On 11/4/2025 6:35 μ.μ., Martijn Katerbarg wrote:
Hi Dimitris,I’m forking this to a new thread to separate from the delayed trust bit removal thread.>Can you please share more details about the risks you see between the following options, for a Root CA with a hierarchy that no longer issues new end-entity certificates (OCSP responder certificates excluded) and all of its subscriber certificates have either been revoked or expired?I want to clarify here that just because the trust bits are being removed from Mozilla/NSS and Chrome it doesn’t mean there won’t be some unexpired leaf certificates, or even new certificate issuance from the hierarchy for a small number of subscriber edge cases.For the sake of this discussion, I’ll leave in the middle whether these edge cases are good or bad. We certainly see both. I also believe the WebPKI is currently in a more active transformational state, one where proper customer education about using the right PKI at the right location is more important than ever.1. Utilizing distrust notAfter/notBefore modern browser methods 2. Removal of "trust bits"3. Removal of the Root CA entirely, especially if there are no "trust bits" enabled.I want to say here that we believe all three mechanisms should be utilized, in the correct order (frankly, the one you specified).
+1
At this moment, number 2 and 3 are utilized by Mozilla for the scheduled deprecations due to CA key lifetimes. The main reason for this is the different removal dates for SMIME and TLS. As an example for a multipurpose root:- 2025-04-15: TLS trust bit removal - 2028-04-15: S/MIME trust bit removal / complete Root CA removal.When looking at single purpose hierarchies, the direct Root CA removal could be the first step.We would like to advocate adding the first option at the beginning of the process, changing the example to:- 2024-04-15: DistrustAfter set for TLS and S/MIME - 2025-04-15: TLS trust bit removal - 2028-05-15: Complete Root CA removalWe believe this approach would help significantly to make root deprecations smoother for CAs and Subscribers alike.What this change would essentially force is setting a single date for both TLS and S/MIME, after which any newly issued certificate won’t be trusted by Mozilla.The issue we see currently is that certificates issued at the same time have different trust removal dates. As an example, with our AAA Certificate Services Root CA:- A TLS certificate issued on 2025-01-15 for 200 days, after 3 months will stop being trusted.- A TLS certificate issued on 2025-02-15 for 30 days, would be trusted for its entire lifetime.- An S/MIME certificate issued on 2026-04-15 for 3 years, would have its trust removed after 2 years.- An S/MIME certificate issued on 2027-04-15 for 1 year, would be trusted for its entire lifetime.We believe (and have noticed in practice) that this easily leads to confusion, whereas having a single “DistrustAfter” date based on the notBefore date for all certificate types allows for more clarity.Additionally, using “DistrustAfter” also has the benefit of showing a non-trusted state immediately after installation of a certificate, rather than (for the Subscriber and Relying Parties perspective) at a random moment.While a CA could obviously choose to halt issuance early on a hierarchy to ensure all leaf certificates expire before the root removal date, that would no longer allow for the sorts of subscriber edge cases that I touched on above.
I assumed all CAs that introduced newer RootCAs in NSS, especially the single-purpose ones, and even more specifically the server TLS cases, would switch issuance to these new hierarchies including a cross-certificate chaining to the previous, more ubiquitous Root. Of course, this tactic increases the bytes in the TLS handshake (more data on the wire), but it's a necessity that comes with the rollover of the Root.
My question is, if the CA has switched issuance of S/MIME and server TLS to other single-purpose hierarchies, like Sectigo, HARICA and others have already done, are there other use cases that might be affected by the entire removal of the Root (option 3) that the ecosystem should be aware of and prevent from breaking?
The updated "ca-certificates" packages in Linux distributions was mentioned, but that would add the new Root the same time as it would remove the old Root, so it shouldn't cause any issues. Is that correct?
Do you, or anyone else, see any other issues that might be caused by the removal of a Root (assuming a new Root is included)?
Thanks, Dimitris.
