On 21/01/15 18:49, Stephen Kent wrote:
Rob,
Hi Steve.
That seems like a good idea. Did you have any particular DNS experts
in mind?
I'd ask Joel Jaeggli <[email protected]> for suggestions.
Thanks. I'll contact Joel.
if not, why state this as a requirement?
I don't think the current text does state this as a requirement.
Your message to me cited this as a reason for not using "*" in this
context. If the different semantics for PRIVATE/? are not a requirement,
that argument isn't applicable. Color me confused.
IIUC, you're saying that since "*" and "PRIVATE/?" are intended to have
different semantics, 6962-bis needs to specify the required semantics
more explicitly than it currently does. Have I got that right?
I think we could address this by adding text to 6962-bis to specify that
CAs MUST NOT use "PRIVATE/?" in a Precertificate domain label if the
corresponding domain label in the Certificate is "*".
If we want TLS Clients to enforce this rule, I think the appropriate
place for that will be the envisaged client behaviour document (rather
than 6962-bis).
I've just created these two tickets:
http://trac.tools.ietf.org/wg/trans/trac/ticket/56
http://trac.tools.ietf.org/wg/trans/trac/ticket/57
If there are multiple certs (which may or may not contain different
unredacted domain label(s)) that match one redacted precert, then
those certs will all share the same Serial Number and be issued by the
same CA.
RFC5280 already disallows this, so 6962-bis doesn't also need to
disallow it. :-)
Can you cite the part of 5280 that you feel is applicable here? 4.1.2.2
notes that a cert serial number uniquely identifies a cert, which doesn't
seem compatible with your comment.
I don't see any incompatibility.
Your question was...
"Do you envision these clients remembering every cert that they
encounter that matches a redacted pre-cert, and check to see if a CA has
issue two certs with different values for a redacted DNS label?"
Even issuing two certs with the same values for a redacted DNS label
would be wrong.
If a client is checking for RFC5280 4.1.2.2 serial number uniqueness,
then they are implicitly checking for multiple certs that match a single
Precertificate.
A Precertificate contains Issuer Name and Serial Number. Every
Certificate that matches that particular Precertificate will, by
definition, contain the same Issuer Name and Serial Number. So, if
there is >1 matching Certificate then RFC5280 Section 4.1.2.2 has been
violated.
Hmmm...
Should we specify that the redaction label (i.e. "(PRIVATE)" or "?" or
whatever the DNS experts prefer) MUST NOT match "*" in the cert?
I can't think of a legitimate reason to hide wildcard certs from the log.
I agree that wildcard certs should be logged.
Great. I think the two new tickets I mentioned above are sufficient.
Steve
--
Rob Stradling
Senior Research & Development Scientist
COMODO - Creating Trust Online
_______________________________________________
Trans mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/trans
