W3C home > Mailing lists > Public > ietf-http-wg@w3.org > October to December 2016

Re: [expect-ct] Is expect-ct policy intended for long-term use? (plus: no user recourse)

From: Emily Stark <estark@google.com>
Date: Mon, 28 Nov 2016 14:25:14 -0800
Message-ID: <CAPP_2SZ+7Te5Piyw7RGbqR-ORSiON+MD1wTZDetteKOYL93Q9A@mail.gmail.com>
To: "=JeffH" <Jeff.Hodges@kingsmountain.com>
Cc: IETF HTTP WG <ietf-http-wg@w3.org>
On Thu, Nov 24, 2016 at 6:00 PM, =JeffH <Jeff.Hodges@kingsmountain.com>
wrote:

> Emily wrote:
> > I anticipate Expect-CT to be useful more than a year and less than 5
> > years. Within 1-2 years, I expect/hope several browsers will be
> > requiring CT for all new certificates. They can still implement
> > Expect-CT to protect sites against backdating and against
> > certificates that were issued before the date that they started
> > requiring CT for all new certs.
>
> ok, by "they" you mean UAs, yes?
>

Yes. To clarify, this is what I meant: "A UA can still implement Expect-CT
to allow sites to protect themselves against backdating and against
certificates that were issued before the date that the UA started requiring
CT for all new certs."


>
>
> > Once a browser is requiring CT for *all* certificates (e.g. because
> > the maximum validity period has elapsed beyond the date that the
> > browser began requiring CT for all new certs), then I don't think
> > Expect-CT is useful for that browser anymore.
>
> by implication you mean "useful" for a server (aka "relying party" (RP))
> and user, yes?
>
> because what we are protecting here is not so much the browser (vendor)
> but the RP and user, yes?
>

Yes. I meant that once a browser is requiring CT for all certs, then the
users who are using that browser to contact an Expect-CT server are not
getting much benefit from Expect-CT.


>
> I could see Expect-CT to be useful for the longer term if it were to
> signal additional RP-desired selective UA behavior such as "no user
> recourse", *if* the browsers were not going to implement such behavior,
> e.g., as a a matter of course in the case of errors during secure
> connection establishment.
>
> =JeffH
>
>
>
> > On Wed, Nov 23, 2016 at 4:47 PM, =JeffH
> > <Jeff.Hodges@kingsmountain.com> wrote:
> > WRT "Expect-CT"
> > <https://tools.ietf.org/html/draft-stark-expect-ct>
> > (aka "the I-D" in the below)...
> >
> > Is the expect-ct policy intended to be used long-term by servers?
> >
> > I.e., is this server-declared expect-ct policy only a stop-gap until
> > all browsers natively enforce their vendors' "ct policies"?
> >
> > At first glance, it seems the answer is "yes, expect-ct has long-term
> > usefulness" given the language in
> > <https://tools.ietf.org/html/draft-stark-expect-ct-00#section-2.1.2>,
> >
> > i.e., a host's declaration of expect-ct policy is stating that the UA
> > must terminate any connection to that host (and port?) that does not
> > satisfy the UA's ct policy.
> >
> > However, given this..
> >
> > On Sunday, November 13, 2016 at 4:47 AM, Emily Stark wrote:
> >> That is, eventually, when browsers require CT for all
> >> certificates, [...] I see Expect-CT as a way that individual sites
> >> can opt in to the future early ("the future" being when browsers
> >> require CT for all certificates)
> >
> > ..it sounds like the browsers intend to do that in any case, and if
> > so, on what timescale?
> >
> > I.e., is it worthwhile to go through all the work to formally define
> > Expect-CT in an RFC?
> >
> > I'm not sure. This is part of the reason why I uploaded this as an
> > experimental draft. I'm not 100% sure what's the right process or
> > venue is for a mechanism that is not meant to stick around forever.
> >
> >
> > Though, if there is some functionality that a server-declared
> > expect-ct policy stipulates that is not intended to be implemented by
> > default in near- to intermediate-term, then formally specifying
> > Expect-CT perhaps has a reasonable cost-benefit regardless. Or also
> > if explicit server-declared "expect-ct" policy would be useful to the
> > long-tail of HTTPS clients other than the dominant browsers.
> >
> > Perhaps one should consider having the expect-ct policy additionally
> > mean that there is "no user recourse" to connection termination as a
> > result of CT-policy violation. I note the I-D does not presently
> > state that.
> >
> > See <https://tools.ietf.org/html/rfc6797#section-12.1> for how this
> > is discussed in HSTS. You might consider adding "no user recourse" to
> > a "UA implementation advice" section.
> >
> > That seems reasonable to include, though I don't think "no user
> > recourse" is enough benefit to justify keeping Expect-CT around after
> > it has otherwise exhausted its usefulness.
> >
> >
> > Though, like any of this (including HSTS), the browsers could in the
> > future decide that they will have a "no user recourse" policy by
> > default for all secure transport establishment failures. It's a
> > question of how far in the future might that occur (in order to
> > justify present-to-intermediate-term work).
> >
> > HTH,
> >
> > =JeffH
> >
> >
> >
>
>
Received on Monday, 28 November 2016 22:26:13 UTC

This archive was generated by hypermail 2.3.1 : Monday, 28 November 2016 22:26:16 UTC