Steve,

As I mentioned, I think log is a good place to ensure that hygiene of the path 
and not just the leaf certificate.

IMPLEMENTATION ISSUES
I do not know if an Appendix would help log operators with some of the past 
erroneous implementations or interpretations of RFC 5280.  I mention this if 
the log operators choose to use some of the existing platforms to implement 
path validation.  Newer versions may or may not have fixed some of these issues 
such as:

1.  Not enforcing name constraints on intermediate certificates in the path
2.  Interpreting name constraints as the only name forms allowed and rejecting 
certificates with other name forms.  For example, if the name constraints 
contain DN based names and SAN contain RFC 822 names, rejecting the 
certificates erroneously.
3. Misapplying name constraints on URI.  This has caused real problems in 
circles I work in.  Since not all URI schemes have authority component, the URI 
name constraint should not be applied to the ones that do not (e.g., even if 
URI name constraints are present, they do not apply to URN and certificates 
should not be rejected.

EKU PROCESSING
I went back to the CAB Forum document and it has no processing rules for EKU 
for the path.  But, based on their desire to constrain the path how some of the 
more popular implementation has done it, the following logic seems to make 
sense.

Assume n certificates in the certificate chain. Set the value of EKU for a 
certificate to Universal Set if the extension is absent or contains 
anyExtendedKeyUsage; otherwise set the value to the set of OIDs in the EKU 
extension.  Now, take the intersection of all n sets.  If the intersection is 
empty, the path validation fails.  If the intersection in Universal set, the 
end certificate can be used for any purpose.  Otherwise, the end certificate 
can be used for purpose(s) in the intersection.

BTW, I am confused by what the CABF document says in Appendix B item:
"Generally Extended Key Usage will only appear within end entity certificates 
(as highlighted in RFC 5280 (4.2.1.12)), however, Subordinate CAs MAY include 
the extension to further protect relying parties until the use of the extension 
is consistent between Application Software Suppliers whose software is used by 
a substantial portion of Relying Parties worldwide"

I only bring it up because 6962 seems to be using that as a baseline.

-----Original Message-----
From: Trans [mailto:[email protected]] On Behalf Of Stephen Kent
Sent: Monday, September 29, 2014 10:42 AM
To: [email protected]
Subject: [Trans] path validation

I realized, after I sent my reply, that Santosh's point wrt path validation is 
important. I believe Ben indicated that he did not envision logs performing all 
of the 5280 checks, just verifying the signatures in the path. So, I misspoke 
when I noted that 6962-bis already called for path checking; it calls for it, 
but only in a very superficial way.

Thus we need to decide if we believe that a sig-only check on a path consistent 
with the CABF guidelines. I don't think it is. The Definitions section of the 
CABF guidelines includes the following entry:

     Valid Certificate: A Certificate that passes the validation procedure 
specified in RFC 5280.

Steve

_______________________________________________
Trans mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/trans

_______________________________________________
Trans mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/trans

Reply via email to