Fwd: GLOBALTRUST 2020's reinclusion in Mozilla's trusted certificates

FYI...

Mailing list post at
<https://groups.google.com/a/mozilla.org/g/dev-security-policy/c/22GoYSOo7NY/m/xLqNxM2VCQAJ>.

---------- Forwarded message ---------
From: Andrew Ayer <[email protected]>
Date: Fri, Jan 10, 2025 at 12:13 PM
Subject: Re: GLOBALTRUST 2020's reinclusion in Mozilla's trusted certificates
To: Mike Benza <[email protected]>
Cc: [email protected] <[email protected]>

Hi Mike,

GLOBALTRUST was never removed from the Mozilla root store.  Rather, it
was tagged with a "Distrust After" date which instructs Firefox to
distrust certificates whose Not Before date is after the root's
Distrust After date.  This is not a security measure (since backdating
certificates is trivial), but rather a mechanism to gracefully sunset
a root so it can be removed without causing problems 398 days later.

However, Curl's mk-ca-bundle.pl script was incorrectly interpreting
the Distrust After date <https://github.com/curl/curl/issues/15547>,
causing GLOBALTRUST to be incorrectly excluded.  Once that bug was
fixed, mk-ca-bundle.pl began emitting GLOBALTRUST again.

There are several reasons why this is unsatisfying.  To begin with,
Mozilla should not be trusting a CA like GLOBALTRUST _at all_, a point
that I and others raised last year
<https://groups.google.com/a/mozilla.org/g/dev-security-policy/c/XpknYMPO8dI/m/j76_U_fMAAAJ>.
Second, root constraints like Distrust After would ideally be
propagated in the PEM bundle through to certificate validators instead
of being dropped by mk-ca-bundle.pl, but there is no widely-supported
mechanism for this at the moment.  For more background, see
https://sslmate.com/blog/post/entrust_distrust_more_disruptive_than_intended

Regards,
Andrew
-- 
Unsubscribe: https://lists.haxx.se/mailman/listinfo/curl-library
Etiquette:   https://curl.se/mail/etiquette.html