[lamps] Re: [TLS] TLS Client Certificates; a survey

[Viktor Dukhovni <[e-mail]>]

On Thu, Mar 26, 2026 at 03:14:49PM +0900, Seo Suchan wrote:

> Wonder if allowing tls root ca to have dedicate clientauth eku only
> (no serverauth) subca is good stop gap. So clientauth cert can't
> backflow to browsers, but allowing clientauth certs to be chained to
> tls root cas

That violates the CRP (Chrome Root Programme) rules as I understand
them, but what is possible is for the CA to not enroll the real root
in the CRP, and instead enroll only subsidiary CAs that don't issue
clientAuth:

        root -> intermediate CRP trust anchor  -> EE cert
         |
         \__--> non-CRP intermediate -> EE cert

but this would require the various popular CAs to be willing to make the
effort to support clientAuth by doing that (thus coördinating with the
CRP before introducing new intermediates).

And of course some servers might need separate certificates if they
need to be valid for both browser clients and in server-to-server
contexts as clients.

-- 
    Viktor.  🇺🇦 Слава Україні!