several small ones that wanted to join and become
> voting members, that might have an impact, and since this is not
> trivial, we are opening up this discussion on the list.
> One draft from 2009 was as follows: "*Product Supplier* - The
> organization supplies a product intended for use by the general
> public that acts as a PKI relying party (making trust decisions on
> behalf of the user by actually processing certificates with
> software) to secure information or transactions."
> Here are the notes of our meeting May 14, 2009:
> The voting category [would be] expanded beyond Web browsers to any
> product that acts as a PKI relying party.
> Would the author of a Web browser with two users be eligible for
> membership? Tim said that it would.
> Nick said that, because we have so few members in this category,
> someone with very little market presence could have
> disproportionate influence on the ballots.
> Johnathan said that he would like to be able to apply common sense
> on a case-by-case basis.
> Tim said that we have to ensure that our criteria are objective.
> Johnathan said that the proposed criteria do not distinguish
> between software that curates its own root store and software that
> relies entirely on another's root store. There are a lot of
> browsers that are extensions on Mozilla. They use the same root
> store and make the same EV decisions. If they were all to join,
> then Mozilla may be over-represented.
> Tim suggested that it should depend upon whether their
> constituents viewed them as their agent in questions of trust.
> Johnathan agreed.
> Aaron said that there are browsers that use the Apple framework to
> make trust decisions and that should not be enough to disqualify
> them; their input as browsers is still very important.
> Johnathan said that there is a case to be made that, if a piece of
> software has meaningfully different trust characteristics, then
> that is closer to a rational test.
> Yngve said that we have encountered this situation with Google. He
> wasn't sure whether Google relies on its own root store or the
> platform's root store. Johnathan said that, without speaking for
> Chrome, his understanding was that they use the Microsoft root
> store on windows, but make their own EV determination, and that it
> is their intention to use 'platform native' cert stores on Mac and
> Linux as well.
> While I realize these notes were from 2009, I do want to chime in that
> we do ship a set of trust roots for Android, Chrome for iOS, ChromeOS,
> as well as other consumer products such as the Google Chromecast.
> Additionally, as noted in our root store policy, we reserve the right
> to remove any certificate from being trusted, even if trusted by the
> underlying OS root store, if we believe the inclusion impacts the
> security of our users. See
> As such, I would presumably see Opera in a similar situation as
> Google, in which the products they ship may derive their primary trust
> value from the decisions made as part of the Chromium projects, but
> for which they reserve the right (organizationally) to remove or
> modify such trusts as appropriate to the security of their users.
> Bjørn said that he would not want to accept anyone who is simply
> 'skinning' Internet Explorer, for instance.
> Tim said that the wording we have is 'acts as a relying party'.
> Bjørn said that it is slightly vague. Yngve suggested 'acting as a
> PKI relying party, making trust decisions on behalf of the user'.
> Bjørn agreed. Tim said that he would incorporate that clause, and
> if there were no objections he would take it to ballot.
> Public mailing list
> Public at cabforum.org <mailto:Public at cabforum.org>
> Public mailing list
> Public at cabforum.org
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the Public