Regards, Martijn*From: *'Dimitris Zacharopoulos' via dev-security-policy@mozilla.org <dev-security-policy@mozilla.org>*Date: *Thursday, 10 April 2025 at 07:15 *To: *dev-security-policy@mozilla.org <dev-security-policy@mozilla.org>*Subject: *Re: Postponement of Removal of Websites Trust Bit for ePKI Root CAMartijn,Can you please share more details about the risks you see between the following options, for a Root CA with a hierarchy that no longer issues new end-entity certificates (OCSP responder certificates excluded) and all of its subscriber certificates have either been revoked or expired?1. Utilizing distrust notAfter/notBefore modern browser methods 2. Removal of "trust bits" 3. Removal of the Root CA entirely, especially if there are no "trust bits" enabled.I'm very interested to hear the ecosystem risks for each of these choices. It feels that the order is correct in terms of risks but I'm more interested to hear what you and others feel about the 3rd option.Best regards, Dimitris.On 9/4/2025 11:10 μ.μ., 'Martijn Katerbarg' via dev-security-policy@mozilla.org wrote:All, > If Mozilla directly removes the "AAA Certificates Servies" and others, more than 12,435,053 websites issued by "Cloudflare TLS Issuing ECC CA 1" will be affected, It has bad impact on the business of CloudFlare's customers. > The above is what I have found out through about few minutes of research, based on the sites count and I think it will have a gravity impact. > I request the community to conduct an assessment to reduce this impact. As already pointed out by Ryan, these certificates can all be validated through multiple chains. With the experience we’ve gained from preparing for this root certificate removal, we do want to add that we believe future scheduled root certificate deprecations would benefit from utilizing the mechanisms for distrust based on notBefore (Mozilla) and SCTNotAfter (Chrome), rather than a hard deadline for trust bit removal. This probably warrants its own discussion thread though, potentially on the CCADB Public list. > Please consider avoiding the DistrustAfter strategy. It causes problems for tools which use Google, Mozilla (and friends) curated lists of trusted CAs. The tools include utilities like cURL and Wget. We don’t agree with this. The DistrustAfter mechanism is one very suitable for a graceful removal of trust, be it for scheduled deprecations, or other forms of trust removal. The tools mentioned, or more broadly, Linux distributions that build their ca-certificates packages based on the data available in Mozilla/NSS, have hopefully chosen to do so for a reason: They believe the open source root store that Mozilla provides fits their needs and provides the transparency the open source community is usually looking for. If the Mozilla root store believes the DistrustAfter mechanism is viable and is a better option (which we agree with in general), then the parties relying on the root store need to adjust to follow that guidance, or re-assess their needs. They should at no point be holding back innovation and improvements of the Mozilla root store / NSS. We’d like to remind everyone of this Mozilla blog post <https://blog.mozilla.org/security/2021/05/10/beware-of-applications-misusing-root-stores/>, which mentions this wiki page <https://wiki.mozilla.org/CA/Additional_Trust_Changes> that discusses additional factors (including DistrustAfter) that application developers need to be aware of if they are using Mozilla’s root store. Regards, Martijn Katerbarg Sectigo Op woensdag 9 april 2025 om 19:08:43 UTC+2 schreef Ryan Dickson: Hi Arabella, The example provided can validate to multiple <https://crt.sh/?graph=15005443728&opt=nometadata> anchors. For example, an alternate path to an SSL.com root is provided below. Anchor: SSL.com TLS ECC Root CA 2022 <https://crt.sh/?q=c32ffd9f46f936d16c3673990959434b9ad60aafbb9e7cf33654f144cc1ba143> ---> SSL.com TLS Transit ECC CA R2 <https://crt.sh/?q=5d1bc399274e649e1c72697de91a54ad725088c5221cb61e17ee9c290bc42a92> ---> Cloudflare TLS Issuing ECC CA 1 <https://crt.sh/?q=2964fd3210ea68faa2b4a849b36243d33f74429d1b43ce019e7b154eac7759ba> ---> www.relialabtest.com <https://crt.sh/?q=133f15fc8303dccb6b319b65c6d9f2ff9ae1c0fea4abf2eaf70939d77240dc0a> Hope this helps! - Ryan On Wed, Apr 9, 2025 at 12:40 PM Arabella Barks <arabel...@gmail.com> wrote: Sorry, I forgot to post one case https://www.relialabtest.com it is the hierarchy I mentioned. On Thursday, April 10, 2025 at 12:36:03 AM UTC+8 Arabella Barks wrote: Ben, I still suggest adopting the distrust-after. Among the root intermediates that Mozilla plans to remove trust, there is the "AAA Certificates Servies" of Sectigo CA, which is being widely issued and used by a subordinate CA of Cloudflare, namely "Cloudflare TLS Issuing ECC CA 1" (https://crt.sh/?caid=282054, and issued by "SSL.com TLS Transit ECC CA R2"). However, SSL.com TLS Transit ECC CA R2 is just a subordinate CA and not a Root. If Mozilla directly removes the "AAA Certificates Servies" and others, more than 12,435,053 websites issued by "Cloudflare TLS Issuing ECC CA 1" will be affected, It has bad impact on the