[cabf_validation] [EXTERNAL] Re: RFC 5280 conflict for SKI in subscriber certificates

Mon Dec 5 18:26:22 UTC 2022

Let's suppose that the SKID is always computed as a SHA1 hash of the public
key info. Then the value of the SKID extension will be 20 bytes, and the
full extension itself (including its OID and length headers) will be 31
bytes of DER.

It's hard to get an estimate of how many SSL handshakes happen globally
every day, so rather than talking in total numbers let's talk in
percentages. A randomly-selected <https://crt.sh/?asn1=8130021554> Let's
Encrypt DV cert for a ~12-character dnsName and a 2048-bit RSA public key
is ~1100 bytes. The SKID extension is nearly 3% of that total. Saving 3% of
the bytes -- or even 1.5% if you include the intermediate that is being
sent at the same time -- of every TLS handshake is huge. Lots of companies
would love to have such a straightforward way to save 3% on their core cost
metrics.

If you assume SKIDs are a SHA256 hash instead (because no one wants to be
running SHA1 code anymore), then nearly double those numbers. If you assume
that chain has ECDSA public keys and signatures, nearly double them again.

I think that saving this large of a percentage of handshake size *is* a
good reason to deviate from RFC 5280.

Aaron

On Mon, Dec 5, 2022 at 7:44 AM Jeremy Rowley <jeremy.rowley at digicert.com>
wrote:

> +1. We really shouldn’t deviate from 5280 unless there is a good reason
> to. Even if SKI has minimal use, both SHOULD and SHOULD NOT are “optional”
> under CAB Forum so we aren’t limiting anyone by matching the RFC language.
>
>
>
> *From:* Validation <validation-bounces at cabforum.org> * On Behalf Of *Tomas
> Gustavsson via Validation
> *Sent:* Monday, December 5, 2022 1:11 AM
> *To:* Aaron Gable <aaron at letsencrypt.org>; Paul van Brouwershaven <
> Paul.vanBrouwershaven at entrust.com>; Lahtiharju, Pekka <
> pekka.lahtiharju at teliacompany.com>; CABforum3 <validation at cabforum.org>
> *Subject:* Re: [cabf_validation] [EXTERNAL] Re: RFC 5280 conflict for SKI
> in subscriber certificates
>
>
>
> When suggesting changes to RFC5280, the argument against it is always that
> it will most certainly break the internet, due to the amount of different
> software out there. We've seen issues in the past due to BRs specifying
> things differently than RFC5280. With that I want to say there is
> definitely a risk in breaking the connection with RFC5280, a risk that is
> not easy to see in the short term. Going astray from RFC5280 should imho
> only be considered when there is a very significant advantage. Since it is
> only a SHOULD in rfc5280, I find it hard to see the big advantage that
> outweighs the risk (even if the risk is considered minor).
>
>
>
> Cheers,
>
> Tomas
>
>
> ------------------------------
>
> *From:* Validation <validation-bounces at cabforum.org> on behalf of
> Lahtiharju, Pekka via Validation <validation at cabforum.org>
> *Sent:* Friday, December 2, 2022 7:53 AM
> *To:* Aaron Gable <aaron at letsencrypt.org>; Paul van Brouwershaven <
> Paul.vanBrouwershaven at entrust.com>
> *Cc:* CA/Browser Forum Validation SC List <validation at cabforum.org>
> *Subject:* Re: [cabf_validation] [EXTERNAL] Re: RFC 5280 conflict for SKI
> in subscriber certificates
>
>
>
> *CAUTION:* External Sender - Be cautious when clicking links or opening
> attachments. Please email InfoSec at keyfactor.com with any questions.
>
>
>
> CA can use SKI for internal purposes to help finding their own
> certificates with the same public key. For example, the new Key Compromise
> revocation reason code specification requires CA to revoke all certificates
> using the same key. Our own certificates use always the same algorithm for
> SKI so we can be sure that search result is correct. We haven’t used it to
> search external certificates.
>
>
>
> Also the CA software we are using currently always add SKI to all
> certificates; there is no option to not use it.
>
>
>
> Pekka
>
>
>
> *From:* Aaron Gable <aaron at letsencrypt.org>
> *Sent:* torstai 1. joulukuuta 2022 20.42
> *To:* Paul van Brouwershaven <Paul.vanBrouwershaven at entrust.com>
> *Cc:* Hubert Chao <hchao at google.com>; Lahtiharju, Pekka <
> pekka.lahtiharju at teliacompany.com>; CA/Browser Forum Validation SC List <
> validation at cabforum.org>; Corey Bonnell <Corey.Bonnell at digicert.com>
> *Subject:* Re: [EXTERNAL] Re: [cabf_validation] RFC 5280 conflict for SKI
> in subscriber certificates
>
>
>
> If you're searching for certificates with the same key, the SKID can
> easily lead you astray -- there's no requirement that two different CAs use
> the same derivation function to compute the SKID from the Public Key. The
> SKID is useful in CA certs because it is required to byte-for-byte match
> the AKID in issued certs. I don't believe the SKID in end-entity certs
> serves any purpose in the modern webpki.
>
>
>
> I'd love to hear more from Corey and/or Ryan Sleevi on the original
> motivation for this from July 2021, in case I'm missing something, but
> obviously I'm convinced already :)
>
>
>
> Aaron
>
>
>
> On Thu, Dec 1, 2022 at 7:18 AM Paul van Brouwershaven <
> Paul.vanBrouwershaven at entrust.com> wrote:
>
> The SKI is useful to quickly search for certificates with the same key.
>
>
>
> Is saving a few bytes a sufficient reason to 'deviate' from RFC 5280,
> where we try to get everyone to focus on RFC 5280 adherence at the same
> time?
>
>
>
> Are we sure that this would not cause any client incompatibility issues?
> Almost all certificates include the SKI today and while this might be
> fine for the major browsers, we also know that there are other
> clients/libraries that interact with web websites.
>
>
>
> Paul
>
>
> ------------------------------
>
> *From:* Hubert Chao <hchao at google.com>
> *Sent:* Thursday, December 1, 2022 15:59
> *To:* Lahtiharju, Pekka <pekka.lahtiharju at teliacompany.com>; CA/Browser
> Forum Validation SC List <validation at cabforum.org>
> *Cc:* Aaron Gable <aaron at letsencrypt.org>; Paul van Brouwershaven <
> Paul.vanBrouwershaven at entrust.com>
> *Subject:* [EXTERNAL] Re: [cabf_validation] RFC 5280 conflict for SKI in
> subscriber certificates
>
>
>
> WARNING: This email originated outside of Entrust.
> DO NOT CLICK links or attachments unless you trust the sender and know the
> content is safe.
> ------------------------------
>
> On Thu, Dec 1, 2022 at 5:21 AM Lahtiharju, Pekka via Validation <
> validation at cabforum.org> wrote:
>
> I support Paul’s idea to change this to SHOULD. Why should we create new
> recommendations against RFC when this extension is useful in several use
> cases and almost everybody is using it now.
>
>
>
> Could you list out the use cases where this extension is useful for a TLS
> certificate? The discussion that Corey linked to (
> https://lists.cabforum.org/pipermail/validation/2021-July/001672.html
> <https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Furldefense.com%2Fv3%2F__https%3A%2Flists.cabforum.org%2Fpipermail%2Fvalidation%2F2021-July%2F001672.html__%3B!!FJ-Y8qCqXTj2!bhb6QGSEpqEOi6JyHDzixLHA_ziEpOs6UQYkMiffRA4PH_9fFgyIiZRW3epCZqq0_V5K5pDehK6XTaH3PNBz1ibt%24&data=05%7C01%7Ctomas.gustavsson%40primekey.com%7Ca77bbc1c0b7840599a3e08dad431ea26%7Cc9ed4b459f70418aaa58f04c80848ca9%7C0%7C0%7C638055608160945478%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=LMcp1hZgr1RctTVx2LAf16ewcktRHuAP9fyF2irydpI%3D&reserved=0>)
> specifically says "... a TLS certificate [SKI] should not be needed ... ".
>
>
>
> /hubert
>
> *Any email and files/attachments transmitted with it are confidential and
> are intended solely for the use of the individual or entity to whom they
> are addressed. If this message has been sent to you in error, you must not
> copy, distribute or disclose of the information it contains. Please notify
> Entrust immediately and delete the message from your system.*
>
>
>
>
> *This email may contain information which is privileged or protected
> against unauthorized disclosure or communication. If you are not the
> intended recipient, please notify the sender and delete this message and
> any attachments from your system without producing, distributing or
> retaining copies thereof or disclosing its contents to any other person.
> Telia Company processes emails and other files that may contain personal
> data in accordance with Telia Company’s Privacy Policy
> <https://nam11.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.teliacompany.com%2Fen%2Fabout-the-company%2Fprivacy%2F&data=05%7C01%7Ctomas.gustavsson%40primekey.com%7Ca77bbc1c0b7840599a3e08dad431ea26%7Cc9ed4b459f70418aaa58f04c80848ca9%7C0%7C0%7C638055608160945478%7CUnknown%7CTWFpbGZsb3d8eyJWIjoiMC4wLjAwMDAiLCJQIjoiV2luMzIiLCJBTiI6Ik1haWwiLCJXVCI6Mn0%3D%7C3000%7C%7C%7C&sdata=b%2BiqCDqfMnfryWWqRLubBULI4Vv5SRTSvuWTm4%2F0Xzw%3D&reserved=0>.*
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.cabforum.org/pipermail/validation/attachments/20221205/38376240/attachment-0001.html>