business of CloudFlare's customers. The above is what I have found out through about few minutes of research, based on the sites count and I think it will have a gravity impact. I request the community to conduct an assessment to reduce this impact. On Thursday, April 10, 2025 at 12:10:21 AM UTC+8 Ben Wilson wrote: Thanks for your feedback. Currently, our proposed strategy for handling this particular issue will be to postpone processing the websites trust bit removal from the Chunghwa Telecom ePKI Root CA by two or three months (until approximately Firefox Release 141 <https://whattrainisitnow.com/release/?version=141>). In other words, we do not anticipate using the distrust-after mechanism in the present case. Thanks again, Ben On Wed, Apr 9, 2025 at 9:55 AM Jeffrey Walton <nolo...@gmail.com> wrote: On Tue, Apr 1, 2025 at 11:03 AM 'Ben Wilson' via dev-secur...@mozilla.org <dev-secur...@mozilla.org> wrote: > > Per - https://bugzilla.mozilla.org/show_bug.cgi?id=1891438#c15: > > "In the interest of transparency, Mozilla received a formal request from Taiwan’s Ministry of Digital Affairs (MODA), dated March 15, 2025, requesting that we delay the removal of the “websites” trust bit for Chunghwa Telecom’s ePKI Root CA, which is currently scheduled to occur on or about April 15, 2025, in accordance with Mozilla’s Root CA Lifecycles Transition Schedule. > > MODA explained that the requested delay is intended to support the ongoing transition of government websites away from certificates issued by CHT’s GTLSCA-G1 subordinate CA. As we understand it, MODA is already implementing a short-term migration plan involving the dual issuance of approximately 12,000 new certificates for government websites—one from Chunghwa Telecom and one from Taiwan CA (TWCA)—to ensure continued availability of government services and minimize user disruption. > > While we have not yet finalized a decision, we are currently contemplating: > > Postponing the removal of the “websites” trust bit; > Implementing a distrust-after date; or > Taking other actions consistent with Mozilla Root Store Policy and ecosystem risk management. > > We note that: > > The ePKI Root CA uses a 4096-bit RSA key, which provides stronger security than other similarly aged root certificates. > Any extension under consideration would be strictly time-bounded (e.g., not to exceed August 1, 2025), reflecting a short-term accommodation, not a change in long-term policy direction. > Mozilla would retain the right to remove or revoke trust at any time, based on new information or evolving risk factors. > > We welcome feedback on any of these approaches." Please consider avoiding the DistrustAfter strategy. It causes problems for tools which use Google, Mozilla (and friends) curated lists of trusted CAs. The tools include utilities like cURL and Wget. See, for example, <https://github.com/curl/curl/issues/15547>. Jeff--You received this message because you are subscribed to the Google Groups "dev-secur...@mozilla.org" group. To unsubscribe from this group and stop receiving emails from it, send an email to dev-security-po...@mozilla.org. To view this discussion visit https://groups.google.com/a/mozilla.org/d/msgid/dev-security-policy/c0794245-c1c8-417c-a40e-a7154a4720d2n%40mozilla.org <https://groups.google.com/a/mozilla.org/d/msgid/dev-security-policy/c0794245-c1c8-417c-a40e-a7154a4720d2n%40mozilla.org?utm_medium=email&utm_source=footer>.-- You received this message because you are subscribed to the GoogleGroups "dev-security-policy@mozilla.org" <mailto:dev-security-policy@mozilla.org> group. To unsubscribe from this group and stop receiving emails from it, send an email to dev-security-policy+unsubscr...@mozilla.org. To view this discussion visit https://groups.google.com/a/mozilla.org/d/msgid/dev-security-policy/136da9cf-e967-401c-9cc9-d2031655d605n%40mozilla.org <https://groups.google.com/a/mozilla.org/d/msgid/dev-security-policy/136da9cf-e967-401c-9cc9-d2031655d605n%40mozilla.org?utm_medium=email&utm_source=footer>. --You received this message because you are subscribed to a topic in the Google Groups "dev-security-policy@mozilla.org" group. To unsubscribe from this topic, visit https://groups.google.com/a/mozilla.org/d/topic/dev-security-policy/uYAm_c_pfos/unsubscribe. To unsubscribe from this group and all its topics, send an email to dev-security-policy+unsubscr...@mozilla.org. To view this discussion visit https://groups.google.com/a/mozilla.org/d/msgid/dev-security-policy/c7ddab01-7145-4b87-a2c0-5532eca6675b%40it.auth.gr <https://groups.google.com/a/mozilla.org/d/msgid/dev-security-policy/c7ddab01-7145-4b87-a2c0-5532eca6675b%40it.auth.gr?utm_medium=email&utm_source=footer>.
-- You received this message because you are subscribed to the Google Groups "dev-security-policy@mozilla.org" group. To unsubscribe from this group and stop receiving emails from it, send an email to dev-security-policy+unsubscr...@mozilla.org. To view this discussion visit https://groups.google.com/a/mozilla.org/d/msgid/dev-security-policy/9d0e8278-d389-4cef-8a6d-04db402a76f1%40it.auth.gr.
smime.p7s
Description: S/MIME Cryptographic Signature
